Back to Archive

LA-AI Insights: What AI's Power Appetite Actually Means for Alabama

Your weekly AI news and updates from Lower Alabama

Friday, April 17, 2026

Share this newsletter:

<p>This week Lower Alabama AI has had the pleasure of welcoming Ezra Langley as our “Spring break Intern.” Ezra is currently a junior at Baldwin Preparatory Academy and some of you might remember Ezra from Hatch-a-thon 2026 where he won the Most Innovative Solution prize. Ezra has been busy getting a local LLM running on a Raspberry PI as he prepares for a UAV drone competition in Montgomery. As part of Ezra’s internship he will be presenting his work at the next LA-AI Fairhope meetup on May 8th. <br><br>In the meantime, I encourage you to connect with Ezra on Linkedin or on Discord. <br><br><strong>Linkedin</strong>: <a target="_blank" rel="noopener noreferrer nofollow" class="text-teal-400 hover:text-teal-300 underline" href="https://www.linkedin.com/in/ezra-langley-7604b5314/">https://www.linkedin.com/in/ezra-langley-7604b5314/</a><br><br><strong>Discord</strong>: @thewoodenmandalorian<br></p>



How much power does it all need?

If you were at the civic club meeting where hundreds of Baldwin County residents packed the room, you already know the basics. 4,500 acres of timberland in north Baldwin County, slated for a 260-megawatt solar farm. The energy feeds Alabama Power's grid. The renewable energy credits go to Meta, which uses them to offset its data center campus in Montgomery.

The room had a lot of questions. The biggest one wasn't really about solar panels or timber. It was simpler than that: why does an AI company need this much power in the first place?

The short answer: AI uses a lot of electricity. More than most people realize.

Training and running large AI models rely on thousands of specialized processors, often operating continuously for weeks or months. Unlike an office building or even a traditional server farm, these facilities don't really have slow hours. They run at or near full capacity around the clock, and they generate enormous amounts of heat, which means they also need massive cooling systems running alongside them.

A single hyperscale data center can consume as much electricity as a mid-sized city.

That's not a figure of speech.

Nationally, the trend is clear. The North American Electric Reliability Corporation, which monitors the U.S. power grid, projects that summer peak electricity demand will increase by 224 gigawatts over the next decade. That's a 20 to 25 percent jump, and it's the fastest acceleration in demand since NERC started tracking this data in 1995. AI infrastructure is a major driver of that growth.

So what does this look like in Alabama?

Alabama Power's most recent long-term planning documents project a capacity shortfall beginning in 2029. The utility hasn't publicly pinned that gap entirely on data centers, but the timing lines up with a wave of large-scale projects either under construction or in planning.

Meta's Montgomery data center campus is expanding beyond $1.5 billion in investment. The Stockton solar project is directly tied to that expansion.

Worth noting how the math actually works. The 260-megawatt Stockton solar farm, once built, would generate somewhere around 500,000 to 600,000 megawatt-hours a year. Solar panels only produce when the sun is up, so a 260-megawatt solar farm puts out far less energy than a 260-megawatt data center draws. Meta gets the renewable energy credits on paper. The physical electrons powering the Montgomery campus still come from wherever Alabama Power happens to be generating at any given hour, which is mostly natural gas and nuclear.

In Bessemer, a separate hyperscale campus known as Project Marvel has been proposed at roughly 1.2 gigawatts across 18 buildings, and it's already generating local debate about water use, emissions, and land clearing.

To put that in local terms: Mobile County as a whole, every home, business, school, and factory combined, uses about 7.2 million megawatt-hours of electricity a year. A 1.2-gigawatt data center campus running around the clock would use closer to 9.5 million. One project, more power than an entire coastal Alabama county. Statewide generation in 2025 came in around 142 million megawatt-hours, which puts Project Marvel alone at roughly 7 percent of Alabama's total electric output.

To shore up capacity, Alabama Power agreed in late 2024 to acquire the Lindsay Hill Generating Station, an 855-megawatt natural gas plant in Autauga County, for approximately $622 million. That purchase goes into the utility's rate base, the pool of assets that gets factored into what customers pay every month.

Alabama's residential electricity price in January 2026 sat at about 16 cents per kilowatt-hour, up roughly 6.4 percent from the same month a year earlier. That increase reflects a mix of factors, not just data centers. But the direction is clear. Demand is growing. Infrastructure is being built to meet it. And those costs eventually show up somewhere.

Those that know me know I'm extremely pro AI. I want the technology to succeed, and I want the Gulf Coast to get the benefit of it. But pro AI doesn't mean pretending the grid isn't strained.

The loud read is that Big Tech is externalizing its power costs onto families and small businesses that never asked to underwrite a data center. The quieter read is that this is how most major American infrastructure has been built, roads, railways, broadband, with costs spread across the grid, not just the companies using the most of it. The truth, as usual, is messier than either camp wants to admit. It depends on how the rates are structured, who gets first claim on new generation, and whether the utility passes the capital costs through to residential customers or negotiates them into the industrial tariff.

Those structural choices are the ones worth watching.

