Energy Digest

Daily Summaries & Key Takeaways of Power & Energy Updates
Powered by Llama 3.2
Last Updated: January 14, 2026 at 06:01 PM
23
News & Articles
7
Research Papers
1

EIA Predicts Sustained Power Growth in 2026 and 2027

Summary

The US Energy Information Administration (EIA) is predicting sustained power growth in 2026 and 2027, with the highest demand growth in a quarter century, largely due to the proliferation of data centers. Solar power output is expected to increase significantly, while natural gas will provide a slightly smaller percentage of U.S. electricity, and coal will decline further. The predicted growth rate for solar power will be over 20% in both 2026 and 2027.
Read Full Article →

Key Takeaways

  • The EIA predicts a 1-3% annual power demand growth from 2026 to 2027, the first time since 2007 that power demand has increased four years in a row.
  • Solar power output is expected to increase by more than 20% in both 2026 and 2027, making it 10% of U.S. power generation by the end of 2027.
  • The share of natural gas in the US power supply is expected to decrease slightly from 40-42% in 2024 to 39% in 2026, while coal's share will decline further from 17% in 2025 to 15% in 2026 and 2027.
2

Bank Consortium Closes $160M Financing for Spanish Wind, Solar Portfolio

Summary

A bank consortium consisting of NORD/LB, Rabobank, and Siemens Bank has provided €135 million ($160 million) in financing for a portfolio of five renewable energy projects in Aragón, Spain. The projects, owned by ENCAVIS, are comprised of three wind farms and two solar farms with a total capacity of 199 MW. This deal highlights the growing demand for renewable energy financing.
Read Full Article →

Key Takeaways

  • A €135 million ($160 million) financing package has been secured for five renewable energy projects in Aragón, Spain.
  • The portfolio consists of three wind farms and two solar farms with a combined capacity of 199 MW.
  • ENCAVIS, the owner of the portfolio, is a leading pan-European independent power producer.
3

How AI’s Energy Challenge is Becoming its Innovation Engine

Summary

The article discusses the growing energy demands of artificial intelligence (AI) systems, which are scaling in size and sophistication at an unprecedented rate. As AI continues to innovate, it is also becoming its own innovation engine, with its energy consumption accelerating faster than any previous wave of digital innovation. This shift highlights the need for more efficient and sustainable energy solutions to support the rapid growth of AI.
Read Full Article →

Key Takeaways

  • The increasing size and sophistication of AI models and workloads are driving up their energy consumption at an accelerated rate.
  • As AI becomes a major driver of innovation, it is also becoming its own primary source of energy demand, highlighting the need for more efficient solutions.
  • The article suggests that the growing energy challenges posed by AI are creating new opportunities for innovation and development in the field.
4

This battery is about to change the world in 3 months, or make this guy a fool

Summary

The article discusses Donut Lab's announcement of a 400 Wh/kg solid-state battery cell that claims to last for 100 years, charge in 5 minutes, and have a 100,000-cycle life. If true, this would be a groundbreaking development in the electric vehicle industry and energy storage. The article questions the credibility of the claim and explores potential implications.
Read Full Article →

Key Takeaways

  • Donut Lab's battery cell claims to achieve a high energy density of ~400 Wh/kg, which is considered a significant breakthrough if confirmed.
  • The claimed 5-minute charge time and 100,000-cycle life are impressive, but battery durability and charging speeds can be affected by various factors, including the type of application and usage patterns.
  • If Donut Lab's claims hold up, this could have far-reaching implications for the electric vehicle industry and energy storage market, potentially changing the game in terms of range, efficiency, and affordability.
5

The Nuclear Industry’s Race Against the Clock: EPRI Experts on Fleet Optimization, SMRs, and What’s Next

Summary

The nuclear industry is facing a significant challenge in meeting growing electricity demand with carbon-free, firm power. The sector's aging fleet and lack of new capacity are creating a "timing problem" that must be addressed to meet the increasing energy needs driven by data centers, AI deployment, and industrial reshoring. Nuclear power is seen as a solution, but can it deliver capacity quickly enough?
Read Full Article →

Key Takeaways

  • The nuclear industry is under pressure to increase capacity to meet growing electricity demand, particularly for carbon-free, firm power.
  • The use of Small Modular Reactors (SMRs) may play a key role in addressing the sector's capacity needs and meeting emerging energy demands.
  • Fleet optimization and revitalization of existing nuclear plants are also crucial strategies for the industry to remain competitive and deliver on its promise of low-carbon energy.
6

