What good is leverage if you don’t use it? Considering the last two weeks of earnings announcements among the country’s largest tech and utility companies, it’s a question worth asking.
Where AI and the power sector meet, there is an unprecedented abundance of capital ready to support both the construction of data centers and power infrastructure at a scale that could serve an entire city. AI companies in the United States — OpenAI, Anthropic, Google, Meta, Microsoft, and Amazon — continue to increase their projected CapEx.
But are these companies insisting on powering the new infrastructure needed for AI with zero-carbon energy? Their aims feel much less clear today than they once did.
Over the past two years, CapEx spending among the largest tech companies has increased by anywhere from 70% (Meta) to 174% (Microsoft). And it’s expected to increase further.
Alphabet indicated $85 billion in spending in 2025, an increase of $10 billion over previous plans. Microsoft said it had spent $88 billion in the past year, and said it would devote $30 billion in the next quarter alone. Meta said it expects to spend between $66 billion and $72 billion in 2025, while Amazon said it would spend over $100 billion.
And it’s worth noting most of this spending is coming from free cashflow, not debt or equity, so it has less of the whiff of a bubble than previous tech cycles did.
This money, it’s worth noting, is largely being spent domestically and in ways that have aligned with the goals of both the Biden and Trump administrations. AI is considered by many to be the next industrial revolution, and will underpin every sector of the economy — if the doubts of skeptics can be overcome, that is.
And utilities, long in the doldrums of flat demand forecasts, are already reaping the rewards. In just the past week, publicly traded utilities have given powerful support to the idea of their imminent golden age. For example:
- DTE Energy said they were in discussion with hyperscalers for 3 GW of data center load and another 4 GW of power for other data center opportunities; the utility plans to spend $24 billion over the next four years on power infrastructure.
- AEP revised its load growth projections up from 20 GW to 24 GW through 2029, with 10 GW from data centers alone; the company described a $70 billion five-year CapEx plan, up from a previous $54 billion base case.
- FirstEnergy discussed the long-term strength of data center requests, up to 11.1 GW, and said it received over 40 new large load studies in 2025 alone.
- PPL Electric reported advanced-stage agreements for roughly 14 GW of data center load in Pennsylvania, an increase of 32% over the latest quarter, and said it will need around 7.5 GW of new generation in the next five to seven years.
Meanwhile, the AI companies themselves are driving ever-bigger builds. OpenAI is building gigawatt-scale Stargates around the world, and Anthropic says it will likely need 50 GW of data center capacity by 2028 for continued training of its frontier AI models. This feels like market power to me (or “dominance” in the Trump era parlance), and therefore like political power.
This begs a question: Why not leverage that political capital to insist on a net-zero energy supply? It could be waved away as naive, given the popular assumption among U.S. companies and politicians that the economic race against China must be won at all cost. But if speed to power is in fact the priority, given the competition, then isn’t solar-plus-storage the fastest path in the U.S.?
There seems to have been a quick, collective resignation that once Trump took power an embrace of clean energy may run afoul of the administration’s clear preference for fossil energy, and even risk political retribution. But as AI companies continue to grow — the collective market cap of the seven largest tech companies stands at roughly $15 trillion today — their political power grows as well. Couldn’t they, as a group in lockstep, take a stance that any new Trump policies or regulations governing AI and energy should align with their original net-zero goals — regardless of the race with each other, and with China?
That hasn’t happened, of course. Instead, companies have waded through a contradictory and conflicting set of laws and executive orders that make the AI market a considerably more confusing place than it was a mere year ago.
Even gas generation developers face impacts from tariffs, the price and availability of fuel, and shortages of equipment and skilled labor. Developers of renewables obviously have it worse. They face expiring incentives, FEOC restrictions, shifting tariffs, and the potential for a project to be wiped out with the stroke of a pen at the Interior Department.
Nuclear projects may be aided by a permitting “rubber stamp,” but who, really, believes that is the right approach? The same could be said for aging coal plants relieved of their retirement plans. Those should be corner cases, not ideas brought to the front of the line.
As AI companies increasingly reap the benefits of their investments — and all are reporting real revenue growth now — it should follow that their next move should be a demand for cleaner energy, not just from IPPs and utilities, but from Washington. The benefits of using that leverage would accrue to the entire economy, tackling the obvious imperative of slowing climate change, while also addressing the administration goals of energy affordability and the reshoring of manufacturing.
A version of this story was published in the AI-Energy Nexus newsletter on August 6. Subscribe to get pieces like this — plus expert analysis, original reporting, and curated resources — in your inbox every Wednesday.
