Technology · Analysis
Data Centers Turn to Batteries and Gas as AI Boom Strains the Grid
As artificial intelligence drives unprecedented power demand, data center operators are pairing battery storage with natural gas generation to bypass grid bottlenecks—while utilities scramble to plan for load growth that's compressing decade-long timelines into three years.
Stake & Paper Editorial TeamApril 24, 2026
In their quest for power, data centers are building gas and battery power off the grid
, according to Bloomberg, as the AI infrastructure boom collides with an aging electrical system that wasn't designed for this moment.
By pairing energy storage with natural gas, hyperscalers are able to get power faster and do so behind the meter
, Bloomberg reported Thursday. The strategy reflects a stark reality:
according to JT Steenkamp, vice president of projects and technology at Prologis Mobility, the US could see roughly 200 GW of new AI-driven demand by the end of the decade, while also retiring more than 100 GW of existing capacity
, creating what he described as a 300-gigawatt gap that needs solving.
Meta Bets Big on Amazon's Custom Chips
Amazon's cloud unit said Friday that Meta has agreed to use Amazon's general-purpose Graviton chips in a deal that will run for at least three years, coming weeks after Meta signed deals worth a combined $48 billion with CoreWeave and Nebius
, CNBC reported.
The deployment starts with tens of millions of Graviton cores, with the flexibility to expand as Meta's AI capabilities grow
, according to Amazon.
The deal reflects a shift in how AI infrastructure gets built: while GPUs remain essential for training large models, the rise of agentic AI is creating massive demand for CPU-intensive workloads—real-time reasoning, code generation, search, and orchestrating multi-step tasks
, Amazon said in a statement.
Meta's head of infrastructure Santosh Janardhan said that "expanding to Graviton allows us to run the CPU-intensive workloads behind agentic AI with the performance and efficiency we need at our scale"
, according to CNBC.
Utilities Face Planning Whiplash
The speed of AI-driven demand is upending traditional utility planning.
The Electric Power Research Institute said on Feb. 26, 2026, that data centers could account for 9% to 17% of total U.S. electricity demand by 2030
, according to Utility Dive.
Electricity demand from data centres rose by 17% in 2025, far outpacing the 3% growth in overall global electricity demand
, according to a recent IEA report.
By 2030, the IEA projects that electricity consumption from data centres will double, while AI-specific demand is expected to triple
.
The IEA notes that U.S. electricity demand growth over 2026–2030 is expected to average around 2% per year—more than twice the pace of the prior decade—and that "data centres' electricity consumption is expected to account for almost half of the projected growth"
, according to analysis of the IEA's Electricity 2026 report.
That's forcing utilities to compress what were traditionally 10-year planning cycles into much shorter timeframes.
Data centers will play a more active role in stabilizing the grid and mitigating cost increases by securing strategic investment and promoting load flexibility via load shedding or curtailment
, according to Clift Pompee, VP of power and emissions at Compass Datacenters, speaking at Data Center World 2026.
NextEra Locks Down Massive Gas Projects
NextEra Energy expects to finalize agreements within about the next three months for giant natural gas-fired power projects that will be backed by Japan and used for data centers in the U.S., Chief Executive Officer John Ketchum said on Thursday
, Reuters reported.
NextEra Energy Resources will develop a 4.3 GW natural gas generation hub in southwest Pennsylvania and a 5.2 GW hub in Anderson County, Texas, as part of a $550 billion trade deal with Japan
, according to Utility Dive.
The hub in Pennsylvania will support up to 3.5 GW of large load demand, connect to existing interstate natural gas pipelines within the Marcellus and Utica shale region and interconnect with the PJM regional transmission network
, the Commerce Department said.
The projects are capital-light for NextEra: "Essentially zero capital for us," CFO Kirk Dunne said. "From a returns perspective, it's essentially infinite"
, according to coverage of the company's earnings call.
Semiconductors Hit Record Streak
The chip sector that powers all this AI infrastructure is on a historic tear.
The Philadelphia Stock Exchange Semiconductor Index rose 2.7% on Wednesday, its 16th straight positive session—its longest such streak on record, based on data that goes back to 1994, with the index climbing about 39%, putting April on track for the biggest one-month percentage gain since February 2000
, Bloomberg reported.
U.S. chip stocks soared to record highs on Friday, with Intel's unexpectedly strong revenue forecast fueling fresh optimism, as the Philadelphia SE Semiconductor Index rose 2.5% to an all-time high and was on track to extend its record-breaking streak of single-day gains to 18, with the index gaining more than 42% so far this year
, according to Reuters.
The global semiconductor industry is expected to reach US$975 billion in annual sales in 2026, a historic peak fueled by an intensifying AI infrastructure boom, with growth reaching 22% in 2025 and projected to accelerate to 26% in 2026
, according to Deloitte's 2026 semiconductor industry outlook.
MarketWatch noted that the rally has triggered warnings.
The semiconductors sub-industry alone is expected to record first-quarter earnings growth of 104.9% – much higher than the broader S&P 500 information technology sector whose earnings growth is seen at 46.2%
, according to LSEG data cited by Reuters.
OpenAI Ships GPT-5.5 at Breakneck Pace
OpenAI announced GPT-5.5, its latest AI model that is better at coding, using computers and pursuing deeper research capabilities, with the launch coming less than two months after OpenAI released GPT 5.4
, CNBC reported Wednesday.
OpenAI said GPT-5.5 is "our smartest and most intuitive to use model yet, and the next step toward a new way of getting work done on a computer," noting that it "understands what you're trying to do faster and can carry more of the work itself" and "excels at writing and debugging code, researching online, analyzing data, creating documents and spreadsheets, operating software, and moving across tools until a task is finished"
, according to the company's announcement.
The model runs on NVIDIA GB200 NVL72 rack-scale systems
, according to Nvidia's blog.
The systems are capable of delivering 35x lower cost per million tokens and 50x higher token output per second per megawatt compared with prior-generation systems
, Nvidia said.
The Energy Constraint Becomes the Compute Constraint
What ties these stories together is a fundamental shift: energy availability is now the binding constraint on AI growth, not chip supply or software capability.
Racks that once pushed 30–40 kW are now measured in hundreds of kilowatts, with designs approaching the megawatt range
, according to Varun Sakalkar, distinguished engineer in Google's datacenter technology and systems group, speaking at Data Center World 2026.
Natural gas-fired power generation is poised to rebound this year amid the proliferation of energy-hungry data centers, though it could face stiff competition regionally from renewables and battery storage, with Wood Mackenzie forecasting power burns to be up 0.8 Bcf/d or 2.2% in 2026
, Natural Gas Intel reported.
Yet
battery storage additions are expected to total 24 GW, beating last year's total of 15 GW
, according to EIA data cited by Natural Gas Intel. The pairing of batteries with gas generation is becoming the pragmatic solution for data centers that can't wait years for grid interconnection.
According to market data, WTI crude traded at $71.50 per barrel on Thursday, up 0.6%, while Henry Hub natural gas settled at $3.25 per MMBtu, down 2.4%—reflecting the complex energy dynamics as AI infrastructure reshapes demand patterns across fuel types.
The question isn't whether AI will transform the energy system. It's whether the energy system can transform fast enough to keep pace.