# Sustainable Computing

Sustainable computing is the field of designing, using, and disposing of computing hardware and software to minimize environmental impacts and energy consumption across the information technology lifecycle.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Training for sustainable scientific software in the AI era: from code production to scientific stewardship

DevFeed: [Training for sustainable scientific software in the AI era: from code production to scientific stewardship](<https://devfeed.tech/articles/training-for-sustainable-scientific-software-in-the-ai-era-from-code-production-to-scientific-stewardship-142177.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23330>)

Author: Juliette Alimena

Published: 2026-11-19T05:00:00Z

Content type: opinion

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [reproducibility](<https://devfeed.tech/topics/reproducibility.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>)

Tags: [agentic-artificial-intelligence](<https://devfeed.tech/tags/agentic-artificial-intelligence.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-assisted](<https://devfeed.tech/tags/ai-assisted.md>), [ai-assisted-workflows](<https://devfeed.tech/tags/ai-assisted-workflows.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [challenge](<https://devfeed.tech/tags/challenge.md>), [code](<https://devfeed.tech/tags/code.md>), [code-generation](<https://devfeed.tech/tags/code-generation.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [hep-ex](<https://devfeed.tech/tags/hep-ex.md>), [physics](<https://devfeed.tech/tags/physics.md>), [privileged-access](<https://devfeed.tech/tags/privileged-access.md>), [productivity](<https://devfeed.tech/tags/productivity.md>), [provenance](<https://devfeed.tech/tags/provenance.md>), [reproducibility](<https://devfeed.tech/tags/reproducibility.md>), [resource](<https://devfeed.tech/tags/resource.md>), [resources](<https://devfeed.tech/tags/resources.md>), [software](<https://devfeed.tech/tags/software.md>), [training](<https://devfeed.tech/tags/training.md>), [workflows](<https://devfeed.tech/tags/workflows.md>)

### AI overview

For high-energy physics, AI-assisted software training should prioritize scientific stewardship alongside code production. Because AI-generated or modified workflows can weaken researchers' command of the choices and evidence behind a result, training should combine dedicated AI instruction with established software and analysis training. Researchers should practice testing, tracking provenance, reproducing and maintaining workflows, judging resource use, and validating results. Competence should be assessed by the evidence supporting scientific results, not code generation alone. The abstract frames sustainability as keeping workflows trustworthy and maintainable, using resources responsibly, and avoiding dependence on privileged access to particular tools or infrastructure.

### Source excerpt

arXiv:2611.23330v1 Announce Type: new Abstract: Generative and agentic artificial intelligence (AI) systems are changing how researchers interact with scientific software. Their effects on productivity vary across tasks and users, but they can weaken the connection between producing a successful computational artifact and retaining scientific command over the consequential choices and evidence on which it depends. For high-energy physics, this shifts part of the training challenge from code production toward scientific stewardship: researchers must be able to validate, explain, reproduce, maintain, and justify software and workflows partly generated or modified by AI. We argue that AI-aware software training should therefore combine dedicated instruction on AI systems with the integration of AI into established software and analysis training. Testing, provenance, reproducibility, resource judgment, and scientific validation should be exercised in AI-assisted workflows, with competence judged by the evidence supporting a result rather than by code generation alone. We frame this as a problem of sustainable scientific software: AI-assisted workflows should remain trustworthy and maintainable, use computational resources responsibly, and avoid making core competence unnecessarily dependent on privileged access to particular tools or infrastructure.

## AI's hidden energy tax: More than 80% of scale-up traffic is just memory moving around

DevFeed: [AI's hidden energy tax: More than 80% of scale-up traffic is just memory moving around](<https://devfeed.tech/articles/ai-s-hidden-energy-tax-more-than-80-of-scale-up-traffic-is-just-memory-moving-around-140757.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/11/18/memory-bottlenecks-ai-how-photonics-shared-pools-help-sc26/>)

Author: Jonathan Anthony

Published: 2026-11-18T14:57:55Z

Content type: article

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [clean energy data centers](<https://devfeed.tech/topics/clean-energy-data-centers.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [ddr5](<https://devfeed.tech/topics/ddr5.md>)

Tags: [accelerated-computing](<https://devfeed.tech/tags/accelerated-computing.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-data-centers](<https://devfeed.tech/tags/ai-data-centers.md>), [ai-inference](<https://devfeed.tech/tags/ai-inference.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [andy-mills](<https://devfeed.tech/tags/andy-mills.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [bottlenecks](<https://devfeed.tech/tags/bottlenecks.md>), [cache](<https://devfeed.tech/tags/cache.md>), [cube-event-coverage](<https://devfeed.tech/tags/cube-event-coverage.md>), [data](<https://devfeed.tech/tags/data.md>), [ddr5](<https://devfeed.tech/tags/ddr5.md>), [disaggregated-memory](<https://devfeed.tech/tags/disaggregated-memory.md>), [energy](<https://devfeed.tech/tags/energy.md>), [energy-efficiency](<https://devfeed.tech/tags/energy-efficiency.md>), [exascale-computing](<https://devfeed.tech/tags/exascale-computing.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [hbm](<https://devfeed.tech/tags/hbm.md>), [high-bandwidth-memory](<https://devfeed.tech/tags/high-bandwidth-memory.md>), [high-performance-computing](<https://devfeed.tech/tags/high-performance-computing.md>), [high-performance-storage](<https://devfeed.tech/tags/high-performance-storage.md>), [high-speed-networking](<https://devfeed.tech/tags/high-speed-networking.md>), [hpc](<https://devfeed.tech/tags/hpc.md>), [hpc-infrastructure](<https://devfeed.tech/tags/hpc-infrastructure.md>), [hpc-unites](<https://devfeed.tech/tags/hpc-unites.md>), [inference](<https://devfeed.tech/tags/inference.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [john-furrier](<https://devfeed.tech/tags/john-furrier.md>), [kv-cache](<https://devfeed.tech/tags/kv-cache.md>), [marvell](<https://devfeed.tech/tags/marvell.md>), [memory](<https://devfeed.tech/tags/memory.md>), [memory-pooling](<https://devfeed.tech/tags/memory-pooling.md>), [memoryai-photonic-appliance](<https://devfeed.tech/tags/memoryai-photonic-appliance.md>), [news](<https://devfeed.tech/tags/news.md>), [optics](<https://devfeed.tech/tags/optics.md>), [penguin-solutions](<https://devfeed.tech/tags/penguin-solutions.md>), [photonic-fabric](<https://devfeed.tech/tags/photonic-fabric.md>), [photonics](<https://devfeed.tech/tags/photonics.md>), [preet-virk](<https://devfeed.tech/tags/preet-virk.md>), [processors](<https://devfeed.tech/tags/processors.md>), [research-computing](<https://devfeed.tech/tags/research-computing.md>), [sc26](<https://devfeed.tech/tags/sc26.md>), [scale-up](<https://devfeed.tech/tags/scale-up.md>), [thecube](<https://devfeed.tech/tags/thecube.md>), [traffic](<https://devfeed.tech/tags/traffic.md>)

### AI overview

Penguin Solutions' MemoryAI Photonic Appliance places HBM and DDR5 in a shared pool tens of meters from GPUs, connected by Marvell's Photonic Fabric. Each appliance holds 16 TB, which servers across multiple racks can access. The approach targets inference memory bottlenecks and long token contexts by reducing the need to move or recompute data. Marvell's Preet Virk says memory transactions account for more than 80% of scale-up traffic and that, compared with copper-based data movement, the solution can use one-quarter the energy per bit. The excerpt presents these as company representatives' claims; it does not provide independent measurements or broader deployment results.

### Source excerpt

Memory bottlenecks are emerging as one of the costliest problems in AI infrastructure as inference overtakes training. GPUs sit idle waiting on data, while high-bandwidth memory, or HBM, stays bolted to processors that cannot share it. Photonics is shaping up as one way out. Marvell Technology Inc. has been pairing optics with photonic memory to [...] The post AI's hidden energy tax: More than 80% of scale-up traffic is just memory moving around appeared first on SiliconANGLE.

## Semantic Communication via Analog RF Computing: When the Radio Computes the Meaning

DevFeed: [Semantic Communication via Analog RF Computing: When the Radio Computes the Meaning](<https://devfeed.tech/articles/semantic-communication-via-analog-rf-computing-when-the-radio-computes-the-meaning-139962.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22100>)

Author: Christo Kurisummoottil Thomas

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [communications](<https://devfeed.tech/topics/communications.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>)

Tags: [analog](<https://devfeed.tech/tags/analog.md>), [analog-computing](<https://devfeed.tech/tags/analog-computing.md>), [broadcast](<https://devfeed.tech/tags/broadcast.md>), [budget](<https://devfeed.tech/tags/budget.md>), [channel](<https://devfeed.tech/tags/channel.md>), [client](<https://devfeed.tech/tags/client.md>), [co-design](<https://devfeed.tech/tags/co-design.md>), [communication](<https://devfeed.tech/tags/communication.md>), [compute](<https://devfeed.tech/tags/compute.md>), [computing](<https://devfeed.tech/tags/computing.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [costs](<https://devfeed.tech/tags/costs.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-it](<https://devfeed.tech/tags/cs-it.md>), [design](<https://devfeed.tech/tags/design.md>), [device](<https://devfeed.tech/tags/device.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-devices](<https://devfeed.tech/tags/edge-devices.md>), [eess-sp](<https://devfeed.tech/tags/eess-sp.md>), [energy](<https://devfeed.tech/tags/energy.md>), [front-end](<https://devfeed.tech/tags/front-end.md>), [math-it](<https://devfeed.tech/tags/math-it.md>), [rf](<https://devfeed.tech/tags/rf.md>), [semantic](<https://devfeed.tech/tags/semantic.md>)

### AI overview

The proposed framework moves semantic encoding to each client's RF receiver: a base station broadcasts encoder weights as RF waveforms, and a passive mixer computes semantic latents for uplink transmission and network-side inference. It models analog-computing and channel noise together, then allocates client energy across clients, encoder layers, and uplinks according to each stage's semantic sensitivity; the allocation also shapes broadcast beamforming. Stochastic refinement and noise-aware multi-task training further co-design the system. In MNIST simulations at matched total client energy, the allocation reduced semantic distortion by 9.2-19.2% versus uniform precision and 5.1-13.1% versus client-blind allocation; co-design yielded a further 46% reduction.

