Thursday, June 25, 2026Vol. III · No. 176Subscribe
The Mining, Energy & Technology Wire
Technology · Analysis

AI's Power Crunch Reshapes Energy Bets

Data center electricity demand will jump 27% this year as AI labs race overseas for power and chipmakers scramble to build software moats. The grid can't keep up.

AI's Power Crunch Reshapes Energy Bets
PhotographData center electricity demand will jump 27% this year as AI labs race overseas for power and chipmakers scramble to build software moats. The grid can't keep up.

Worldwide data center power demand will rise 27% in 2026 and reach 132 gigawatts, up from 104GW in 2025 , according to Gartner. That's roughly equivalent to adding the entire electricity consumption of Spain in a single year. The IEA estimates data centers consumed around 415 terawatt hours in 2024—about 1.5% of global electricity—and projects that figure will double to 945 TWh by 2030 .

The numbers are staggering, but the real story is playing out in hiring queues and chip deals. Anthropic is hiring for AI data center roles in Australia and Japan as it rushes to expand compute capacity overseas , CNBC reported Tuesday. A listing for a data center energy role in Australia specifically mentions the company's "rapidly expanding AI compute footprint across the region" and talks of leading "multi-hundred megawatt procurement efforts" . The AI lab raised $65 billion in May at a $965 billion valuation, with its revenue run-rate crossing $47 billion that month—multiple times higher than the "around $9 billion" figure from the end of 2025 .

Meanwhile, Qualcomm is in advanced talks to acquire Modular Inc. in a transaction valuing the artificial intelligence infrastructure software company at about $4 billion , Bloomberg reported. The deal, which CNBC said could be announced within weeks, signals that the chip wars have moved beyond silicon. Qualcomm plans to acquire AI startup Modular in an all-stock deal valued at about $3.92 billion, a move that gives Qualcomm a software layer built to run AI models across different chips without forcing developers to rewrite code for each processor .

Can the Grid Handle What's Coming?

A January 2026 report by Bloom Energy predicts that U.S. data centers' total combined energy demand will nearly double between 2025 and 2028, jumping from 80 to 150 gigawatts—like adding a country with the energy needs of Spain in just three years , Consumer Reports noted. By 2028, data centers could use 12 percent of all the electricity consumed in the U.S. , according to Lawrence Berkeley National Laboratory.

The strain is already visible. Bloomberg's analysis found that data centers accounted for almost 40 percent of Virginia's total electricity consumption in 2024 . Areas with high concentrations of data centers saw electricity prices jump 267 percent over the past five years , the same analysis found.

Dominion's 2024 resource plan projects nearly 27 GW of new generation by 2039, including 21 GW of renewable energy and 5.9 GW of gas, while Virginia's energy rates are increasing—Dominion proposed its first base-rate increase since 1992, adding about $8.51 per month in 2026 and $2.00 per month in 2027 for a typical household , according to a Belfer Center analysis.

The bottleneck isn't just generation—it's the grid itself. Most of the U.S. grid was built between the 1950s and 1970s, and today, approximately 70% of the grid is approaching the end of its life cycle , an industry executive told Data Center Knowledge.

Why Australia and Japan Matter Now

Anthropic's overseas push isn't about diversification—it's about survival. Australia has excess land, abundant renewable energy potential and a stable political and regulatory environment, plus "distance from military threats, which have proved such a vulnerability for the Gulf states" , an Australian Strategic Policy Institute analyst told CNBC. Conflict in the Middle East has tested the region's credentials as a secure place to build AI infrastructure, with two Amazon data centers targeted early in the year.

Japan has evolving grid infrastructure and significant government interest in domestic AI infrastructure , according to Anthropic's job listing. Wood Mackenzie principal analyst Xiaonan Feng noted that "securing power is becoming more challenging than securing land, financing or permits" .

The math is brutal. AI data center racks have gone from each having eight GPUs to 72 starting two years ago, requiring around 150kW of power—and Nvidia's new Rubin GPU and rack system coming out later this year will eventually need around 300kW to run, with the industry bracing for chips that bring racks closer to 1 megawatt, or enough to power 750 U.S. homes on average , Bloomberg reported.

Can Nuclear Fill the Gap?

