For years, Big Tech told a reassuring story about the future of energy: AI would scale on clean power.
Solar. Wind. Batteries. A smooth, inevitable transition.
That story just collided with reality.
According to Reuters Events (December 11, 2025), the companies building the world’s largest AI models and data centers are no longer betting on a single energy solution. Instead, they are quietly — and rapidly — adopting what executives now describe as an “all-of-the-above” power strategy: renewables plus natural gas plus nuclear.
Not because they want to — but because they have to.
The AI Power Problem No One Can Ignore
AI data centers are not like traditional industrial loads. They don’t power down at night. They don’t tolerate intermittency. And they can’t wait a decade for grid upgrades.
They need:
- Massive power
- 24/7 reliability
- Immediate availability
That combination is breaking assumptions across the energy sector.
The International Energy Agency reports that natural gas–fired plants are already the largest source of power for U.S. data centers, and that this will remain true through at least 2030 — not because gas is “preferred,” but because it is dispatchable and fast to deploy.
Clean power will play a larger role later, the IEA says — once grid constraints, permitting delays, and capacity build-out catch up. But AI is not waiting.
The Scale Is Staggering
According to S&P Global, utility power supplied to U.S. data centers will jump 22% in a single year, reaching 61.8 gigawatts in 2025, and is projected to more than double to 134.4 GW by 2030.
That growth curve is unprecedented.
Utilities are responding the only way they realistically can:
- Building new gas-fired generation for reliability
- Expanding renewables where possible
- Looking to nuclear for long-term baseload stability
Dominion Energy, which serves Northern Virginia’s “data center alley,” reported that 44% of its generation mix was natural gas in 2024, even as it plans to add 47 GW of new generation and storage over the next two decades — roughly 80% carbon-free.
Entergy, serving four South Central states, has announced more than 19 GW of new natural gas capacity, while also advancing solar projects and building new gas plants specifically to power Meta’s Hyperion data center in Louisiana, according to Entergy’s Chief Customer Officer.
Big Tech’s Quiet Pivot
The most telling shift isn’t coming from utilities — it’s coming from the tech giants themselves.
Companies like Meta, Microsoft, Google, and Amazon Web Services have been among the world’s largest buyers of renewable energy for years.
That hasn’t changed.
What has changed is the timeline.
Meta’s Global Head of Energy told Reuters Events that the company is now working with utilities to bring “all forms of power” onto the grid, with a focus on what can be built fastest. Microsoft, despite committing to match all electricity use with zero-carbon energy by 2030, is sourcing power for new data centers from utilities planning additional natural gas plants.
Google, which has matched 100% of its global electricity consumption with renewables since 2017, is now supporting:
- Nuclear plant restarts
- Life extensions of existing reactors
- Small Modular Reactors (SMRs)
- Even early-stage fusion partnerships
In October 2024, Google signed a 25-year power purchase agreement for a nuclear facility in Iowa expected to return online in 2029. In 2025, it expanded its nuclear strategy further, backing both SMRs and fusion as future solutions.
This is not a retreat from clean energy.
It’s an admission that clean energy alone cannot scale fast enough to power AI right now.
Renewables Are Still Growing — But Not Fast Enough
The U.S. Energy Information Administration projects a record 33 GW of new utility-scale solar in 2025, and nearly 200 GW of solar capacity added between 2025 and 2030.
Yet even this historic build-out faces headwinds:
- Grid interconnection backlogs
- Permitting delays
- Increased federal oversight
- Local opposition
- And the phase-out of key tax credits after 2027
As Michael Thomas of Cleanview told Reuters Events, solar, wind, and batteries can get us much of the way there — but barriers are slowing deployment at exactly the wrong moment.
The Nuclear Revival Is No Longer Theoretical
Perhaps the most surprising outcome of the AI boom is the acceleration of nuclear energy — not as a future concept, but as a near-term necessity.
Big Tech is now actively:
- Funding reactor restarts
- Extending the life of existing plants
- Signing long-term nuclear PPAs
- Backing SMR developers
- Laying groundwork for fusion in the 2030s
As Google’s Head of Advanced Energy put it:
“In the near term, life extensions, restarts and upratings will have the most impact.”
That statement alone would have sounded unthinkable a decade ago.
The Uncomfortable Truth
AI is exposing a hard truth the energy industry has danced around for years:
We do not have the power capacity we thought we had.
Not fast enough.
Not reliably enough.
Not at the scale demanded by always-on digital infrastructure.
And building new generation — of any kind — takes time, capital, permits, and political will.
Which brings us to the question few are asking loudly enough.
If We Can’t Produce Enough Power Fast Enough — What Do We Do?
If generation can’t scale overnight, efficiency becomes the fastest lever available.
Not symbolic efficiency.
Not marginal savings.
But real, measurable capacity recovery inside existing electrical infrastructure.
Across the grid today, enormous amounts of power are lost, distorted, or underutilized — through reactive power, poor power factor, unnecessary current draw, and inefficiencies that rarely make headlines but quietly constrain capacity.
When AI demand surges, these inefficiencies stop being academic.
They become limiting factors.
The emerging reality is this:
The AI era won’t be powered by generation alone. It will be powered by optimization.
Before we build more plants.
Before we widen transmission corridors.
Before we wait on permits.
We have to extract more usable power from what already exists.
Because the shock isn’t that Big Tech is turning to gas and nuclear.
The shock is how quickly AI has forced everyone to admit the grid was already stretched thin.
And that realization is only just beginning.
Source: Reuters Events / Thomson Reuters, December 11, 2025. Data from International Energy Agency (IEA), U.S. Energy Information Administration (EIA), S&P Global.
