Introducing the AI-Energy Nexus newsletter

Introducing the AI-Energy Nexus newsletter

Welcome to the AI-Energy Nexus. I’m Stephen Lacey, executive editor at Latitude Media and one of the authors of this new weekly newsletter.?

Every week, we’ll share a preview of this newsletter here on LinkedIn. This week, we’re bringing you the full AI-Energy Nexus to give you a taste of what to expect in the email version over the coming weeks and months. After today, you’ll need to subscribe to receive the full newsletter in your inbox every Wednesday (it’s free).

Thanks for reading.


After the release of GPT-3 in the fall of 2022 the discourse around AI focused largely on societal and economic impacts.

But in 2024, as Silicon Valley’s AI era entered its second year, attention shifted to another consequence: an overwhelmed electric grid.

I’ll be totally honest: I didn’t take OpenAI’s Sam Altman all that seriously when he said the future of AI depends on energy breakthroughs. Altman’s infatuation with nuclear fusion and advanced fission — instead of the actual needs of an old, backlogged electric grid — felt deeply out of touch with the problem.?

Early projections of a rapid doubling of data center power demand felt like echoes of warnings from the early 2000s that the internet would eat up half of America’s power demand. Those early predictions were laughably inaccurate.

But the tone shifted quickly this year. Experts with a window into AI energy consumption started speaking out. Tech companies started sliding backward on their emissions. Projections for rapid electricity growth piled up. And utilities overwhelmed with new load requests from data centers proposed what they know best: gigawatts of new fossil gas.?

In our conversations, experts inside tech companies and the federal government are expressing deep concern about how the rapid surge of new data center capacity will complicate decarbonization. And we haven’t even felt the full energy impact of the AI boom yet.

Still, the conversation can feel incomplete and hyperbolic as people declare that America is “running out of power.”

The reality is that this is a local grid capacity problem. We have many of the tools to alleviate it today — without major breakthroughs, at least for now. And we need to look at the sector as part of the overall load growth picture, driven by broader electrification and manufacturing reshoring.

Plus, as we have covered extensively at Latitude Media, the AI boom will bring enormous benefits to mapping, managing, and modernizing the grid.?

The AI-Energy Nexus will focus on both of these stories: load growth challenges and AI use cases to support the energy transition. We’ll bring together a range of insights — reporting from Latitude Media and curated news — for readers across the data center, computing, tech, utility, and energy technology industries.

Our team will cover:

  • Data center capacity expansions
  • Data center efficiency and grid interactivity
  • Chip design and AI modeling efficiency
  • Grid-enhancing tech, storage, VPPs, and grid capacity expansions
  • Contracting strategies for clean, firm power
  • Utility integrated resource plans
  • Startups and vendor strategies
  • AI applications for grid management and clean energy integration

If you are looking for a compilation of insights into how AI is shaping load, changing the data center industry, and creating new creative opportunities for the power system, this is your place. We hope you’ll subscribe and get in touch with your feedback and story ideas.

– Stephen Lacey, executive editor?


FEATURED STORY

Are we thinking about the data center energy problem in the right ways?

STEPHEN LACEY | Peter Freed, the former director of energy strategy at Meta, is worried about the evolving challenge of decarbonizing data centers. But he recently explained to Latitude Media why we’re sometimes framing the energy intensity of hyperscale computing in unhelpful ways. Freed outlines some creative approaches that could help. (Freed will also take the stage alongside Stephen Lacey at the Transition-AI conference in Washington, DC on December 3.)


WHAT WE’RE WATCHING THIS WEEK

Big bets on nuclear, Iberdrola’s novel model, and OpenAI’s 5 GW fantasies

SCOTT CLAVENNA | Tech companies are giving the nuclear industry increasingly strong demand signals. Microsoft signed a $16-billion, 835 MW PPA with Constellation Energy to support the restart of Three Mile Island’s Unit 1 reactor, now dubbed the Crane Clean Energy Center, with no impact to ratepayers. (Unit 2, next door, infamously shut down in 1979 and is still being decommissioned.) Also this month, both Google and Oracle said they want to use small nuclear reactors to power data centers, while the Biden administration announced it will provide a $1.52 billion loan guarantee to Holtec International to restart the Palisades nuclear power station in Michigan, which was retired in 2022. AWS continues to make progress on its deal with Talen to power its data center in Pennsylvania, despite questions from FERC about the direct-connect deal with Talen’s Susquehanna nuclear power plant.?