Vistra to Bolster Gas-Fired Fleet by 5.5 GW With $4B Cogentrix Acquisition

Summary

Vistra Corp. has acquired Cogentrix Energy in a $4 billion deal, expanding its gas-fired generation fleet by 5.5 GW across three regional power markets. This acquisition aims to bolster Vistra's presence in PJM Interconnection, ISO New England, and the Electric Reliability Council of Texas (ERCOT). The expansion is expected to enhance Vistra's overall energy production capacity.
Read Full Article →

Key Takeaways

  • Vistra Corp. has acquired Cogentrix Energy in a $4 billion deal, expanding its gas-fired generation fleet by 5.5 GW.
  • The acquisition will bolster Vistra's presence in PJM Interconnection, ISO New England, and the Electric Reliability Council of Texas (ERCOT) power markets.
  • This expansion aims to enhance Vistra's overall energy production capacity across the three regional markets.
7

Dutch long-duration energy storage project targets sub-€50/kWh costs with innovative technologies

Summary

A Dutch consortium is developing a long-duration energy storage project, SLDBatt, with the goal of reducing costs below €50/kWh for storing renewable electricity. The project brings together innovative battery technologies and involves national partners to accelerate knowledge sharing and collaboration. With over €30 million in funding, the project aims to address challenges related to grid congestion.
Read Full Article →

Key Takeaways

  • The SLDBatt project targets a cost reduction of sub-€50/kWh for long-duration energy storage solutions using innovative TRL 7 battery technologies.
  • The project involves a consortium of Dutch start-ups and internationally operating energy companies, as well as industrial partners, to develop and deploy new battery technologies.
  • Battery Competence Cluster NL will act as a national catalyst for knowledge sharing and collaboration across the battery value chain in the Netherlands.
8

India deploys 37.9 GW of solar in 2025

Summary

India added a record 37.9 GW of solar and 6.3 GW of wind capacity in calendar year 2025, marking its highest annual renewable energy additions to date. The country's cumulative installed renewable energy capacity reached 258 GW as of December 31, 2025, with solar accounting for about 53% of the total. This growth was driven by accelerated construction and commissioning of projects, as well as a rise in domestic manufacturing capacity.
Read Full Article →

Key Takeaways

  • India added a record 37.9 GW of solar capacity in 2025, with a significant increase of 54.6% year on year.
  • Utility-scale solar accounted for the majority of new installations (28.6 GW), driven by the commissioning of long-delayed projects and the open-access segment.
  • The country's cumulative installed renewable energy capacity reached 258 GW as of December 31, 2025, with solar accounting for about 53% of the total.
  • Domestic manufacturing capacity also played a crucial role in supporting deployment, with cumulative module and cell production capacity exceeding 200 GW.
9

NewYork GreenCloud Acquires Buena Vista Biomass Power Facility to Launch First Carbon-Negative AI Factory

Summary

NewYork GreenCloud has acquired the Buena Vista Biomass Power facility in California, which will be transformed into a carbon-negative AI factory that integrates biomass-to-pyrolysis energy systems and liquid-cooled AI compute. The acquisition marks the beginning of a national platform for NYGC to launch its AI technology. This move is expected to have significant implications for the company's growth and expansion.
Read Full Article →

Key Takeaways

  • NewYork GreenCloud has acquired the Buena Vista Biomass Power facility in California, which will be repurposed as a carbon-negative AI factory.
  • The facility will integrate biomass-to-pyrolysis energy systems with behind-the-meter liquid-cooled AI compute to create a sustainable and environmentally friendly computing solution.
  • This acquisition is part of NYGC's efforts to launch its first large-scale carbon-negative AI Factory, marking the beginning of a national platform for its technology.
10

Honda reveals an affordable, solar-powered camper that’s built with EVs in mind

Summary

Honda has unveiled a solar-powered travel trailer called the Base Station Prototype, designed to be towed by various vehicles including electric vehicles (EVs). The trailer is intended to be "competitively priced," suggesting an affordable option for outdoor enthusiasts. It features a built-in EV system, taking into account the specific needs of EV owners.
Read Full Article →