### Source excerpt

arXiv:2611.22100v1 Announce Type: cross Abstract: Semantic communication (SC) systems transmit continuous latent representations of meaning, but computing the semantic encoder at resource-constrained edge devices incurs substantial energy and memory costs. To address this challenge, we propose an end-to-end SC framework that executes the semantic encoder in the client's RF receiver front end using analog RF computing. A base station (BS) broadcasts encoder weights as RF waveforms, while each edge device reuses its receiver's passive mixer to compute semantic latents at ultra-low energy. These latents are transmitted over the uplink for network-side task inference. We show that analog computing noise across encoder layers and channel noise affecting transmitted latents admit a unified mathematical model. We define a task-posterior semantic distortion metric and derive its second-order decomposition into per-client, per-stage semantic sensitivities. Using this decomposition, we obtain a closed-form semantic-aware water-filling allocation of the total client energy budget across clients, encoder layers, and uplinks, assigning tighter accuracy targets to semantically sensitive stages. This allocation induces a broadcast beamforming problem that accounts for client-specific, per-layer accuracy requirements. We further refine the allocation through stochastic optimization and jointly design it with multi-task noise-aware training of the shared semantic model. MNIST simulations show that, at matched total client energy, the proposed allocation reduces network semantic distortion by 9.2\%-19.2\% compared with uniform precision and 5.1\%-13.1\% compared with client-blind allocation. Noise-aware co-design yields a further 46\% reduction.

## The Language-Energy Divide: Measuring Energy Costs of Multilingual LLM Inference

DevFeed: [The Language-Energy Divide: Measuring Energy Costs of Multilingual LLM Inference](<https://devfeed.tech/articles/the-language-energy-divide-measuring-energy-costs-of-multilingual-llm-inference-139784.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.21869>)

Author: Naihao Deng, Alissa Shen, Yiming Feng, Joan Nwatu, Jae-Won Chung, Mosharaf Chowdhury, Yulong Chen, Rada Mihalcea

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [energy](<https://devfeed.tech/tags/energy.md>), [energy-consumption](<https://devfeed.tech/tags/energy-consumption.md>), [energy-costs](<https://devfeed.tech/tags/energy-costs.md>), [energy-efficiency](<https://devfeed.tech/tags/energy-efficiency.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [inference](<https://devfeed.tech/tags/inference.md>), [language](<https://devfeed.tech/tags/language.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [languages](<https://devfeed.tech/tags/languages.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llm](<https://devfeed.tech/tags/llm.md>), [llm-deployment](<https://devfeed.tech/tags/llm-deployment.md>), [low-resource-languages](<https://devfeed.tech/tags/low-resource-languages.md>), [measuring](<https://devfeed.tech/tags/measuring.md>), [mitigations](<https://devfeed.tech/tags/mitigations.md>), [multilingual-llm](<https://devfeed.tech/tags/multilingual-llm.md>), [performance](<https://devfeed.tech/tags/performance.md>), [reporting](<https://devfeed.tech/tags/reporting.md>), [scripts](<https://devfeed.tech/tags/scripts.md>)

### AI overview

The study reports that multilingual LLM inference energy varies substantially by language: up to 8.3x per output token and, for a fixed request set, from 17.6 kJ for English to 3,147 kJ for Pashto. It attributes the gap to higher per-token costs for complex or rare scripts and more generated tokens for low-resource languages. Languages with the highest energy use also tend to have the lowest task accuracy, and the reported divide persists across models, hardware, and tasks. The authors recommend including energy in evaluations, reporting checklists, and model cards, and using deployment-side efficiency measures.

### Source excerpt

arXiv:2606.21869v3 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly deployed in multilingual settings, yet the energy costs of serving these models across different languages remain poorly understood. We present a systematic study of inference energy consumption across languages with ML.Energy framework (Chung et al., 2026). We find striking disparities: energy consumption per output token varies by up to 8.3 times across languages, while total energy for a fixed set of requests varies by up to 179 times between the cheapest (English, 17.6 kJ) and the most expensive (Pashto, 3,147 kJ) languages. Our analysis shows that this disparity is driven by two compounding factors: (1) higher per-token energy costs for languages using complex or rare scripts, and (2) more tokens generated for low-resource languages. Moreover, we find a double cost + performance penalty: languages with the highest energy footprints also tend to achieve the lowest task accuracy. We reveal that the energy divide persists across models, hardware, and tasks, suggesting a systemic energy inequity in multilingual LLM deployment. Finally, we recommend that the community treat energy as a first-class evaluation axis, extend reporting checklists and model cards to include it, and adopt deployment-side mitigations for better energy efficiency.

## Collecting Infrastructure-Level Measures for Energy Analysis in Modern Computing Infrastructures

DevFeed: [Collecting Infrastructure-Level Measures for Energy Analysis in Modern Computing Infrastructures](<https://devfeed.tech/articles/collecting-infrastructure-level-measures-for-energy-analysis-in-modern-computing-infrastructures-140383.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22160>)

Author: Isac Pasianotto, Stefano Cozzini, Niccolo Tosato

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [data centers](<https://devfeed.tech/topics/data-centers.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [infrastructure monitoring](<https://devfeed.tech/topics/infrastructure-monitoring.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>)

Tags: [accounting](<https://devfeed.tech/tags/accounting.md>), [acquisition](<https://devfeed.tech/tags/acquisition.md>), [aggregates](<https://devfeed.tech/tags/aggregates.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [analyze](<https://devfeed.tech/tags/analyze.md>), [approach](<https://devfeed.tech/tags/approach.md>), [case-studies](<https://devfeed.tech/tags/case-studies.md>), [cluster](<https://devfeed.tech/tags/cluster.md>), [collecting](<https://devfeed.tech/tags/collecting.md>), [collection](<https://devfeed.tech/tags/collection.md>), [complex](<https://devfeed.tech/tags/complex.md>), [component](<https://devfeed.tech/tags/component.md>), [components](<https://devfeed.tech/tags/components.md>), [compute](<https://devfeed.tech/tags/compute.md>), [computing](<https://devfeed.tech/tags/computing.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [data](<https://devfeed.tech/tags/data.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [energy](<https://devfeed.tech/tags/energy.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [modern](<https://devfeed.tech/tags/modern.md>)

### AI overview

The paper presents a cluster-wide pipeline for energy analysis that synchronizes fine-grained measurements across infrastructure subsystems, addressing dynamics that measurements from individual compute nodes can miss. It reports a dataset sampled at 1 Hz for one month in a production research data center running HPC, Kubernetes services, and distributed storage. Case studies of distributed I/O and large-language-model inference illustrate the value of infrastructure-wide measurements. The dataset and method are intended to support future predictive modeling and energy-aware resource management.

### Source excerpt

arXiv:2611.22160v1 Announce Type: new Abstract: Modern HPC facilities are complex environments in which all components are tightly coupled; consequently, precisely modeling and accounting for energy consumption requires approaches that go beyond instrumenting individual compute nodes. Aiming for a cluster-wide vision, we designed a scalable data collection pipeline that aggregates fine-grained measurements from multiple subsystems-spanning component loads, software and hardware counters, sensors and energy-ensuring consistent sampling frequencies and synchronized data acquisition across the infrastructure. We present a dataset of comprehensive metrics collected at 1 Hz over one month from a production research data center supporting heterogeneous workloads, including HPC, Kubernetes services, and distributed storage. To illustrate what this approach makes possible in practice, we analyze two case studies-distributed I/O and large-language-model inference-showing how cluster-wide measurement reveals system-level dynamics missed by localized instrumentation. The proposed dataset and collection methodology provide a foundation for future work on predictive modeling, energy-aware resource management, and more efficient computing infrastructures.

## AI Is Giving Optical Computing Another Shot

DevFeed: [AI Is Giving Optical Computing Another Shot](<https://devfeed.tech/articles/ai-is-giving-optical-computing-another-shot-138573.md>)

Original publisher: [Read original article](<https://spectrum.ieee.org/optical-computing-ai>)

Author: Dina Genkina

Published: 2026-11-17T14:00:03Z

Content type: article

Language: en

Sources: [IEEE Spectrum](<https://devfeed.tech/sources/ieee-spectrum.md>)

Topics: [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [communications](<https://devfeed.tech/topics/communications.md>), [fourier analysis](<https://devfeed.tech/topics/fourier-analysis.md>), [Electrical engineering and computer science (EECS)](<https://devfeed.tech/topics/electrical-engineering-and-computer-science-eecs.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-inference](<https://devfeed.tech/tags/ai-inference.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [algorithms](<https://devfeed.tech/tags/algorithms.md>), [computing](<https://devfeed.tech/tags/computing.md>), [data-centers](<https://devfeed.tech/tags/data-centers.md>), [energy](<https://devfeed.tech/tags/energy.md>), [energy-efficiency](<https://devfeed.tech/tags/energy-efficiency.md>), [fourier-transform](<https://devfeed.tech/tags/fourier-transform.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [industry](<https://devfeed.tech/tags/industry.md>), [inference](<https://devfeed.tech/tags/inference.md>), [investment](<https://devfeed.tech/tags/investment.md>), [light](<https://devfeed.tech/tags/light.md>), [optical-computing](<https://devfeed.tech/tags/optical-computing.md>), [photonics](<https://devfeed.tech/tags/photonics.md>), [silicon-photonics](<https://devfeed.tech/tags/silicon-photonics.md>)

### AI overview

Optical computing is attracting investment as AI inference drives data-center power demand, but its practical advantages depend on workload and system design. Passive optical matrix multiplication could reduce energy as matrix sizes grow; however, encoding and readout costs can erase the savings at smaller sizes, and only part of inference maps to optical operations. Free-space systems such as Lumai's target compute-intensive prefill, while integrated photonics trades some speed for easier manufacturing; analog-to-digital conversion remains a major energy barrier. Other approaches target specialized operations or compute within optical networks, including memory-bound decode tasks. The excerpt describes promising prototypes and company performance claims, alongside substantial scaling and benchmarking limits. It suggests photonics may complement CPUs and GPUs in specialized roles rather than replace them broadly.