Tech giants are betting on it. Big tech companies signed contracts for more than 10 gigawatts of possible new nuclear capacity in the United States over the last year, with Microsoft committing to a 20-year, 835-megawatt power purchase agreement to restart Three Mile Island, Google ordering up to 500 megawatts of small modular reactors from Kairos Power, and Amazon investing over $20 billion converting the Susquehanna site into a nuclear-powered AI data center campus , according to industry analysis.

Small Modular Reactors are a new class of nuclear reactors designed to be smaller, safer, and more flexible than traditional plants, typically generating between 50 and 300 MW, making them suitable for single data center campuses or regional clusters . Unlike the massive gigawatt-scale reactors that require large facilities, SMRs are typically 5–300 MW in size and are engineered to be quickly manufactured in factories and transported to their final location, then installed on site on a much smaller footprint than conventional reactors .

But SMRs remain largely unproven at commercial scale. The first units aren't expected online until the late 2020s at the earliest.

What About the Old Energy Economy?

While AI devours electricity, the traditional energy sector is quietly automating. Robotics is rapidly transforming oil and gas operations as advances in AI and cloud computing unlock the next phase of industrial automation—robots can now operate autonomously, collaborate, and access cloud-based data in real time, enabling advanced decision-making, navigation in complex environments, and reduced reliance on human intervention, reshaping how oil and gas companies approach inspection, maintenance, and monitoring , GlobalData reported.

The global robotics market was valued at $90.2bn in 2024 and is expected to grow at a CAGR of 15% to reach $205.5bn by 2030 , according to GlobalData forecasts. An EY poll found that 62% of Gen Z teens find a career in the oil and gas industry unappealing, while only 26% of the broader Gen Z demographic find such roles attractive , making automation less a choice than a necessity.

The Financial Times noted Tuesday that robots are coming to the oil patch—a shift driven as much by workforce constraints as efficiency gains.

The Chip Wars Move to Software

Qualcomm's Modular deal is a direct shot at Nvidia's CUDA ecosystem, the software platform that has locked developers into Nvidia's hardware for years. Modular's platform is anchored by the Mojo programming language and the MAX inference engine, letting developers write code once and run it across wildly different hardware, pushing models across CPUs, GPUs, NPUs, and custom ASICs without forcing a rewrite for every chip .

The timing matters. Cerebras said revenue almost doubled in the AI chipmaker's first earnings report since its initial public offering last month, but the stock fell 10% in extended trading as the company forecast a drop in its gross margin , CNBC reported Tuesday. The company's revenue increased 92% in the first quarter from $99.5 million a year earlier, with net loss narrowing to $14 million from $23.9 million a year ago .

Cerebras CEO Andrew Feldman told CNBC the margin forecast was "misunderstood," but the market's reaction underscores the pressure on AI chipmakers to prove unit economics at scale.

What Changed This Week

Three forces converged: Anthropic's hiring spree revealed the global scramble for power capacity outside traditional data center hubs. Qualcomm's near-$4 billion bet on Modular showed that software—not just silicon—will determine who wins the AI infrastructure race. And Cerebras's first earnings as a public company exposed the margin pressure facing anyone trying to challenge Nvidia's dominance. The common thread: power availability, not chip performance, is becoming the binding constraint on AI growth.

What to Watch

Cerebras reports full Q1 results after the close today, with a conference call at 5 p.m. ET. Anthropic's Australia and Japan hiring will signal whether the offshore data center push accelerates or stalls. And watch for Qualcomm's formal Modular announcement—expected within weeks—which could trigger a wave of similar software acquisitions as chipmakers race to build developer ecosystems. The IEA's next data center energy update, due later this summer, will reveal whether 2026 demand is tracking toward the high or low end of projections.

Original reporting and analysis by the Stake & Paper editorial team. See linked sources within the article.

Share this story

More from Stake & Paper

Was this article helpful?

ClaimWatch

Mining claims intelligence — from query to report, in minutes.

Every unpatented mining claim across all twelve BLM states. Leadfile audits, due diligence, site selection, regional prospecting, entity investigations, and AOI monitoring — delivered as complete report packages.

4.4M+
Claims Tracked
12
BLM States
7
Report Types
Request a Sample Report
Stake & Paper AM

One morning brief. The whole energy sector.

Original analysis, the day's most important wire stories, and market data — delivered before your first cup of coffee. Free.