Meta's Nuclear Bet: How Big Tech Is Financing the Grid

Zach Martin

Welcome to Nuclear Capital on KMKT I R Hub Radio. I'm Zach Martin, and it's Tuesday, October 6, 2026. Before we get into it, a clear note. Every company we discuss today is a case study, for informational purposes only. Nothing here is investment advice, and we are not making buy or sell recommendations, ratings, or price targets.

Emily Martin

And I'm Emily Martin. Today we're treating Meta's nuclear deals as a capital structure story, not a product launch. Nobody is unveiling a gadget. What we have is a stack of contracts, some tied to plants that already run and some tied to plants that don't exist yet. The question is what those contracts actually do for the people who have to finance them.

Zach Martin

Let's lay out the stack first. Based on the reporting from Utility Dive, Meta struck deals with Oklo, Vistra, and TerraPower to supply up to 6.6 gigawatts of nuclear power by 2035. All of it is aimed at the PJM region. It follows a December request for proposals where Meta said it wanted developers who could help accelerate new nuclear generation and create enough scale, by deploying multiple units, to bring costs down materially.

Emily Martin

And it sits on top of something from June. Meta signed a twenty year nuclear deal with Constellation to supply its data centers in Illinois. So this is a pattern. One deal is a procurement decision. A string of them starts to look like a financing strategy.

Zach Martin

Walk me through the three layers, because they're very different animals. Start with Oklo.

Emily Martin

Oklo is the greenfield piece. The company described a 1.2 gigawatt power campus in Pike County, Ohio, to support Meta's data centers in the region. That includes the Prometheus AI supercluster in New Albany, Ohio, which is expected to come online in 2026 and consume at least one gigawatt. Now look at the timeline Oklo gave. Pre construction and site characterization are slated to begin in 2026. The first phase is targeted to come online as early as 2030, and the campus is expected to build out incrementally to the full 1.2 gigawatts by 2034.

Zach Martin

Notice the words there. Slated. Targeted. As early as. That's the language of a plan, not a delivery. And I want listeners to hold one picture in their heads. Prometheus is expected online this year. The Oklo campus serving that region starts its first phase, at best, around 2030. The load shows up years before the generation does.

Emily Martin

Which is exactly why the second layer matters. Vistra is the existing fleet and uprate piece. Meta described twenty year agreements to purchase more than 2.1 gigawatts from Vistra's Beaver Valley plant in Pennsylvania and its Perry and Davis Besse plants in Ohio, plus uprates at the Ohio plants. Those are operating reactors. The electrons are real today, and the uprates add some incremental capacity on top.

Zach Martin

Think of it like a training plan. Vistra is the miles you've already run. Oklo is the marathon you've registered for. Both go in the log, but they carry very different odds when you toe the line.

Emily Martin

I like that, and the third layer stretches the analogy. TerraPower is the advanced reactor piece. Meta said its agreement provides funding to support development of two TerraPower sodium fast reactors, each 345 megawatts, the Natrium design. Those two units won't start supplying power until at least 2032. And Meta also gets rights to energy from up to six other Natrium units, capable of producing 2.1 gigawatts and targeted for delivery by 2035.

Zach Martin

So now the capital markets question. Why does any of this matter to a lender or an investor? Because a hyperscaler contract is a form of credit. A developer with no revenue yet, and Oklo and TerraPower are both in that category as developers of unbuilt plants, has a hard time raising project debt on a promise. A twenty year agreement with one of the largest technology companies in the world turns a hope about demand into a counterparty a bank can underwrite against.

Emily Martin

Right, and the mechanism is worth saying slowly. A merchant generator sells into a market where prices move. A long contract with a creditworthy buyer converts volatile exposure into something closer to contracted cash flow. For a fleet owner like Vistra, that can change how the existing plants are valued and funded. For a pre revenue developer, it can be the difference between a drawing and a financeable project, because it helps support raising debt, raising equity, and ordering long lead components.

Zach Martin

And Meta's own framing is that it wants to meet the demand from the data centers it intends to bring online over the next decade without leaving consumers to bear the expenses. That's Meta's statement, so listeners should weigh it as a stated goal. But structurally, it points to a model where the large buyer carries more of the commitment so ordinary ratepayers don't.

Emily Martin

Here's where I want to slow down, though, because credit is not the same as completion. A signed offtake guarantees a buyer for future power. It guarantees zero electrons. Between a 2026 announcement and 2030 or 2035 delivery sits a gap of four to nine years, and a lot can happen in it.

Zach Martin

Let's name the risks one at a time. First, regulatory. Advanced designs face licensing milestones, review periods, and design certification steps. Each one is a gate, and a gate can move a schedule even when nothing is wrong with the engineering.

Emily Martin

Second, construction and supply chain. Nuclear construction has a long history of cost overruns and delays. So the question I'd ask of any of these deals is who absorbs the inflation. If costs rise, does it land on the developer's balance sheet, or does the contract let the price adjust toward the buyer? And I'll be direct. The public reporting we're working from does not disclose those terms. We know Meta is funding TerraPower development for the first two units, which means Meta carries some development risk there. But beyond that, the allocation of cost overrun risk is not something we can tell you from what's public.

Zach Martin

That's an important discipline point. When you can't see the term, you don't fill in the blank with a comfortable assumption. You write it down as an open question. Because a take or pay style comfort, if it exists, is only as strong as the contract language, and we haven't seen the language.

Emily Martin

Third risk, and this one is about the grid, not the reactor. All 6.6 gigawatts is pointed at PJM. And a market monitor reported that data centers accounted for about 40 percent of PJM capacity costs in its last auction. So the region is already feeling the load. Interconnection queues, transmission bottlenecks, and local grid stability are gating factors that no hyperscaler contract can wave away.

Zach Martin

Which gives us a clean way to say what an offtake does and does not do. It does give a developer a bankable demand anchor. It does help justify ordering equipment and raising capital. It does signal to regulators and to suppliers that the load is serious. What it does not do is license the reactor, finish construction on time, secure fuel, or connect the plant to the grid. Those are separate jobs, with separate owners.

Emily Martin

And that's the difference between the announcement date and the energized date. Everything we've said so far lives on the announcement side. The energized side is earned one milestone at a time.