### Source excerpt

The idea of computing with light has been around for over 70 years, but until recently it's remained largely in the realm of research. Over the past few years, though, optical computing has emerged as a fast-growing industry: Market valuation of companies in this sector now totals US $2.22 billion, up from $0.4 billion in 2022. One projection suggests that the market will continue to grow exponentially in the coming years. The reason for this sudden surge will surprise no one: AI. In particular, AI inference--the process of using AI models to respond to user queries--has driven much of the construction of huge, power-hungry data centers, and relatively energy-efficient optical systems are seen as a potential way forward. "Most of the interest in [photonics] is driven by the fact that AI inference has literally become limited by power, and in a rather striking manner," says Peter McMahon, associate professor of applied and engineering physics at Cornell University, in Ithaca, N.Y., who has researched optical computing for over a decade. Optics, at least on paper, carries the promise of doing the same math as today's GPUs at a drastically lower level of power. Switching from electronics to optics would be a pretty radical solution. Nevertheless, like mushrooms after a rainstorm, optical-computing startups are popping up in the United States, China, Germany, and Israel. And while competing against such GPU makers as Nvidia is an ambitious goal, the investment in optical computing has spurred development of complementary approaches, such as designing alternative AI algorithms, taking on scientific computing, and enhancing in-network computing. Optalysys, based in England, is inserting optical-computing elements into the networking path, where data centers already use light. The company's basic on-chip optical element, shown in the inset, is a Fourier transform calculator.Optalysys "I think it would be good for the technology if many of these companies made a good run at i

## Lasso: Towards Sustainable AI-Generated Media Detection with Intel QAT On-Chip Compression

DevFeed: [Lasso: Towards Sustainable AI-Generated Media Detection with Intel QAT On-Chip Compression](<https://devfeed.tech/articles/lasso-towards-sustainable-ai-generated-media-detection-with-intel-qat-on-chip-compression-137360.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.18895>)

Author: Darsh Asher, Joshua Kalyanapu, Kaushal Mhapsekar, Bita Aslrousta, Amna Hassan, Debayan Sengupta, Samira Mirbagher Ajorpaz

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [Reverse Engineering](<https://devfeed.tech/topics/reverse-engineering.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Compression](<https://devfeed.tech/topics/compression.md>), [Generative AI](<https://devfeed.tech/topics/generative-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-generated-content](<https://devfeed.tech/tags/ai-generated-content.md>), [ai-generated-media](<https://devfeed.tech/tags/ai-generated-media.md>), [artifacts](<https://devfeed.tech/tags/artifacts.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [binoculars](<https://devfeed.tech/tags/binoculars.md>), [chip](<https://devfeed.tech/tags/chip.md>), [code](<https://devfeed.tech/tags/code.md>), [compression](<https://devfeed.tech/tags/compression.md>), [content](<https://devfeed.tech/tags/content.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [data](<https://devfeed.tech/tags/data.md>), [data-structure](<https://devfeed.tech/tags/data-structure.md>), [detection](<https://devfeed.tech/tags/detection.md>), [device](<https://devfeed.tech/tags/device.md>), [domains](<https://devfeed.tech/tags/domains.md>), [edge](<https://devfeed.tech/tags/edge.md>), [energy](<https://devfeed.tech/tags/energy.md>), [intel](<https://devfeed.tech/tags/intel.md>), [media](<https://devfeed.tech/tags/media.md>)

### AI overview

Lasso uses timing differences in Intel QuickAssist Technology's compression accelerator as a signal for detecting AI-generated media. For text, the abstract reports 100% accuracy on original DetectRL text, 92% average after paraphrasing, and 82% across original, paraphrased, and perturbed text; accuracy reaches 95% for matched instruction-tuned generators, including an unseen generator. On RAID, it averages 82% when tested on a different generator and 74% across eight held-out domains. Detection takes 0.74 microseconds, at least 83,784 times faster than five baselines, and uses over 99% less energy per document than Binoculars and DetectGPT. It also reports 67.5% accuracy for video, 70% for code, and 78% for images. These are abstract-level reported results; evaluation details are not included in the supplied excerpt.

### Source excerpt

arXiv:2611.18895v1 Announce Type: cross Abstract: Synthetic text, images, and video are rising. Software-only detectors are power-hungry, too slow, and costly, and do not scale with demand. When software cannot scale, historically hardware often comes to the rescue. We perform the first reverse engineering of a compression accelerator--Intel QuickAssist Technology (QAT) microarchitecture and--find different amount of time on each, a novel data structure-dependent property. Using this information, we then build Lasso, a low-cost, high-performance detector for AI-generated content that turns the accelerator into a measuring instrument. For text detection, Lasso achieves 100% accuracy on original DetectRL text and retains 92% average accuracy after paraphrasing, compared with 78% for Binoculars and 55% for DetectGPT. Across original, paraphrased, and perturbed text, it achieves 82% average accuracy, outperforming all five baselines. With topic, length, and formatting matched, accuracy reaches 95% for instruction-tuned generators, including when the tested generator is unseen. On RAID, accuracy averages 82% when trained on one generator and tested on another, and 74% across eight held-out domains. Detection takes 0.74 microseconds, a speedup of at least 83,784x across five baselines. At the same time, Lasso consumes over 99% less energy per document than both Binoculars and DetectGPT. At the same time, Lasso detects AI-generated videos, software code, and images with 67.5%, 70% and 78% accuracy, respectively, with low latency and power opening avenues for practical real time and edge device detection of multimodality AI artifacts for the first time. No other compared technique supports detection across all four modalities: text, video, images, and code.

## A predictive, per-engine energy roofline of the Apple Neural Engine

DevFeed: [A predictive, per-engine energy roofline of the Apple Neural Engine](<https://devfeed.tech/articles/a-predictive-per-engine-energy-roofline-of-the-apple-neural-engine-138834.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19101>)

Author: Spencer H. Bryngelson

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [performance-engineering](<https://devfeed.tech/topics/performance-engineering.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>)

Tags: [apple](<https://devfeed.tech/tags/apple.md>), [apple-neural-engine](<https://devfeed.tech/tags/apple-neural-engine.md>), [compression](<https://devfeed.tech/tags/compression.md>), [compute](<https://devfeed.tech/tags/compute.md>), [coreml](<https://devfeed.tech/tags/coreml.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-ce](<https://devfeed.tech/tags/cs-ce.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-pf](<https://devfeed.tech/tags/cs-pf.md>), [dispatch](<https://devfeed.tech/tags/dispatch.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [energy](<https://devfeed.tech/tags/energy.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [graph](<https://devfeed.tech/tags/graph.md>), [int4](<https://devfeed.tech/tags/int4.md>), [int8](<https://devfeed.tech/tags/int8.md>), [memory](<https://devfeed.tech/tags/memory.md>), [operators](<https://devfeed.tech/tags/operators.md>), [performance](<https://devfeed.tech/tags/performance.md>), [throughput](<https://devfeed.tech/tags/throughput.md>), [weight](<https://devfeed.tech/tags/weight.md>)

### AI overview

Using a CoreML-free fused-graph dispatch path and per-engine energy accounting, the study finds that the consumer-available roofline does not predict Apple Neural Engine performance. It models separate slow activation and fast weight-stream bandwidths, yielding two ridges; the weight-stream ridge lies below the GPU's. The ANE's single-operand compute peak falls rather than plateauing. For weight streams, fp16 is byte-bound, while int8 is limited by dequantizer element rate. Int8 compression saves energy at unchanged throughput; an int4 lookup-table stream improves throughput and efficiency at a cost of 2.55 ImageNet top-1 points. The constructed device map reproduces across four ANE generations, and the results support placing operators by arithmetic intensity and fusing them for performance.

### Source excerpt

arXiv:2611.19101v1 Announce Type: cross Abstract: The Apple Neural Engine (ANE) is in over a billion devices, but is characterized indirectly via CoreML. Using a CoreML-free fused-graph dispatch path, we measure the ANE, along with the Apple GPU and CPU on the same device, with per-engine energy accounting. The consumer-available roofline does not predict the ANE in practice. A critical correction is that the ANE sees two effective memory bandwidths: a slow activation path and a fast weight-stream path. Two roofline ridges follow, and the weight-stream ridge is below that of the GPU. The ANE single-operand compute peak decreases instead of plateauing. Its weight stream is byte-bound in fp16 but limited by the dequantizer's element rate in int8. As a result, int8 compression saves energy at unchanged throughput and an int4 lookup-table stream raises throughput and efficiency at a cost of 2.55 points of ImageNet top-1. The corrected roofline identifies the lower-energy ANE paths and bounds it from below. A constructed device map reproduces across four unique ANE generations. Results show that operators should be placed by arithmetic intensity and fused for optimal performance.

## Towards Sustainable Large Language Model Inference: A Systematic Literature Review

DevFeed: [Towards Sustainable Large Language Model Inference: A Systematic Literature Review](<https://devfeed.tech/articles/towards-sustainable-large-language-model-inference-a-systematic-literature-review-137956.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.18570>)

Author: Sabiya Banu Masthan Ali, Syed Muhammad Danish, Rupinder Kaur, Marin Litoiu, Arghavan Asad

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [energy-consumption](<https://devfeed.tech/tags/energy-consumption.md>), [energy-efficiency](<https://devfeed.tech/tags/energy-efficiency.md>), [environmental-impact](<https://devfeed.tech/tags/environmental-impact.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [inference](<https://devfeed.tech/tags/inference.md>), [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [latency](<https://devfeed.tech/tags/latency.md>), [literature](<https://devfeed.tech/tags/literature.md>), [llm-inference](<https://devfeed.tech/tags/llm-inference.md>), [llms](<https://devfeed.tech/tags/llms.md>), [measurement](<https://devfeed.tech/tags/measurement.md>), [metrics](<https://devfeed.tech/tags/metrics.md>), [performance](<https://devfeed.tech/tags/performance.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [review](<https://devfeed.tech/tags/review.md>), [software](<https://devfeed.tech/tags/software.md>), [survey](<https://devfeed.tech/tags/survey.md>)

### AI overview

The review examines ways to reduce energy use in LLM inference across input, model, system, and hardware levels. It considers how energy savings relate to latency and accuracy, noting that studies evaluate these trade-offs inconsistently. It also surveys energy-measurement tools, datasets, and metrics, including software- and hardware-based approaches, and identifies limitations in current evaluation practices. The supplied text is an abstract and does not specify which techniques perform best or quantify their effects.