Key Takeaways

  • The Honda Base Station Prototype is a solar-powered travel trailer that can be towed by nearly any vehicle, including electric vehicles (EVs).
  • The trailer is designed to be "competitively priced," implying an affordable option for outdoor enthusiasts.
  • The Base Station Prototype's design takes into account the specific needs of EV owners, suggesting integration with the vehicle's electrical system.
11

Illinois AG files objections to ComEd data center agreements at FERC

Summary

The Illinois Attorney General has filed objections to agreements between ComEd and federal regulators (FERC) regarding the use of data centers, citing concerns that payment models do not guarantee sufficient revenue to cover transmission costs. The state argues that these agreements may not provide enough financial certainty for the grid operator. The dispute revolves around the payment models used in these agreements.
Read Full Article →

Key Takeaways

  • The Illinois Attorney General has filed objections to ComEd's data center agreements with FERC, citing concerns about the reliability of the payment models.
  • The state alleges that these payment models do not guarantee enough revenue to cover transmission costs, potentially leaving ComEd vulnerable to financial losses.
  • The dispute highlights issues related to grid financing and revenue certainty, which are critical for maintaining a stable and reliable power system.
12

Data Center Developer, Major Investment Group Plan Gigawatts of New Capacity

Summary

A data center development company, backed by former Amazon Web Services (AWS) executives, has partnered with a major investment group to launch a platform for building large-scale hyperscale data centers in North America. This move aims to capitalize on the growing demand for cloud computing and data storage services. The project plans to generate gigawatts of new capacity.
Read Full Article →

Key Takeaways

  • A data center development company and an investment group are launching a platform for hyperscale data centers in North America, targeting growing demand for cloud computing and data storage.
  • The partnership brings together expertise from former Amazon Web Services (AWS) executives, indicating a strong understanding of the market needs for large-scale data centers.
  • The project plans to generate significant capacity, with no specific details provided on the scale or timeline of the development.
13

Coal-Powered Thermal Generation Declines In China & India, But Surges In US

Summary

The article reports that coal-powered thermal generation has declined in China and India due to increased renewable energy adoption, but has surprisingly seen a surge in the US. This trend suggests that the shift towards cleaner energy sources is not uniform across all regions. The disparity may be attributed to various factors, including policy differences, economic conditions, or regional energy demand patterns.
Read Full Article →

Key Takeaways

  • Coal-powered thermal generation declined in China and India last year due to increased renewable energy adoption.
  • In contrast, coal-powered thermal generation has surged in the US, indicating a divergence in energy transition trajectories between regions.
  • The trend highlights the need for more nuanced analysis of energy policy and transition strategies to account for regional differences in energy demand and production patterns.
14

Xcel defends storage ownership in distributed capacity pilot

Summary

Xcel Energy, a Minnesota utility, is defending its ownership of storage batteries in its distributed capacity pilot program, CapacityConnect. The proposal has undergone modifications, but Xcel maintains its claim to the batteries as it expands the program. Stakeholders have argued for more competition in the ownership of these energy assets.
Read Full Article →

Key Takeaways

  • Xcel Energy is maintaining its ownership of storage batteries in its CapacityConnect program despite stakeholder opposition.
  • The utility's stance on battery ownership has been a subject of debate, with some arguing for increased competition among owners.
  • Modifications have been made to the proposal, but the specifics of these changes are not provided in the article.
15

Eolian and CyrusOne team up to deploy Texas data centre at site of 100MW BESS

Summary

CyrusOne and Eolian have partnered to deploy a 200MW data centre campus at a 100MW BESS site in Fort Worth, Texas, marking a significant expansion of the data centre developer's presence in the region. The project aims to reduce reliance on traditional power sources by leveraging the stored energy from the grid-scale battery energy storage system (BESS). This partnership highlights the growing demand for sustainable and renewable energy solutions in the data centre sector.
Read Full Article →

Key Takeaways

  • The partnership between CyrusOne and Eolian demonstrates a growing trend of collaboration between data centre developers and independent power producers to deploy large-scale, sustainable energy solutions.
  • The deployment of a 200MW data centre campus at a 100MW BESS site in Fort Worth, Texas, underscores the increasing demand for renewable energy sources in the region.
  • This project is likely to reduce reliance on traditional power sources, contributing to a more sustainable and environmentally friendly data centre infrastructure.
16

HD Renewable Energy puts energy storage at centre of 2026 ‘overseas expansion year’ strategy

