The $32 Billion Blank Cheque

Here's a sentence I had to read three times before it stopped sounding fake: a company with about fifty employees, no product, and not a single published paper is worth $32 billion, and the man who runs it briefly fired the CEO of OpenAI as a side quest before starting it. That's not my line, by the way. That's roughly how Yahoo Finance's own headline put it, and business journalists don't usually write headlines that read like a joke unless the underlying fact actually is one.

The company is Safe Superintelligence, SSI for short, founded in 2024 by Ilya Sutskever, formerly OpenAI's chief scientist and one of the small group of board members who voted to remove Sam Altman that November before the whole thing reversed inside a week. Since founding, SSI has raised roughly $6 billion at that $32 billion valuation, shipped nothing you can go touch, and published nothing you can go read. Then on July 27, 2026, Nvidia announced it's adding another $5 billion on top of that, for a company whose entire public output remains a one-line mission statement and a closed door.

Ilya Sutskever (left) and Sam Altman, photographed together at Tel Aviv University in June 2023, five months before Sutskever voted to remove Altman as OpenAI's CEO. Photo by Eladkarmel, licensed under CC BY-SA 4.0.

I don't think the number is actually the interesting part, even though it's the part every headline leads with. The interesting part is what the number is paying for.

---

Every other company at this scale sells you evidence before it sells you a story. OpenAI has ChatGPT. Anthropic has Claude. Even Thinking Machines Lab, which I wrote about a few weeks back, took almost a year in stealth but eventually shipped something you could go touch. SSI's entire pitch, as far as I can tell from the outside, is closer to: this specific person was right about scaling before almost anyone else, so extend him the same trust again, this time with no visible checkpoints along the way.

That's not nothing. Sutskever co-authored some of the papers that the current generation of AI is quite literally built on top of. If you're going to bet on one person's judgment being worth ten figures with zero intermediate proof, his name is about as strong a case as exists. But it's still worth noticing how unusual it is that a market this size will fund total silence, when it demands weekly shipping cadences and public benchmarks from almost everyone else in the room.

Meta clearly agreed the name was worth more than the silence around it. In June 2025, Mark Zuckerberg's team reportedly tried to buy SSI outright. Sutskever said no. So Meta pivoted and simply hired away SSI's co-founder and sitting CEO, Daniel Gross, straight into its own superintelligence lab instead. Sutskever confirmed the whole thing himself on social media, plainly: "Meta tried to buy us." He took over as CEO the following month. You can lose the acquisition and still lose the executive, apparently, and the company just absorbs it and keeps raising billions regardless.

---

The part that actually made me sit with this for longer than I expected wasn't the funding round or the Meta drama. It was a specific line from Sutskever's November 2025 conversation with Dwarkesh Patel, the one where he said, on the record, that the "age of scaling is over" and the "age of research has begun." Read that again slowly. That is the founder of a company built entirely on the premise of scaling one research bet in total isolation, telling the world that the scaling era, the thing that got him funded in the first place, just ended. If the person holding the eleven-figure checkbook is the one saying the old playbook is done, that tells you something the funding announcement doesn't: even total conviction has a shelf life once you're the one sitting inside it, watching the runway and not the headlines.

I keep landing on the same read of this. SSI isn't really being valued as a product company, or even really as a research company in the normal sense. It's being valued as an option on one man's judgment, priced by people who've decided that scarcity (there's exactly one of these to buy) is worth more than proof. Options can pay off enormously. They can also expire worthless. What makes this one strange is how comfortable everyone involved seems with not knowing which for a while yet.

---

I don't have a tidy conclusion here, and I think that's honest rather than lazy. If SSI eventually opens that door and something real is behind it, this becomes the story venture capital tells about disciplined patience for the next decade. If it doesn't, this is the cleanest case study the industry has ever produced for a simple idea: a great enough reputation can get funded like a product, priced like a bet, and questioned by absolutely no one, for years, before it has to prove anything at all. Nvidia clearly thinks the odds favor the first outcome. Twice now, with real money. I'm not confident enough either way to bet alongside them. I'm just going to keep watching the door.

FAQ

How much has Safe Superintelligence actually raised, and what has it shipped?

Around $11 billion total: roughly $1 billion at a $5 billion valuation in its 2024 seed round, another $2 billion at a $32 billion valuation in 2025 with Nvidia and Alphabet joining as strategic investors, and a further $5 billion from Nvidia announced in July 2026. In return, it has shipped no commercial product and published no research papers as of this writing.

Why would anyone put billions into a company with nothing to show for it?

The case investors keep making is some mix of Sutskever's track record as one of the researchers whose work the current AI boom is built on, the scarcity of a pure superintelligence-research company existing to invest in at all, and option value: if the bet lands, the payoff is enormous, and if it doesn't, the position was presumably sized for that outcome.

Is Ilya Sutskever still committed to SSI's "no product until we're done" strategy?

Reportedly not entirely. In a November 2025 conversation with Dwarkesh Patel, Sutskever said the "age of scaling is over" and the "age of research has begun," which reads as the founder of a scaling-focused, isolation-first company publicly questioning the very premise SSI was built on.

Did Meta actually try to buy Safe Superintelligence?

Yes. In June 2025, Meta reportedly tried to acquire SSI outright and was turned down. It then hired away SSI's co-founder and CEO, Daniel Gross, into Meta's own superintelligence lab instead. Sutskever confirmed the acquisition attempt himself on social media and became SSI's CEO the following month.

What's Sutskever's connection to the OpenAI leadership crisis in 2023?

He was one of the OpenAI board members who voted to remove Sam Altman as CEO that November. Within days, facing near-total staff revolt and Altman's reinstatement, he said he regretted the decision, and he left OpenAI for good in 2024 to start SSI.