### Source excerpt

arXiv:2611.18570v1 Announce Type: new Abstract: Large language model (LLM) inference is becoming a significant contributor to both operational cost and environmental impact in modern AI deployments. Energy consumption during inference depends on multiple factors across the pipeline, including input and output length, model design, system-level decisions, and hardware characteristics. This paper presents a systematic literature review (SLR) of energy-efficient and sustainable LLM inference. We analyze selected primary studies to identify techniques proposed to reduce energy consumption across different levels of the inference pipeline, including input, model, system, and hardware. We further examine the trade-offs between energy consumption, latency, and accuracy, highlighting inconsistencies in how these relationships are evaluated. In addition, we review the tools, datasets, and metrics used for energy measurement, covering both software-based and hardware-based approaches. This survey provides a comprehensive synthesis of existing literature, offering insights into the effectiveness of different techniques in terms of energy savings and their associated impact on accuracy. It highlights the trade-offs between energy efficiency and performance and identifies limitations in current evaluation practices. Overall, this work aims to support future research toward more efficient and environmentally responsible deployment of LLMs. Our repository of relevant references is maintained on GitHub.

## Investigating the Impact of CI-Source Sensitivity in Carbon-Aware Resource Management Systems

DevFeed: [Investigating the Impact of CI-Source Sensitivity in Carbon-Aware Resource Management Systems](<https://devfeed.tech/articles/investigating-the-impact-of-ci-source-sensitivity-in-carbon-aware-resource-management-systems-136994.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20830>)

Author: Gon\\c{c}alo Ferreira, Jaime Iglesias Blanco, Mar\\'ia Castrillo, \\'Alvaro L\\'opez Garc\\'ia, Li Wu, Shashikant Ilager

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Green Software](<https://devfeed.tech/topics/green-software.md>), [clean energy data centers](<https://devfeed.tech/topics/clean-energy-data-centers.md>), [performance-engineering](<https://devfeed.tech/topics/performance-engineering.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [artificial-intelligence-ai](<https://devfeed.tech/tags/artificial-intelligence-ai.md>), [carbon](<https://devfeed.tech/tags/carbon.md>), [ci](<https://devfeed.tech/tags/ci.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [emissions](<https://devfeed.tech/tags/emissions.md>), [energy](<https://devfeed.tech/tags/energy.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [framework](<https://devfeed.tech/tags/framework.md>), [grid](<https://devfeed.tech/tags/grid.md>), [latency](<https://devfeed.tech/tags/latency.md>), [management](<https://devfeed.tech/tags/management.md>), [operational](<https://devfeed.tech/tags/operational.md>), [performance](<https://devfeed.tech/tags/performance.md>), [providers](<https://devfeed.tech/tags/providers.md>), [resource](<https://devfeed.tech/tags/resource.md>), [resource-management](<https://devfeed.tech/tags/resource-management.md>), [source](<https://devfeed.tech/tags/source.md>)

### AI overview

Using Wattnet and ElectricityMaps traces across five Central European zones, the study evaluates how carbon-intensity (CI) signal differences affect carbon-aware placement. Its ESCAPE framework tests provider switching and controlled CI perturbations in a reproduced setting, using CarbonEdge to measure placement decisions, carbon proxy, energy use, and latency. Switching CI sources changed the dominant placement site in 43.50% of cases and altered reported operational outcomes, despite high rank correlation between sources. The authors argue that CI uncertainty should be accounted for explicitly in resource-allocation decisions.

### Source excerpt

arXiv:2611.20830v1 Announce Type: new Abstract: The sustainable growth of data centres increasingly depends on Resource-Management Systems (RMSs) that aim to reduce carbon emissions of computing workloads without degrading performance. This need is intensified by the rise of Artificial Intelligence (AI) workloads, which demand exponentially growing amounts of compute and energy across geographically distributed infrastructure. Recent carbon-aware RMSs exploit temporal and spatial variation in grid Carbon Intensity (CI) to shift workloads towards lower-carbon periods or locations. In practice, CI estimates and forecasts can be inaccurate and uncertain, and CI providers' methods differ substantially. Such discrepancies may have a significant impact on RMS placement decision-making, raising concerns about the fragility of existing solutions. Thus, in this work, we ask how discrepancies between plausible CI estimates propagate into resource-management decisions and operational outcomes. To answer this question, we harmonise and characterise CI traces from Wattnet and ElectricityMaps across five Central European zones. We then propose our evaluation framework: External CI Signal Characterisation And Perturbation Evaluation (ESCAPE), which assesses provider switching and controlled CI perturbations in a reproduced placement setting. We implement our framework using CarbonEdge as a use case, and measure changes in placement decisions and operational outcomes, such as carbon proxy, energy use, and latency. Our results show that switching CI-source changes the dominant placement site in 43.50% cases, and alters reported operational outcomes. These changes occur despite high rank correlation and structured disagreement between CI-sources. This observed sensitivity to CI-source choice motivates robustness-aware carbon-aware placement, in which CI uncertainty should be explicitly accounted for in resource-allocation decisions.

## A Matter of Stages: Energy-Latency Trade-offs in Multimodal LLM Serving

DevFeed: [A Matter of Stages: Energy-Latency Trade-offs in Multimodal LLM Serving](<https://devfeed.tech/articles/a-matter-of-stages-energy-latency-trade-offs-in-multimodal-llm-serving-136996.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.21165>)

Author: Yifan Sun, Jiantong Jiang, Zhexiang Zhang, Xu Bai, Feng Liu, Rajkumar Buyya, Adel N. Toosi

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [LLM Techniques](<https://devfeed.tech/topics/llm-techniques.md>), [performance-engineering](<https://devfeed.tech/topics/performance-engineering.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [disaggregated-serving](<https://devfeed.tech/tags/disaggregated-serving.md>), [disaggregation](<https://devfeed.tech/tags/disaggregation.md>), [energy](<https://devfeed.tech/tags/energy.md>), [latency](<https://devfeed.tech/tags/latency.md>), [llm](<https://devfeed.tech/tags/llm.md>), [llm-serving](<https://devfeed.tech/tags/llm-serving.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [serving](<https://devfeed.tech/tags/serving.md>), [stage](<https://devfeed.tech/tags/stage.md>)

### AI overview

Across the evaluated workloads, separating visual encoding, prefill, and decode across GPUs reduced GPU energy per successful request by 13.3-37.0% at nearly unchanged output-token throughput, but increased mean time to first token (TTFT). The study finds that stage-specific frequency reductions have different effects: lower power can be offset by queueing and longer pipeline execution. Uniform downclocking to 80% reduced GPU energy by 11.7-18.6% with modest mean-latency changes. To meet latency constraints, the authors' budgeted exploration strategy used four probes to identify the lowest-energy feasible option among 25 measured configurations across 288 evaluated positive-margin SLO scenarios. Results cover four multimodal models under a fixed three-GPU budget and 18 workload combinations.

### Source excerpt

arXiv:2611.21165v1 Announce Type: new Abstract: Online serving of multimodal large language models requires deciding whether to separate visual encoding (E), prefill (P), and decode (D) across GPUs and how to configure their frequencies. EPD disaggregation enables independent stage control, but it also changes communication and queueing behavior. Lowering one stage's frequency can increase queueing and prolong execution across the EPD pipeline. Consequently, reducing power at one stage does not necessarily reduce total GPU energy per request. In this paper, we systematically characterize fused and EPD-disaggregated serving across four multimodal models under a fixed three-GPU budget, followed by single-stage and uniform downclocking experiments across 18 workload combinations. We find that disaggregation reduces GPU energy per successful request by 13.3-37.0% at nearly identical output-token throughput, but increases mean TTFT. Downclocking exhibits distinct stage sensitivities, and aggressive frequency reductions can offset power savings through queueing and longer execution. Across the evaluated workloads, uniform downclocking to 80% reduces GPU energy by 11.7-18.6% with modest mean-latency changes. Guided by these findings, we develop a budgeted frequency exploration strategy to select energy-efficient configurations under latency constraints. Evaluation shows that four probes can identify the lowest-energy feasible configuration among 25 measured candidates across all 288 evaluated positive-margin SLO scenarios. These results provide practical guidance for energy-efficient multimodal serving.

## Measuring and Optimizing Energy Efficiency in ICT: A Multi-vocal Literature Review on Data Centers, Cloud Computing, and LLM-Driven Code Optimization

DevFeed: [Measuring and Optimizing Energy Efficiency in ICT: A Multi-vocal Literature Review on Data Centers, Cloud Computing, and LLM-Driven Code Optimization](<https://devfeed.tech/articles/measuring-and-optimizing-energy-efficiency-in-ict-a-multi-vocal-literature-review-on-data-centers-cloud-computing-and-llm-driven-code-optimization-137962.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.18576>)

Author: Tobias B\\"ohm, Benjamin Weigell, Ezgi Sarikayak, Philipp M. Z\\"ahl, Tina Vartziotis, Simon Hornung, Julien Siebert, Anika Hennig, Bernhard Bauer, Andreas Biesdorf

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Green Software](<https://devfeed.tech/topics/green-software.md>), [clean energy data centers](<https://devfeed.tech/topics/clean-energy-data-centers.md>), [data centers](<https://devfeed.tech/topics/data-centers.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Cloud Architecture](<https://devfeed.tech/topics/cloud-architecture.md>), [reproducibility](<https://devfeed.tech/topics/reproducibility.md>)

Tags: [cloud-computing](<https://devfeed.tech/tags/cloud-computing.md>), [code-generation](<https://devfeed.tech/tags/code-generation.md>), [container](<https://devfeed.tech/tags/container.md>), [containers](<https://devfeed.tech/tags/containers.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [data-centers](<https://devfeed.tech/tags/data-centers.md>), [energy](<https://devfeed.tech/tags/energy.md>), [energy-efficiency](<https://devfeed.tech/tags/energy-efficiency.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [literature](<https://devfeed.tech/tags/literature.md>), [llms](<https://devfeed.tech/tags/llms.md>), [measuring](<https://devfeed.tech/tags/measuring.md>), [memory](<https://devfeed.tech/tags/memory.md>), [network](<https://devfeed.tech/tags/network.md>), [optimizing](<https://devfeed.tech/tags/optimizing.md>), [python](<https://devfeed.tech/tags/python.md>), [resource-management](<https://devfeed.tech/tags/resource-management.md>), [review](<https://devfeed.tech/tags/review.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [software](<https://devfeed.tech/tags/software.md>), [system-design](<https://devfeed.tech/tags/system-design.md>)

### AI overview

This multi-vocal review synthesizes 181 sources published or collected from 2010-2025 on measuring and reducing energy use in ICT, with emphasis on data centers and cloud computing. It organizes sustainability measures by domain--power and energy, efficiency, and environmental impact--and by component scale, from applications to system-of-systems. The reviewed optimization approaches include dynamic power and resource management, VM and container consolidation, energy-aware design, and cross-layer methods. For LLM-assisted code generation and transformation, most studies focus on Python and report runtime and memory; fewer directly measure energy. The review identifies gaps in vendor-neutral telemetry, consistent reporting, network metrics, and unified cross-layer optimization, which limit reproducibility and comparison.