What people are watching

Alabama PSC President Twinkle Cavanaugh said publicly in late 2025 that the state needs to make sure families don't "get stuck with the bill" for large-scale data center infrastructure, and that the standard should be "no free rides for Big Tech." Those are signals about where the commission's thinking is headed, though they haven't translated into formal policy changes yet.

Across the country, at least 11 states introduced legislation in 2026 proposing temporary pauses on new data center development. Maine became the first to actually pass one. The common thread isn't anti-technology. It's a desire to study the cumulative effects on power grids, water supplies, and electricity rates before approving more projects.

Alabama hasn't pursued a moratorium. But with multiple large projects in the pipeline and a projected capacity gap just a few years away, the conversation about how these costs are shared is only getting started.

Why this matters for our community

You don't need a data center in your county to feel the ripple effects. When utilities build new generation and transmission to meet large industrial demand, those capital costs work their way into the rate structure. How much of that cost lands on residential customers versus the companies driving the demand is a policy question, and it's one Alabama regulators will be answering in real time over the next several years.

For a small business owner in Fairhope or a nonprofit director in Mobile, the electric bill is the utility cost you can't negotiate and can't outrun. Worth knowing what's coming.

— Kai



Upcoming Events

LA-AI Mobile

Invalid Date

Innovation Portal: 358 St. Louis St. Mobile

LA-AI Fairhope

Invalid Date

Hatch Fairhope: 314 Magnolia Ave., Fairhope

This Week in AI

1. Anthropic releases a new Opus model amid Mythos Preview buzz

Anthropic has released Claude Opus 4.7, a new flagship model, amid broader speculation surrounding its Mythos Preview program. The release arrives with notable cybersecurity-focused capabilities, signaling Anthropic's intent to compete directly in enterprise security and high-stakes reasoning workloads. For teams evaluating frontier models, Opus 4.7 represents an updated performance ceiling from Anthropic and warrants immediate benchmarking against GPT-5 and Gemini Ultra. The timing alongside Mythos Preview buzz suggests Anthropic is managing a layered product roadmap targeting both current enterprise demand and longer-horizon capability deployment.

The VergeRead more

2. OpenAI’s big Codex update is a direct shot at Claude Code

OpenAI has significantly upgraded Codex into an always-on coding agent capable of monitoring screen activity and operating continuously on macOS, positioning it as a direct competitive response to Anthropic's Claude Code. This marks a strategic shift from Codex as a point-use tool to an ambient development partner embedded in the developer workflow. For engineering organizations choosing AI coding infrastructure, this changes the competitive calculus — persistent agents that observe context over time represent a meaningfully different capability tier than prompt-response coding assistants. Developer tool lock-in implications are significant.

The VergeRead more

3. Google Opens Gemma 4 Under Apache 2.0 with Multimodal and Agentic Capabilities

Google has released Gemma 4 under the permissive Apache 2.0 license, adding multimodal understanding and agentic capabilities to its open-weight model family. This is strategically significant: Apache 2.0 licensing enables unrestricted commercial deployment without royalty obligations, lowering barriers for enterprise and startup adoption. Adding agentic and multimodal features to an open-weight model narrows the capability gap between proprietary APIs and self-hosted alternatives. Organizations seeking data privacy, cost control, or customization will find Gemma 4 a materially stronger candidate than prior open-weight options from Google.

4. 2026 AI Index Report released

Stanford's annual AI Index Report for 2026 has been released, providing the field's most comprehensive data-driven assessment of AI progress, investment, adoption, policy, and societal impact across geographies and sectors. The AI Index is a primary reference document for policymakers, investors, and technical leaders establishing strategic baselines. Key dimensions tracked include model performance trajectories, compute trends, labor market effects, and global R&D competitiveness. For organizations conducting annual AI strategy reviews or competitive benchmarking, this report represents essential context that will inform decisions and public discourse throughout the remainder of 2026.

ΑΙhubRead more

5. China has erased the US lead in AI, Stanford HAI’s 2026 AI index reveals

Stanford HAI's 2026 AI Index, one of the most authoritative annual assessments of global AI progress, finds that China has effectively closed the gap with the United States across key AI capability benchmarks. The report documents parity or Chinese advantages in areas including model performance, research output, and deployment scale. For US technology and policy leaders, this is a significant strategic data point — export controls and compute restrictions have not prevented China from reaching competitive AI capability levels, prompting reassessment of current policy frameworks and corporate competitive strategies.

AI – SiliconANGLERead more

6. OpenAI's leaked memo says new "Spud" model will make all its products "significantly better"

A leaked internal OpenAI memo reveals a forthcoming model codenamed 'Spud,' described internally as capable of making all OpenAI products 'significantly better.' The disclosure provides rare visibility into OpenAI's near-term product roadmap and signals a broad capability uplift across its entire portfolio — not just a standalone release. For enterprise teams evaluating OpenAI integrations, this suggests meaningful performance improvements may be imminent. The memo also frames competitive urgency against Anthropic, offering insight into how OpenAI's leadership is internally framing the frontier model race at this moment.

The DecoderRead more

Know someone who would enjoy this newsletter?

Forward this email or share the link below

https://la-ai.io/newsletter/view/2026-04-17
Subscribe to LA-AI Newsletter

Join Our AI Community

Get weekly insights on AI innovations and exclusive updates on LA-AI events

Subscribe Now