Hyperscalers are getting involved in earlier stages of power development.
Tech giants are sick of waiting around for electricity. But locking in future power means taking on more upfront risk.
The race to build up more AI data centers has created a strain on the existing power grid while also bottlenecking the ambitious plans of the world's largest tech companies. AI systems consume far more energy than more typical servers and other computing gear, but new energy generation facilities don't exactly go up overnight. As a result, companies better known for driving advertising clicks and social network likes are now diving into the power business.
In a landmark deal, Google-parent Alphabet last month agreed to buy renewable energy developer Intersect Power for $4.75 billion, plus the assumption of debt. It was the first time a tech company brought an energy developer in-house. The deal was a surprise for the energy industry: "The market always thought [tech companies] would outsource because development is too nitty-gritty, kind of like the real-estate business," said Prashant Khorana, director of power and renewables consulting at Wood Mackenzie.
Other hyperscalers aren't quite going that far, but are getting more involved in energy projects. Amazon is poised to buy a 1.2 gigawatt development-stage solar project with battery storage capacity in Oregon after winning it recently in a bankruptcy auction. In 2024, Amazon agreed to fund early development for a project developed by small modular reactor company X-Energy, in which it also owns an equity stake. Meta last week announced that it would be funding the development of small modular reactors from Oklo and TerraPower.
This is a departure from the more passive role tech companies have played in past energy projects. In the traditional model, developers and external investors -- often infrastructure funds and banks -- took on the risk of developing and constructing the project, with financing backstopped by the creditworthy tech company's power purchase agreement. This model worked just fine a couple of years ago when tech companies were occasionally looking for power purchase agreements to get green credentials rather than to fill immediate power needs.
Now, electricity is one of the main barriers to hyperscalers' AI build-out. And in the case of Google's parent, its latest effort means energy will become a capital expenditure, not just an operating expense.
Capex on energy projects is nothing to sneeze at: If the company plans to build out the multiple gigawatts of energy in Intersect Power's pipeline, that could represent billions of dollars of investment beyond what it paid for the acquisition. And Alphabet is already spending record amounts on capex to fund its AI build-out, as are its peers. According to estimates from Visible Alpha, Alphabet spent about $91 billion on capex in 2025 -- nearly triple its annual average over the prior five years.
Meta and Amazon haven't disclosed how much of the development cost they are shouldering for SMR projects, but these can be high-risk, substantial outlays. Early development capital, or funding that is needed to know if an SMR project is even viable, can run from $500 million to $600 million, said Ted Brandt, chief executive officer of Marathon Capital, a clean energy-focused investment bank. These costs include permitting at local and federal levels, testing and engineering to get licensing from the Nuclear Regulatory Commission, and site preparation.
One natural advantage that tech companies have is their massive cash pile and high credit rating. In a note discussing Google's Intersect Power deal, Jefferies analysts said the deal "emphasizes the quantum of capital needed by project developers, and the appeal of simply raising this with the hyperscalers" rather than through infrastructure funds or public markets. Hyperscalers' cost of capital is cheaper than that of a typical developer that needs to raise external equity and debt, according to an industry banker.
And among its deep-pocketed peers, Google's are the deepest -- which puts the company in the best position to try this new approach. The company has about $141 billion in cash net of debt, the highest balance by far of any of its megacap tech rivals. It also currently sports the highest annual operating cash flow -- $151 billion -- of any public company, according to data from S&P Global Market Intelligence.
By focusing on new energy projects, tech companies also get to sidestep thorny affordability questions that arise when they try to strike deals with existing power plants. On Monday, President Trump said in a Truth Social post that the administration is working with tech companies to "ensure that Americans don't 'pick up the tab' for their power consumption." Notably, Alphabet made it clear in its announcement that it bought Intersect Power's new development pipeline, not its operating assets, which will go to another group of investors.
The flip side is that if the AI business stalls, tech companies could be left with more than just data centers as stranded assets. But it is the opposite scenario -- losing the AI race -- that is giving them more sleepless nights these days.
Comments