### Source excerpt

arXiv:2611.18576v1 Announce Type: new Abstract: Rapid growth in Information and Communication Technology (ICT) increases electricity use and environmental impact. Addressing this challenge requires a systematic, software-centric view of how energy and related sustainability attributes are measured, reported, and improved. We present a comprehensive Multi-vocal Literature Review (MLR) integrating peer-reviewed and gray literature to map current practice in measuring and optimizing energy use in ICT software, with emphasis on data centers and cloud computing environments (DCCCEs). First, we synthesize how sustainability is measured across ICT by proposing a taxonomy organized along metric domains (power/energy, energy efficiency, environmental impact) and component granularity (from applications, to containers/VMs, and system-of-systems levels). We summarize measurement approaches and the tooling landscape that implements them. Second, we review optimization strategies at the software and system levels, covering dynamic power and resource management, VM/container consolidation, energy-aware software and system design, and cross-layer, system-wide methods relevant to DCCCEs. Third, we examine how large language models (LLMs) are applied to code generation and transformation for resource-efficient software, noting that most studies target code generation (predominantly in Python) and evaluate runtime and memory, with fewer works directly reporting energy. Hardware- and network-level telemetry are treated as data sources for software-level analysis. We identify gaps in vendor-neutral telemetry, consistent measurement and reporting practices, network-level metrics, and unified, cross-layer optimization frameworks, and outline directions for improving reproducibility and comparability. We screened 635 records (2010-2025) and included 181 sources (122 peer-reviewed, 59 gray).

## Re-Reading Is Not Free: Token Traces and Cache-Aware Energy Estimation for LLM Agents in Software Engineering

DevFeed: [Re-Reading Is Not Free: Token Traces and Cache-Aware Energy Estimation for LLM Agents in Software Engineering](<https://devfeed.tech/articles/re-reading-is-not-free-token-traces-and-cache-aware-energy-estimation-for-llm-agents-in-software-engineering-137625.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20873>)

Author: Tina Vartziotis, Ketevan Peranidze, Rodopi Kosteli, Ezgi Sarikayak, George Dasoulas, Michael Keckeisen, Sotirios Kotsopoulos, Francesca Dominici

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [agentic workflows](<https://devfeed.tech/topics/agentic-workflows.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Cache](<https://devfeed.tech/topics/cache.md>), [long-context](<https://devfeed.tech/topics/long-context.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-workloads](<https://devfeed.tech/tags/agentic-workloads.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [cache](<https://devfeed.tech/tags/cache.md>), [cached-context](<https://devfeed.tech/tags/cached-context.md>), [context](<https://devfeed.tech/tags/context.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [energy](<https://devfeed.tech/tags/energy.md>), [environmental-footprint](<https://devfeed.tech/tags/environmental-footprint.md>), [estimator](<https://devfeed.tech/tags/estimator.md>), [inference](<https://devfeed.tech/tags/inference.md>), [llm](<https://devfeed.tech/tags/llm.md>), [llm-agents](<https://devfeed.tech/tags/llm-agents.md>), [reuse](<https://devfeed.tech/tags/reuse.md>), [software](<https://devfeed.tech/tags/software.md>), [software-development](<https://devfeed.tech/tags/software-development.md>), [software-engineering](<https://devfeed.tech/tags/software-engineering.md>), [token](<https://devfeed.tech/tags/token.md>)

### AI overview

Across 42 sessions and 752 software-development task episodes, cached-context reads made up 94.8% of API-reported tokens, with about 260 cached tokens read per generated token. Extending an analytical inference-energy model to include cached-context processing yielded an estimated 250 kWh, versus about 10 kWh from the output-based estimator EcoLogits. Neither estimator has been validated on long cached contexts, and the tested modeling assumptions did not close the gap. The cache-aware estimate fell about 60% when each task started with fresh context. Parent-session logs also missed most subagent activity, undercounting delegated tokens tenfold; correcting that changed this workload's energy estimate by about 5%, though token-volume studies are more exposed to the omission.

### Source excerpt

arXiv:2611.20873v1 Announce Type: cross Abstract: As large language models (LLMs) evolve from conversational assistants into autonomous agents embedded in software development workflows, estimating their environmental footprint requires moving beyond conventional token-based accounting. Existing energy estimators focus on isolated request-response interactions, overlooking the repeated processing of accumulated context, cache reuse, and work delegated to subagents. We analyze nearly one billion tokens collected across 42 sessions and 752 task episodes over ten weeks of real-world AI-assisted software development. Cached-context reads make up 94.8% of all API-reported tokens, while uncached input accounts for only 0.017%; for every generated token, approximately 260 cached tokens are read. Context grows 5.3-fold over a session compared with 1.4-fold within individual tasks. To quantify the energy implications, we extend an existing analytical inference-energy model to account for cached-context processing, yielding a cache-aware hybrid estimator. It estimates approximately 250 kWh for the workload, compared with approximately 10 kWh from the output-based estimator EcoLogits, a 25-fold difference largely driven by cached-context reads. Neither estimator has been validated on long cached contexts, but no modeling assumption we varied closes the gap. Starting each task from a fresh context would lower our estimate by about 60%. We also find that parent-session logs report only a subagent's final internal call, undercounting delegated tokens by a factor of ten; for this workload the correction changes the energy estimate by about 5%, but token-volume studies are fully exposed to it. Our findings expose a substantial gap between how current energy estimators and agentic workloads treat context, highlighting the need for workflow-level energy accounting that captures context accumulation, cache reuse, and delegated execution.

## Arm enables the world's fastest and most energy-efficient supercomputers

DevFeed: [Arm enables the world's fastest and most energy-efficient supercomputers](<https://devfeed.tech/articles/arm-enables-the-world-s-fastest-and-most-energy-efficient-supercomputers-137165.md>)

Original publisher: [Read original article](<https://newsroom.arm.com/blog/arm-worlds-fastest-energy-efficient-supercomputers>)

Author: Arm Editorial Team

Published: 2026-11-17T00:48:18Z

Content type: article

Language: en

Sources: [Arm Newsroom](<https://devfeed.tech/sources/arm-newsroom.md>)

Topics: [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [Supercomputing](<https://devfeed.tech/topics/supercomputing.md>), [Arm](<https://devfeed.tech/topics/arm.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>)

Tags: [arm](<https://devfeed.tech/tags/arm.md>), [computing](<https://devfeed.tech/tags/computing.md>), [ecosystem](<https://devfeed.tech/tags/ecosystem.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [energy](<https://devfeed.tech/tags/energy.md>), [high-performance-computing](<https://devfeed.tech/tags/high-performance-computing.md>), [high-performance-computing-hpc](<https://devfeed.tech/tags/high-performance-computing-hpc.md>), [hpc](<https://devfeed.tech/tags/hpc.md>), [november-2026](<https://devfeed.tech/tags/november-2026.md>), [performance](<https://devfeed.tech/tags/performance.md>), [rankings](<https://devfeed.tech/tags/rankings.md>), [supercomputers](<https://devfeed.tech/tags/supercomputers.md>)

### AI overview

The supplied text says Arm-based systems hold the top positions in the November 2026 TOP500 and Green500 rankings, with four of the 10 fastest and five of the 10 most energy-efficient supercomputers using Arm technology. It distinguishes Arm-compatible custom processors, including those in LineShine and Fugaku, from Arm Neoverse-based NVIDIA Grace Hopper systems in JUPITER, Alps, and the Green500 leaders. It attributes HPC performance and efficiency to flexible hardware and software tuning: high-bandwidth memory and CPU-GPU connectivity for data movement; vector and matrix extensions, compilers, and math libraries for numerical work; parallel execution and data-locality tuning; and power tuning to balance throughput with power limits.

### Source excerpt

The November 2026 TOP500 and Green500 rankings showcase the breadth of the Arm high-performance computing (HPC) ecosystem, and how it combines performance, efficiency, and the freedom to innovate. Highlights include: Arm technology across the world's leading supercomputers From the world's most powerful supercomputers to the most energy-efficient systems, Arm technology is helping advance HPC through [...] The post Arm enables the world's fastest and most energy-efficient supercomputers appeared first on Arm Newsroom.

## Data center boom faces challenges beyond water and power

DevFeed: [Data center boom faces challenges beyond water and power](<https://devfeed.tech/articles/data-center-boom-faces-challenges-beyond-water-and-power-136599.md>)

Original publisher: [Read original article](<https://www.networkworld.com/article/4249332/data-center-boom-faces-challenges-beyond-water-and-power.html>)

Author: Andy Patrizio

Published: 2026-11-16T22:41:39Z

Content type: news

Language: en

Sources: [Network World](<https://devfeed.tech/sources/network-world.md>)

Topics: [data centers](<https://devfeed.tech/topics/data-centers.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-data-center](<https://devfeed.tech/tags/ai-data-center.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [artificial-intelligence-ai](<https://devfeed.tech/tags/artificial-intelligence-ai.md>), [artificial-intelligence-data-center-industry-markets-technology-industry](<https://devfeed.tech/tags/artificial-intelligence-data-center-industry-markets-technology-industry.md>), [bbc](<https://devfeed.tech/tags/bbc.md>), [building](<https://devfeed.tech/tags/building.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [challenges](<https://devfeed.tech/tags/challenges.md>), [communities](<https://devfeed.tech/tags/communities.md>), [computing](<https://devfeed.tech/tags/computing.md>), [construction](<https://devfeed.tech/tags/construction.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [data-centers](<https://devfeed.tech/tags/data-centers.md>), [demand](<https://devfeed.tech/tags/demand.md>), [developers](<https://devfeed.tech/tags/developers.md>), [edition-us](<https://devfeed.tech/tags/edition-us.md>), [elections](<https://devfeed.tech/tags/elections.md>), [electricity](<https://devfeed.tech/tags/electricity.md>), [industry](<https://devfeed.tech/tags/industry.md>), [language-en](<https://devfeed.tech/tags/language-en.md>), [markets](<https://devfeed.tech/tags/markets.md>), [power](<https://devfeed.tech/tags/power.md>), [storytype-news](<https://devfeed.tech/tags/storytype-news.md>), [technology-industry](<https://devfeed.tech/tags/technology-industry.md>), [water](<https://devfeed.tech/tags/water.md>)

### AI overview

Rapid data center expansion for AI faces financial and community risks alongside heavy resource use. The excerpt reports that U.S. data centers used about 66 billion liters of water in 2023, and cites a poll in which 69% opposed an AI data center in their neighborhood. Facilities can bring tax revenue and infrastructure upgrades, but typically create only hundreds of permanent jobs, and promised benefits may leave noise or environmental concerns unresolved. The larger financial risk is that costly AI facilities may not earn enough to cover construction and debt: economists estimated major technology companies planned more than $750 billion in U.S. infrastructure and equipment spending by the end of 2026, while data centers would need roughly 2.7 times greater productivity to support an assumed 15% return. If AI demand or revenue falls short, some capacity could become stranded or serve less demanding workloads.