These nuclear bets are fraught with risks — permitting, supply chain, construction costs — which makes the recent announcement by ECL of a 1 GW hydrogen-powered off-grid data center deal with Lambda worth considering. The planned site, east of Houston, has direct access to hydrogen pipelines, and could be put into service in a year, far faster than any nuclear solution. We looked at green hydrogen for data centers in August, which is highly limited except for cases where access to hydrogen is local.?

In Spain, Iberdrola is taking its data center ambitions beyond PPAs and into equity. In a recent webinar , the company described its new in-house company, CPD4Green, that will offer data center developers land, renewable energy supply, interconnection, and a 24/7 contract in exchange for a 20% ownership interest in the data center. It plans to use this throughout Spain, with a target of 20% market share of the national data center market by 2030.

OpenAI says it needs multiple 5 GW data centers to power the next wave of ChatGPT — and it wants the government to help. Is this more magical thinking? There is no way to secure the power and infrastructure today to serve that size facility, nor is it likely to be the optimal way to build and power data centers in the future.?


STARTUPS AND INNOVATIONS

We covered AI-powered “firetech” in detail, with mentions of Rhizome , Pano AI , and Google’s FireSat system.?

BrightBand raised a $10 million Series A to develop and commercialize its AI-driven weather forecasting technology.

The utility FirstEnergy will deploy an AI solution developed in-house called Advanced Vegetation Analytics Tool (AVAT), a prediction model that pulls data — including soil, weather, roadway, historical outage data, geographical typography and high-resolution aerial patrol photos — to improve vegetation management.

CBS profiled Gridware and its AI solution for wildfire detection.

Fast Company reports on G42 , a UAE-based AI startup developing weather tech, in partnership with NVIDIA, and starting a climate tech lab in Abu Dhabi.

TechCrunch added Exa Labs to their list of startups to watch coming out of Y Combinator’s Demo Day.


MORE NEWS

Latitude Media | $30 billion for AI and energy infrastructure? Put some of it in the grid — Microsoft, BlackRock, Global Infrastructure Partners, and MGX plan to raise and invest billions in U.S. data centers and power infrastructure. “There seems to be a tendency in the tech industry to think you can bypass the electric grid and build nuclear plants to run everything,” Brian Janous, Cloverleaf Infrastructure CEO, said.?

Bloomberg | Constellation CEO Says US Should Copy China to Meet AI Power Use — Joe Dominguez, Constellation Energy CEO, said the U.S. should do what China is doing, and build data centers alongside power plants, skipping the headaches of transmission lines.

Houston Chronicle | Texas Politicians Weigh Energy Regulation to Accommodate Data Centers — In Texas, data centers’ power and water needs have politicians weighing new regulations for their development, as Texas utility Oncor recently noted that more than 70% of the connection requests of the past year came from data centers. Meanwhile, a tech hub in Houston secured $40 million to meet rising data center needs (Innovation Map ), using C-PACE, a financing mechanism that was historically used to finance energy and water conservation systems at commercial buildings

Latitude Media | In Chile, Google X is taking AI-powered grid tools out of pilot purgatory — In Chile, Google X’s Tapestry is using AI to make the grid more visible for planning and operating purposes, in order to speed up the country’s plans to phase out coal.

Bloomberg | The energy boss overhauling the grid for AI and net zero — John Pettigrew, CEO of National Grid, spoke with other utility executives recently in the U.S. and is skeptical of the absolute numbers being used around data center interconnection requests.?


RESOURCES + EVENTS

On December 3, join us in DC for Transition-AI 2024 , a one-day in-person conference that will convene a broad range of experts on the impacts of AI on the energy transition (like the energy-first approach to AI infrastructure covered in our recent event with Crusoe , and everything we talk about in this newsletter).

Tapestry and AES are proposing a digital grid architecture divided into three categories:?governing frameworks and market mechanisms; open standards and software systems; and physical infrastructure.?

This new McKinsey report highlights opportunities for investors in power infrastructure and adjacent sectors, including transmission and distribution.?

This handy Data Centers 101 , published by Orennia, summarizes key points at the AI-energy nexus.?

You can download JLL’s mid-year U.S. Data Center Report here , and compare it with CBRE’s H1 2024 North America Data Center Trends report here . They both agree power is the limiting factor on growth, and markets like Atlanta, Austin, Houston, and Phoenix are new hot spots.?

In this short economic bulletin, the Federal Reserve Bank of Kansas City weighs in on the recent surge in U.S. electricity demand, driven in part by the adoption of artificial intelligence.?


Like what you see? Subscribe to receive the full newsletter in your inbox every Wednesday.

要查看或添加评论,请登录

社区洞察

其他会员也浏览了