Summary

HD Renewable Energy is planning an overseas expansion in 2026, with a focus on energy storage playing a key role in its strategy. The company is a Taiwan-headquartered developer and services provider. This expansion aims to capitalize on the growing demand for renewable energy.
Read Full Article →

Key Takeaways

  • HD Renewable Energy is targeting further overseas expansion in 2026, indicating a significant growth plan.
  • Energy storage will be a major component of this expansion, suggesting that the company sees it as a crucial element in its strategy.
  • The focus on energy storage implies that HDRE aims to provide innovative solutions for renewable energy projects, which could help drive its growth and success.
17

Neoen’s battery to pioneer grid-forming services on French transmission network

Summary

Neoen's 92 MW/183 MWh Breizh Big Battery is part of a trial with French transmission system operator RTE to provide grid-forming services on France's transmission network. This marks a significant step towards enhancing the stability of the French power grid, as traditional battery storage facilities currently operate in grid-following mode. The trial involves retrofitting inverters with grid-forming technology, which replicates key behaviors of synchronous generators to provide essential stability services.
Read Full Article →

Key Takeaways

  • Neoen's Breizh Big Battery will be one of the first battery projects to pioneer grid-forming services on France's transmission network.
  • Grid-forming technology has progressed beyond experimental stages and is being deployed globally to support power systems with high renewable penetration.
  • The trial aims to enhance the stability of the French power grid by providing essential stability services, such as frequency regulation and voltage control, using advanced control architectures.
18

Who’s driving the 300GWh boom in demand for AI data centre battery storage?

Summary

The article discusses the growing demand for AI data centre battery storage, driven by the rapid expansion of the global AI market. To address these energy challenges, HiTHIUM is providing products and strategies that cater to this booming industry. The article highlights the need for innovative solutions to support the increasing energy demands of AI data centres.
Read Full Article →

Key Takeaways

  • The global AI data centre boom is driving a significant increase in demand for AI data centre battery storage.
  • Companies like HiTHIUM are emerging as key players in addressing the energy challenges faced by this rapidly expanding industry.
  • Innovative battery storage solutions are crucial to support the increasing energy demands of AI data centres and mitigate potential power outages or downtime.
19

Australian solar installer fined $6,000 over unsafe battery installs

Summary

A Melbourne-based solar installation company, Greenova Pty Ltd, has been fined $6,000 by an Australian court for pleading guilty to unsafe installations of residential battery energy storage systems at five properties. The installations were found to have exposed batteries to uncontrolled voltage, causing fires and damage to property. This incident highlights the importance of independent inspections before connecting battery systems under Victoria's electrical safety laws.
Read Full Article →

Key Takeaways

  • Greenova Pty Ltd was fined $6,000 for pleading guilty to unsafe installations of residential battery energy storage systems at five properties in Victoria.
  • The company's installations were found to have exposed batteries to uncontrolled voltage, causing fires and damage to property, including a minor house fire that destroyed the battery and damaged the exterior of the house.
  • The incident emphasizes the importance of independent inspections before connecting battery systems under Victoria's electrical safety laws to prevent similar incidents.
20

CATL Batteries Stand Out For Minimal Degradation In EVs And Long-Term Energy Storage Systems

Summary

Morgan Stanley conducted research on battery degradation in China, analyzing data from ride-hailing firms on 12 EV models and found that CATL batteries showed minimal degradation. The study compared the battery capacity decline of these vehicles to identify the most resilient battery technology for electric vehicles (EVs) and long-term energy storage systems.
Read Full Article →

Key Takeaways

  • CATL batteries demonstrated minimal degradation, standing out from other EV models in terms of battery capacity retention.
  • The research involved analyzing data from 100 cars across 12 different EV models, highlighting significant variability in battery degradation among the tested vehicles.
  • The study suggests that CATL batteries are a promising option for electric vehicles and long-term energy storage systems due to their minimal degradation.
21

Europe Built Hydrogen Infrastructure Instead of the Power Grid It Needed

Summary

The article discusses how Europe invested heavily in building a hydrogen infrastructure network, despite having no suppliers or offtakers, as opposed to developing a power grid. This highlights the importance of demand realism over technological aspirations in achieving decarbonization goals. The European experience serves as a cautionary tale for policymakers and industry leaders.
Read Full Article →