### Source excerpt

There is rising concern around the rapid expansion of data centers to support artificial intelligence (AI) workloads beyond electricity just power consumption and water availability. Experts fear the smaller companies can't afford the hardware bill. The industry's rapid construction boom could create significant financial risks, leave communities with billions of dollars in partially completed infrastructure if demand for high-performance AI computing fails to meet expectations. The concerns come as technology companies commit enormous sums to building new data center capacity as questions are emerging about whether the industry can sustain its rate of investment. According to a report published by BBC Science Focus, the United States had more than 4,000 data centers by the end of 2025, representing approximately 50 gigawatts of capacity. Meanwhile, U.S. data centers consumed approximately 66 billion liters of water in 2023, highlighting the substantial environmental footprint of the industry. These resource requirements have driven considerable opposition to new developments. A recent NBC News poll cited by the publication found that 69% of respondents would oppose construction of an AI data center in their neighborhoods. This is reflected in the recent elections, where pro-data center candidates lost badly while those who publicly opposed for the construction of data centers did well at the ballot box. Data centers can provide economic benefits to the communities where they operate, including increased property tax revenues, permanent employment and infrastructure improvements. Developers sometimes finance upgrades to electrical substations, roads and emergency services to accommodate their facilities. However, data centers are designed to be self-operating and need minimal hands-on maintenance. The number of permanent jobs created by individual data centers is generally relatively modest, often numbering only in the hundreds. Dr. Ben Green, an assistant professor

## BPU-SpikeMM:Exact Matrix Multiplication via Spatio-Temporal Spike Encoding on Real Neuromorphic Hardware

DevFeed: [BPU-SpikeMM:Exact Matrix Multiplication via Spatio-Temporal Spike Encoding on Real Neuromorphic Hardware](<https://devfeed.tech/articles/bpu-spikemm-exact-matrix-multiplication-via-spatio-temporal-spike-encoding-on-real-neuromorphic-hardware-134897.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.18275>)

Author: Rongyuan Guo, Yufeng Zhang, Haotian Wang, Canming Yao, Yuxiang Huan, Wangdong Yang, Kenli Li

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Encoding](<https://devfeed.tech/topics/encoding.md>)

Tags: [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [encoding](<https://devfeed.tech/tags/encoding.md>), [event-driven](<https://devfeed.tech/tags/event-driven.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [matrix](<https://devfeed.tech/tags/matrix.md>), [neuromorphic](<https://devfeed.tech/tags/neuromorphic.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [temporal](<https://devfeed.tech/tags/temporal.md>)

### AI overview

BPU-SpikeMM presents an exact spatio-temporal spike encoding framework for matrix multiplication on real Brain Processing Unit (BPU) hardware. The paper distinguishes mathematical equivalence from native direct-readout equivalence, which avoids additional data recovery, bit-weight reconstruction, or state readout after execution. It describes a spatio-temporal offloading pipeline supporting multiple encoding schemes and evaluates them for correctness, temporal and spatial resource costs, and event activity. The abstract reports native direct-readout equivalence for all evaluated schemes, while their applicability depends on joint trade-offs among those costs and event activity.

### Source excerpt

arXiv:2611.18275v1 Announce Type: new Abstract: Matrix multiplication is the dominant computational kernel in deep learning, yet its growing scale strains conventional von Neumann hardware in energy efficiency and bandwidth. Neuromorphic architectures offer a promising alternative, but mathematical equivalence between matrix multiplication and spiking computation does not guarantee direct execution on real neuromorphic hardware, as the specific hardware paradigm may still require additional data recovery, bit-weight reconstruction, or state readout after neuromorphic execution. This paper presents BPU-SpikeMM (Brain Processing Unit-based Spike Matrix Multiplication), an exact spatio-temporal spike encoding framework for matrix multiplication on real neuromorphic hardware. We prove the exact equivalence of spike-encoded matrix multiplication under the BPU execution paradigm, distinguish mathematical equivalence from native direct-readout equivalence on BPU hardware, and build a spatio-temporal offloading pipeline supporting multiple encoding schemes. Furthermore, we extend the framework to a multi-dimensional encoding design space and evaluate the applicability of different encodings on real BPU hardware in terms of correctness, temporal resource cost, spatial resource cost, and event-driven metrics. The results confirm native direct-readout equivalence across all evaluated encoding schemes and show that their applicability to BPU execution depends on joint trade-offs among temporal cost, spatial resource consumption, and event activity.

## Measured from the feed: the unsustainable growth of AI video generation

DevFeed: [Measured from the feed: the unsustainable growth of AI video generation](<https://devfeed.tech/articles/measured-from-the-feed-the-unsustainable-growth-of-ai-video-generation-135151.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.17938>)

Author: Tristan Coignion, Aur\\'elie Bugeau, Gael Guennebaud, Anne-Laure Ligozat

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Generative AI](<https://devfeed.tech/topics/generative-ai.md>), [Green Software](<https://devfeed.tech/topics/green-software.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-cy](<https://devfeed.tech/tags/cs-cy.md>), [cs-et](<https://devfeed.tech/tags/cs-et.md>), [emissions](<https://devfeed.tech/tags/emissions.md>), [growth](<https://devfeed.tech/tags/growth.md>), [video](<https://devfeed.tech/tags/video.md>), [video-generation](<https://devfeed.tech/tags/video-generation.md>)

### AI overview

Providers rarely disclose video volumes or energy needs, so the authors measured yearly output from two commercial text-to-video platforms and modeled one service's footprint from training through storage. Their projections indicate that continued growth could raise emissions nearly 75-fold by 2030 despite efficiency improvements; holding usage at 2026 levels would let those improvements reduce emissions by about two-fifths. The estimates are a conservative lower bound, covering one service among many.

### Source excerpt

arXiv:2611.17938v1 Announce Type: new Abstract: Text-to-video generation runs at commercial scale, but its broader environmental impacts remain, to our knowledge, understudied. Providers rarely disclose video volumes or energy needs; we measured the yearly video output of two commercial platforms and built a prospective model of such services' absolute footprint from training to storage. Our projections suggest that reducing emissions depends on both limiting growth in video generation and achieving substantial efficiency gains. Under continued growth, emissions rise nearly seventy-five-fold by 2030 despite technological improvements. When usage remains at 2026 levels, the same improvements reduce emissions by about two-fifths. These estimates are a conservative lower bound: they cover one AI service among many; other services and uses would push the total higher.

## Pruning Extensions and Efficiency Trade-Offs for Sustainable Time Series Classification

DevFeed: [Pruning Extensions and Efficiency Trade-Offs for Sustainable Time Series Classification](<https://devfeed.tech/articles/pruning-extensions-and-efficiency-trade-offs-for-sustainable-time-series-classification-134971.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2604.07953>)

Author: Raphael Fischer, Angus Dempster, Sebastian Buschj\\"ager, Matthias Jakobs, Urav Maniar, Geoffrey I. Webb

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Time Series](<https://devfeed.tech/topics/time-series.md>), [Post-training optimization](<https://devfeed.tech/topics/post-training-optimization.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Hyperparameter optimization](<https://devfeed.tech/topics/hyperparameter-optimization.md>)

Tags: [classification](<https://devfeed.tech/tags/classification.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [energy-consumption](<https://devfeed.tech/tags/energy-consumption.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [extensions](<https://devfeed.tech/tags/extensions.md>), [pruning](<https://devfeed.tech/tags/pruning.md>), [time-series](<https://devfeed.tech/tags/time-series.md>), [trade](<https://devfeed.tech/tags/trade.md>)

### AI overview

The paper evaluates time series classifiers across more than 4,000 configurations, 20 MONSTER datasets, 10 methods, and three compute setups to examine resource-quality trade-offs. It proposes a bounded-error pruning strategy for Hydra and Quant, plus Hydrant, an extension combining interval representations, random convolutions, and dictionary learning. The authors report that pruning reduces energy consumption by up to 70% with only marginal losses in predictive quality; results are framed as dependent on data, model, and hardware.

### Source excerpt

arXiv:2604.07953v2 Announce Type: replace-cross Abstract: Time series classification (TSC) at scale requires balancing predictive performance against resource consumption, which is subject to the given data, model, and hardware. This paper introduces a multi-dimensional evaluation framework that allows for explicitly benchmarking and exploring the resource-versus-quality trade-offs in TSC. To improve efficiency, we present a pruning strategy with bounded error for leading hybrid classifiers - Hydra and Quant - and moreover propose Hydrant, a novel extension that integrates interval representations with random convolutions and dictionary learning. With over 4000 evaluated configurations across 20 MONSTER datasets, 10 methods, and three compute setups, we systematically analyze how model design, hyperparameters, and hardware choices impact practical TSC performance. Our results show that pruning can significantly reduce energy consumption by up to 70% while only marginally lowering predictive quality. As such, our contributed evaluation framework, methodological extensions, experimental results, and software repository advance TSC toward sustainable and reproducible practice.

## The Day Microsoft 365 Went Down Worldwide: The Environmental Cost of Digital Dependence

DevFeed: [The Day Microsoft 365 Went Down Worldwide: The Environmental Cost of Digital Dependence](<https://devfeed.tech/articles/the-day-microsoft-365-went-down-worldwide-the-environmental-cost-of-digital-dependence-134012.md>)

Original publisher: [Read original article](<https://joho-todai.com/en/microsoft-365-outage-digital-dependency-environmental-cost/>)

Author: Toru Misaki

Published: 2026-11-13T23:28:23Z

Content type: opinion

Language: en

Sources: [Joho no Todai](<https://devfeed.tech/sources/joho-no-todai.md>)

Topics: [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [clean energy data centers](<https://devfeed.tech/topics/clean-energy-data-centers.md>), [microsoft 365](<https://devfeed.tech/topics/microsoft-365.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-data-center-demand](<https://devfeed.tech/tags/ai-data-center-demand.md>), [carbon](<https://devfeed.tech/tags/carbon.md>), [choice](<https://devfeed.tech/tags/choice.md>), [cost](<https://devfeed.tech/tags/cost.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [data-center-electricity](<https://devfeed.tech/tags/data-center-electricity.md>), [digital](<https://devfeed.tech/tags/digital.md>), [electricity](<https://devfeed.tech/tags/electricity.md>), [electronic](<https://devfeed.tech/tags/electronic.md>), [environment](<https://devfeed.tech/tags/environment.md>), [fees](<https://devfeed.tech/tags/fees.md>), [government](<https://devfeed.tech/tags/government.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [microsoft-365](<https://devfeed.tech/tags/microsoft-365.md>), [software](<https://devfeed.tech/tags/software.md>), [subscription](<https://devfeed.tech/tags/subscription.md>), [switching](<https://devfeed.tech/tags/switching.md>), [waste](<https://devfeed.tech/tags/waste.md>), [work](<https://devfeed.tech/tags/work.md>)

### AI overview

The November 13 Microsoft 365 outage exposed how dependence on one provider can disrupt organizations, while its environmental and switching costs remain outside subscription fees. The article cites IEA estimates that data centers used 415 TWh of electricity in 2024 and could use 945 TWh by 2030; Microsoft reported a 25% year-over-year rise in total 2025 emissions, despite falling emissions per operation. It also links software support requirements to hardware replacement and e-waste, while clarifying that global e-waste figures are not attributable to Microsoft alone. UK competition scrutiny of Microsoft's workplace ecosystem and cloud switching barriers highlights a related concern: lock-in may constrain provider choice, including lower-emission options, though alternatives are not necessarily greener. The article argues sustainability requires portable data, open formats, repairable hardware, transparent emissions reporting, and contracts that permit switching.

### Source excerpt

Microsoft 365 went down on November 13, disrupting even the Japanese government's work environment. Burdens that software subscription fees alone obscure--environmental impacts from data center electricity consumption, electronic waste, and barriers to switching--are accumulating behind dependence on a single company.

## Energy Aware Optimization of Non-Deterministic Task-Based Intermittent Computing Nodes with Thimble

DevFeed: [Energy Aware Optimization of Non-Deterministic Task-Based Intermittent Computing Nodes with Thimble](<https://devfeed.tech/articles/energy-aware-optimization-of-non-deterministic-task-based-intermittent-computing-nodes-with-thimble-131583.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.15681>)

Author: John Madden, Morgan Masters, Alec Levy, Steve McGuire, Colleen Josephson

Published: 2026-11-13T05:00:00Z

Content type: comparison

Language: en

Sources: [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [scheduling](<https://devfeed.tech/topics/scheduling.md>), [Combinatorial optimization](<https://devfeed.tech/topics/combinatorial-optimization.md>)

Tags: [capacitor](<https://devfeed.tech/tags/capacitor.md>), [capacitors](<https://devfeed.tech/tags/capacitors.md>), [computing](<https://devfeed.tech/tags/computing.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [data](<https://devfeed.tech/tags/data.md>), [demand](<https://devfeed.tech/tags/demand.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [energy](<https://devfeed.tech/tags/energy.md>), [energy-storage](<https://devfeed.tech/tags/energy-storage.md>), [leak](<https://devfeed.tech/tags/leak.md>), [mapping](<https://devfeed.tech/tags/mapping.md>), [markov](<https://devfeed.tech/tags/markov.md>), [markov-chain](<https://devfeed.tech/tags/markov-chain.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [run](<https://devfeed.tech/tags/run.md>), [sensor](<https://devfeed.tech/tags/sensor.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [static](<https://devfeed.tech/tags/static.md>), [throughput](<https://devfeed.tech/tags/throughput.md>), [trace](<https://devfeed.tech/tags/trace.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

Thimble maps tasks to energy-storage capacitors at runtime for intermittent nodes whose data-dependent workloads can change. It models task transitions as a Markov chain and weighs capacitor leakage, task energy demand, and likely future tasks when replanning. When deployment resembles training, its single design-time solve comes within 0.3-4% of LEACS's throughput. Under drift on the lowest-power source studied, LEACS's static plan can stall or fail to wake, while Thimble achieves 22-29% higher throughput at the reference workload. Across 25 workload profiles on that source, depth-based replanning leads LEACS by 22-42%; fixed-interval remapping matches or beats the static map at 21 points, by up to 7%.

### Source excerpt

arXiv:2611.15681v1 Announce Type: new Abstract: Intermittent computing nodes run on harvested energy, which they store in capacitors and spend in short bursts of computation. Choosing those capacitors is a trade-off: large capacitors can power large tasks but leak more current, wasting harvested energy, while small capacitors leak less but cannot power large tasks. Recent work, LEACS, showed that selecting capacitors and mapping tasks to them at compile time improves inference throughput. However, LEACS optimizes the mapping for a predetermined workload, whereas sensor nodes often run data-dependent workloads in which the next task depends on the outcome of earlier ones. We present Thimble, an energy storage system that maps a node's tasks to capacitors at runtime. Thimble models the workload as a Markov chain over tasks and assigns each task to a capacitor by weighing capacitor leakage, the task's energy demand and, when it replans, the likelihood of each future task. We compare Thimble against a reimplementation of LEACS, which ranks every static plan by simulation. When the training trace resembles deployment, Thimble's single design-time solve comes within 0.3--4% of LEACS's throughput. When the deployment conditions drift from the characteristics of the training set, LEACS's throughput-focused static plans can fail sharply. On the lowest-power source we consider, LEACS's plan stalls on deployment traces that harvest less than its training trace, and on some never wakes; Thimble avoids this, with 22--29% higher throughput at the reference workload. A sensitivity sweep over 25 workload profiles on that source shows that Thimble with depth replanning leads LEACS by 22--42% throughout, and what run-time remapping adds: replanning at a fixed interval matches or improves on the static map at 21 of 25 points, by up to 7%.

## Sustainable Federated Fine-Tuning with LoRA: Environmental Impact and Performance Footprint

DevFeed: [Sustainable Federated Fine-Tuning with LoRA: Environmental Impact and Performance Footprint](<https://devfeed.tech/articles/sustainable-federated-fine-tuning-with-lora-environmental-impact-and-performance-footprint-131590.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.16856>)

Author: Pedro Augusto Londe Ribeiro, Rodrigo Moreira, Larissa Ferreira Rodrigues Moreira, Joberto S. B. Martins, Tereza C. Carvalho, Fl\\'avio de Oliveira Silva

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>), [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [lora](<https://devfeed.tech/topics/lora.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [peft](<https://devfeed.tech/topics/peft.md>), [Green Software](<https://devfeed.tech/topics/green-software.md>)

Tags: [a100](<https://devfeed.tech/tags/a100.md>), [carbon](<https://devfeed.tech/tags/carbon.md>), [costs](<https://devfeed.tech/tags/costs.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [emissions](<https://devfeed.tech/tags/emissions.md>), [energy](<https://devfeed.tech/tags/energy.md>), [energy-consumption](<https://devfeed.tech/tags/energy-consumption.md>), [environmental-impact](<https://devfeed.tech/tags/environmental-impact.md>), [federated](<https://devfeed.tech/tags/federated.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llms](<https://devfeed.tech/tags/llms.md>), [local](<https://devfeed.tech/tags/local.md>), [lora](<https://devfeed.tech/tags/lora.md>), [models](<https://devfeed.tech/tags/models.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [peft](<https://devfeed.tech/tags/peft.md>), [performance](<https://devfeed.tech/tags/performance.md>), [rtx](<https://devfeed.tech/tags/rtx.md>)

### AI overview

In the evaluated federated LoRA fine-tuning setup, using fewer local steps achieved comparable accuracy on NVIDIA A40 and RTX 6000 Ada GPUs while consuming roughly 6-8 times less energy and delivering higher accuracy per kWh. The study tested 135M and 360M models at 10 or 100 local steps across A40, RTX 6000 Ada, and A100 GPUs, tracking energy and estimated CO₂-equivalent emissions with CodeCarbon. Within those tested ranges, local steps accounted for 84.8-91.2% of the variation in energy and emissions. The findings support smaller local-step budgets in this setting; the excerpt does not report the same accuracy-energy result for A100.

### Source excerpt

arXiv:2611.16856v1 Announce Type: cross Abstract: Federated Learning (FL) with Parameter-Efficient Fine-Tuning (PEFT) enables efficient adaptation of Large Language Models (LLMs), yet its energy and carbon costs across hardware remain poorly understood. We evaluated federated Low-Rank Adaptation (LoRA) fine-tuning on NVIDIA A40, RTX 6000 Ada, and A100 GPUs using a $2^2$ full factorial design with 135M/360M models and 10/100 local steps. CodeCarbon tracked energy and estimated CO$_2$-equivalent emissions. Descriptive allocation of variation assigned 84.8-91.2% of energy and emissions variation to local steps within these ranges. On A40 and RTX 6000 Ada, configurations with fewer steps achieved comparable accuracy with roughly 6-8-fold lower energy consumption and higher accuracy per kWh. These findings support smaller local budgets for energy-efficient federated fine-tuning within the evaluated setting.