Key Takeaways

  • Europe's failure to develop a functional hydrogen infrastructure network, despite significant investment, underscores the need for robust demand and supply chains in the transition to low-carbon energy.
  • The article emphasizes that decarbonization success depends on realistic policy goals and strategies, rather than solely relying on technological advancements.
  • The European experience suggests that policymakers should prioritize building a reliable power grid over investing in unproven technologies like hydrogen infrastructure.
22

Solar accounted for 7.24 million jobs in 2024

Summary

The International Renewable Energy Agency (IRENA) has reported that solar energy accounted for over 43% of global jobs in the renewable energy sector in 2024, with a total of 7.24 million jobs worldwide. This represents an increase from 7.11 million jobs in 2023. The majority of these jobs are concentrated in China, India, and other countries in Asia.
Read Full Article →

Key Takeaways

  • Solar energy is the largest employer in the renewable energy sector globally, accounting for over 43% of total renewable energy jobs.
  • China holds a significant share of global solar jobs, with over 4.2 million positions, followed by India and Brazil.
  • The top five countries with the most solar jobs are China, India, Brazil, the United States, and Pakistan, which collectively account for around 82% of the global total.
23

MNRE & BIRAC Unveil INR 100 Crore CfP for Biomass-Based Green Hydrogen Pilots

Summary

The Ministry of New and Renewable Energy (MNRE) and the Biotechnology Industry Research Assistance Council (BIRAC) have launched a call for proposals to support pilot-scale green hydrogen production in India. The initiative aims to utilize India's vast biomass reserves, reduce waste, and promote public-private partnerships. The project seeks to develop innovative technologies that can be scaled up within 12 months.
Read Full Article →

Key Takeaways

  • The MNRE and BIRAC are offering INR 100 crore in funding for pilot-scale green hydrogen projects, with each project receiving up to INR 25 crore.
  • The initiative targets technologies already proven at bench scale, such as biomass gasification, pyrolysis, biological fermentation, and solar-integrated electrolysis.
  • The project aims to bridge the gap between lab experiments and real-world deployment by targeting Technology Readiness Levels 5-6, with a focus on reducing upfront risk for the private sector.

Technical papers from ArXiv - for readers interested in academic research

1

Large Artificial Intelligence Model Guided Deep Reinforcement Learning for Resource Allocation in Non Terrestrial Networks

Summary

The article proposes a novel approach to resource allocation in Non-Terrestrial Networks (NTN) using a Deep Reinforcement Learning (DRL) agent guided by a Large Language Model (LLM). The LLM generates textual guidance that shapes the reward of the DRL agent during training, leading to improved performance. The results demonstrate significant advantages of this approach over traditional DRL methods.
Read Full Paper →

Key Takeaways

  • The proposed Large AI Model (LAM)-DRL agent outperforms traditional DRL by 40% in nominal weather scenarios and 64% in extreme weather scenarios.
  • The LLM-guided DRL agent achieves improved performance through reduced task-specific training requirements, thanks to its generalization capabilities.
  • The approach offers benefits in terms of throughput, fairness, and outage probability in NTN resource allocation.
2

Grid-Aware Charging and Operational Optimization for Mixed-Fleet Public Transit

Summary

The article presents a comprehensive model to optimize charging schedules and trip assignments for mixed-fleet public transit systems, addressing the challenges of dynamic electricity pricing, vehicle capacity, and route constraints. The proposed approach uses a hierarchical method to tackle potential computational intractability, allowing it to be applied to relatively small fleets. By using real-world data from Chattanooga, Tennessee, the authors demonstrate significant savings in operating costs.
Read Full Paper →

Key Takeaways

  • The model uses mixed-integer linear programming (MILP) to optimize charging and trip assignments for mixed-fleet public transit systems.
  • The approach addresses the challenges of dynamic electricity pricing, vehicle capacity, and route constraints, which can pose significant operational challenges for transit agencies.
  • By employing a hierarchical method, the authors aim to make the model computationally tractable even with relatively small fleets, enabling its application in real-world scenarios.
3

AUV Trajectory Learning for Underwater Acoustic Energy Transfer and Age Minimization

Summary

This article proposes a novel approach for sustainable underwater acoustic energy transfer and age minimization using autonomous underwater vehicles (AUVs). The authors develop two deep-reinforcement learning algorithms to optimize frequency division duplex (FDD) and time division duplex (TDD) solutions for simultaneous information uplink and energy transfer. The results show significant reductions in average age of information and improved data collection fairness.
Read Full Paper →