## Designing Edge AI SoCs At Near-Threshold: Why Foundation IP Can Help You Break Down The Power Wall

DevFeed: [Designing Edge AI SoCs At Near-Threshold: Why Foundation IP Can Help You Break Down The Power Wall](<https://devfeed.tech/articles/designing-edge-ai-socs-at-near-threshold-why-foundation-ip-can-help-you-break-down-the-power-wall-130164.md>)

Original publisher: [Read original article](<https://semiengineering.com/designing-edge-ai-socs-at-near-threshold-why-foundation-ip-can-help-you-break-down-the-power-wall/>)

Author: Andrew Appleby

Published: 2026-11-12T08:05:04Z

Content type: article

Language: en

Sources: [Semiconductor Engineering](<https://devfeed.tech/sources/semiconductor-engineering.md>)

Topics: [tinyml](<https://devfeed.tech/topics/tinyml.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>)

Tags: [3nm](<https://devfeed.tech/tags/3nm.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-networking](<https://devfeed.tech/tags/ai-networking.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [area](<https://devfeed.tech/tags/area.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [behavior](<https://devfeed.tech/tags/behavior.md>), [benefit](<https://devfeed.tech/tags/benefit.md>), [blocks](<https://devfeed.tech/tags/blocks.md>), [blogs-lphp](<https://devfeed.tech/tags/blogs-lphp.md>), [edge-ai](<https://devfeed.tech/tags/edge-ai.md>), [energy](<https://devfeed.tech/tags/energy.md>), [foundation-ip](<https://devfeed.tech/tags/foundation-ip.md>), [ip](<https://devfeed.tech/tags/ip.md>), [ip-and-lp-in-socs](<https://devfeed.tech/tags/ip-and-lp-in-socs.md>), [logic-libraries](<https://devfeed.tech/tags/logic-libraries.md>), [low-power-high-performance](<https://devfeed.tech/tags/low-power-high-performance.md>), [low-voltage-design](<https://devfeed.tech/tags/low-voltage-design.md>), [near-threshold](<https://devfeed.tech/tags/near-threshold.md>), [performance](<https://devfeed.tech/tags/performance.md>), [power](<https://devfeed.tech/tags/power.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [scaling](<https://devfeed.tech/tags/scaling.md>), [synopsys](<https://devfeed.tech/tags/synopsys.md>), [voltage](<https://devfeed.tech/tags/voltage.md>), [voltage-scaling](<https://devfeed.tech/tags/voltage-scaling.md>)

### AI overview

Near-threshold operation around 0.4-0.5 V can improve energy per operation for edge AI SoCs, but delay and variability rise and noise margins shrink, making predictable design a cross-stack challenge. The excerpt says success depends on carefully selected and silicon-validated Foundation IP, accurate modeling, and close alignment with EDA flows. Memory can set system Vmin; a dual-rail architecture is cited as decoupling logic and memory voltage, with a 38 Mb/mm² SRAM validated at 380/540 mV in 3nm FinFET. It also describes different memory needs across MAC tiles, activation units, and shared buffers, and notes that I/O reliability can constrain system Vmin. In one high-volume AI networking SoC engagement, tuned low-voltage logic cells, custom level shifters, a dual-rail memory compiler, selected PVT corners, and application engineering support were used; the stated result was predictable execution without schedule or silicon risk. No quantitative power or performance result is provided.

### Source excerpt

Traditional voltage scaling is no longer sufficient to deliver real-time performance within tight energy and thermal limits. The post Designing Edge AI SoCs At Near-Threshold: Why Foundation IP Can Help You Break Down The Power Wall appeared first on Semiconductor Engineering.

## Mobility-Aware Task Offloading in Mobile Edge Computing-Empowered Space-Air-Ground Integrated Internet-of-Things Networks

DevFeed: [Mobility-Aware Task Offloading in Mobile Edge Computing-Empowered Space-Air-Ground Integrated Internet-of-Things Networks](<https://devfeed.tech/articles/mobility-aware-task-offloading-in-mobile-edge-computing-empowered-space-air-ground-integrated-internet-of-things-networks-127846.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.14102>)

Author: Minghao Sun, Jinghuai Xiong, Xiaoli Chu, Zitian Zhang

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [Internet of things](<https://devfeed.tech/topics/iot.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Retrieval Augmented Generation (RAG)](<https://devfeed.tech/topics/retrieval-augmented-generation-rag.md>), [Vectors & LLMs](<https://devfeed.tech/topics/vectors-llms.md>)

Tags: [algorithm](<https://devfeed.tech/tags/algorithm.md>), [communication](<https://devfeed.tech/tags/communication.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [edge-computing](<https://devfeed.tech/tags/edge-computing.md>), [energy](<https://devfeed.tech/tags/energy.md>), [energy-consumption](<https://devfeed.tech/tags/energy-consumption.md>), [internet-of-things](<https://devfeed.tech/tags/internet-of-things.md>), [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [mobile](<https://devfeed.tech/tags/mobile.md>), [mobility](<https://devfeed.tech/tags/mobility.md>), [networks](<https://devfeed.tech/tags/networks.md>), [offloading](<https://devfeed.tech/tags/offloading.md>), [planning](<https://devfeed.tech/tags/planning.md>), [rag](<https://devfeed.tech/tags/rag.md>), [relay](<https://devfeed.tech/tags/relay.md>), [resource](<https://devfeed.tech/tags/resource.md>), [satellites](<https://devfeed.tech/tags/satellites.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [space](<https://devfeed.tech/tags/space.md>)

### AI overview

The proposed RLD method jointly plans a UAV's trajectory and splits tasks across the UAV's edge server and multiple LEO satellites to reduce total system energy, subject to task, communication, computation, flight-region, and satellite service-window constraints. The non-convex problem is divided into trajectory planning, handled by a RAG-assisted, LLM-enhanced DDPG algorithm, and task splitting/resource allocation, which the abstract says is convex and solvable with CVXPY for a fixed trajectory. A weighted retrieval score balances device mobility-pattern similarity with the energy performance of past UAV trajectories. Simulations reportedly show task completion in previously unseen scenarios with less total energy than benchmark algorithms; the abstract gives no numerical results or benchmark details.

### Source excerpt

arXiv:2611.14102v1 Announce Type: new Abstract: Mobile edge computing (MEC)-empowered spaceair-ground integrated Internet-of-Things (IoT) networks (SAGIINs) provide a promising solution for computation-intensive tasks generated by mobile IoT devices (IDs). However, task offloading in SAGIINs is challenging due to time-varying IDto-unmanned aerial vehicle (UAV) channels, intermittent LEO satellite service availability, and the coupling among UAV trajectory, communication, computation, and energy decisions. This paper investigates an energy-efficient task offloading and UAV trajectory planning problem in an MEC-empowered SAGIIN, where one UAV acts as both an aerial relay and an MEC server, and multiple LEO satellites provide supplementary computing resources. The UAV trajectory planning, task splitting, and computing resource allocation are jointly optimized to minimize the total system energy consumption under task completion, communication, computation, flight region, and satellite service window constraints. The resulting non-convex problem is decomposed into a UAV trajectory planning subproblem and a task splitting and computing resource allocation subproblem. To solve the former, a retrieval-augmented generation (RAG)-assisted large language model (LLM)-enhanced deep deterministic policy gradient (DDPG) algorithm, termed RLD, is proposed. The latter is shown to be convex and can be solved by using CVXPY for a given UAV trajectory. A weighted retrieval score is further designed to balance ID mobility-pattern similarity and historical UAV trajectory energy performance. Simulation results show that RLD completes all computational tasks in previously unseen test scenarios while consuming less total energy than the benchmark algorithms.

## Useful Work Can Increase Energy Use: Demand Elasticity and Proof of Useful Work

DevFeed: [Useful Work Can Increase Energy Use: Demand Elasticity and Proof of Useful Work](<https://devfeed.tech/articles/useful-work-can-increase-energy-use-demand-elasticity-and-proof-of-useful-work-128894.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.13131>)

Author: Isaiah Hull, Or Sattath

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [proof of work](<https://devfeed.tech/topics/proof-of-work.md>), [Sustainable Computing](<https://devfeed.tech/topics/sustainable-computing.md>)

Tags: [2017](<https://devfeed.tech/tags/2017.md>), [2026](<https://devfeed.tech/tags/2026.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [bitcoin](<https://devfeed.tech/tags/bitcoin.md>), [can](<https://devfeed.tech/tags/can.md>), [concrete](<https://devfeed.tech/tags/concrete.md>), [costs](<https://devfeed.tech/tags/costs.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-gt](<https://devfeed.tech/tags/cs-gt.md>), [demand](<https://devfeed.tech/tags/demand.md>), [discount](<https://devfeed.tech/tags/discount.md>), [effective](<https://devfeed.tech/tags/effective.md>), [elastic](<https://devfeed.tech/tags/elastic.md>), [electricity](<https://devfeed.tech/tags/electricity.md>), [energy](<https://devfeed.tech/tags/energy.md>), [introduced](<https://devfeed.tech/tags/introduced.md>), [jevons-paradox](<https://devfeed.tech/tags/jevons-paradox.md>), [less](<https://devfeed.tech/tags/less.md>), [useful](<https://devfeed.tech/tags/useful.md>), [work](<https://devfeed.tech/tags/work.md>)

### AI overview

Proof-of-Useful-Work (PoUW) can increase total electricity use when demand for the useful computation is elastic: mining subsidies lower its effective price, and the resulting increase in demand can outweigh the electricity saved. With inelastic demand, the model predicts lower total consumption, but PoUW-related electricity use does not disappear because the subsidy funds work that otherwise would not have been done. The result is proved under assumptions of electricity-only costs, no PoUW overhead and an unchanged mining subsidy; the analysis extends to less restrictive settings with adjusted thresholds.

### Source excerpt

arXiv:2611.13131v1 Announce Type: new Abstract: A serious criticism of Bitcoin's Proof-of-Work is its enormous electricity consumption. Proof-of-Useful-Work (PoUW), introduced by Ball, Rosen, Sabin, and Vasudevan (2017), in which mining performs useful computations, was suggested as an alternative to reduce electricity expenditure, with matrix multiplication (MatMul) as a recent concrete instantiation by Komargodski and Weinstein (2026). Perhaps surprisingly, we show that switching to PoUW may increase the total electricity consumption, echoing Jevons' paradox. Intuitively, the PoUW mechanism acts as a subsidy that lowers the effective price of the useful work. Recall that demand is elastic if a discount of x percent in the price leads to an increase of more than x percent in the quantity demanded. If demand is elastic, the resulting increase in quantity demanded leads to a net increase in total electricity consumption; if inelastic, total electricity consumption decreases. Even then, the electricity consumption due to PoUW never vanishes, since the mining subsidy is spent on useful work that would not have been produced without the subsidy. This result is proved in a simple model with electricity-only costs, no PoUW overhead, and an unchanged mining subsidy. The analysis extends to less restrictive settings with adjusted thresholds.

[Next page](<https://devfeed.tech/topics/sustainable-computing.md?cursor=WyIyMDI2LTExLTExVDA1OjAwOjAwKzAwOjAwIiwgImQxZTc3MjEzLWNmZjAtNDFjOS04YmUxLWU3Mjk1ZDA0MTA1MSJd>)