Key Takeaways

  • The proposed AUV-based approach enables underwater acoustic energy transfer, potentially allowing IoUT devices to operate indefinitely without battery limitations.
  • Two deep-reinforcement learning algorithms are developed to optimize FDD and TDD solutions for simultaneous information uplink and energy transfer.
  • The proposed solutions significantly reduce average age of information and improve data collection fairness compared to baseline approaches.
4

Current and temperature imbalances in parallel-connected grid storage battery modules

Summary

The article discusses the challenges of heterogeneous currents and temperatures in parallel-connected grid storage battery modules, a common issue in large battery systems. Researchers conducted simulations and experiments to quantify these imbalances and identified key factors contributing to them, such as contact resistances and cell resistances. The study aims to provide safety thresholds for cell-to-cell variability and robustness metrics to mitigate failures.
Read Full Paper →

Key Takeaways

  • Heterogeneous currents and temperatures in parallel-connected grid storage battery modules can be a significant issue, affecting performance and lifetime of battery cells.
  • Contact resistances and cell resistances are strong contributors to temperature differences between cells, and safety thresholds for cell-to-cell variability have been defined based on these factors.
  • Lowering C-rates (charge/discharge rates) and narrowing SOC (state of charge) ranges can help mitigate failures in grid storage battery modules.
5

Multiobjective Model Predictive Control for Residential Demand Response Management Under Uncertainty

Summary

The article proposes a multiobjective model predictive control approach for residential demand response management, aiming to minimize electricity costs and user dissatisfaction under real-time pricing. The proposed method uses Laguerre functions to parameterize control signals and incorporates a constrained multiobjective evolutionary algorithm to ensure feasible solutions. This approach is shown to outperform existing methods in simulations.
Read Full Paper →

Key Takeaways

  • The study proposes a novel approach to residential demand response management, integrating electricity cost minimization with user satisfaction under real-time pricing.
  • The proposed method uses Laguerre functions and a constrained multiobjective evolutionary algorithm to efficiently explore feasible solutions and mitigate uncertainties.
  • The simulation results show that the proposed approach can limit cost increases to 0.52% under uncertainties, significantly outperforming existing methods which result in cost increases of at least 2.3%.
6

Blockchain-Enabled Renewable Energy Certificate Trading: A Secure and Privacy-Preserving Approach

Summary

The article discusses the implementation of Blockchain-Enabled Renewable Energy Certificate Trading, which aims to secure and protect consumer privacy in the tracking of renewable energy production and consumption. The study utilizes a directed acyclic graph (DAG) blockchain system to facilitate REC trading between suppliers and consumers, resulting in improved transaction times and reduced energy consumption compared to traditional proof-of-stake methods.
Read Full Paper →

Key Takeaways

  • The existing Renewable Energy Certificate (REC) schema has two main challenges: inconsistent global adoption and inadequate consumer privacy protection, which are addressed by the proposed Blockchain-Enabled Trading System.
  • The study demonstrates a 41% reduction in transaction time and a 65% decrease in energy consumption when using the DAG blockchain system for REC trading compared to traditional proof-of-stake methods.
  • The introduction of the trading schema helps protect consumer information while facilitating secure and transparent REC trading, aligning with global efforts to mitigate climate change by promoting renewable energy adoption.
7

Coordinated Cooling and Compute Management for AI Datacenters

Summary

The article discusses the need to optimize power and energy consumption in AI datacenters, particularly those supporting large-language models. Current research has focused on compute-side scheduling without considering thermal objectives or constraints, which can lead to increased energy consumption and reduced efficiency. The authors propose a novel hierarchical control framework that balances computing and thermal management.
Read Full Paper →

Key Takeaways

  • Most prior research on optimizing LLM inference has focused on compute-side scheduling without considering thermal effects, potentially leading to increased total energy consumption.
  • GPU-intensive inference generates substantial heat, which can degrade datacenter performance if ignored, highlighting the importance of thermal objectives in AI datacenter design.
  • The proposed hierarchical control framework aims to co-optimize computing and thermal management by identifying optimal GPU parallelism, frequency (DVFS), and cooling control knobs to balance serving latency and thermal constraints.