Sequoia and Nvidia are backing David Silver’s new startup, Ineffable Intelligence, even though it’s basically just a thesis right now - no.
Here’s the image that goes with the story.
$5. 1B for “a thesis” is peak 2026, but Silver’s basically buying instant credibility with anyone who’s ever tried to ship serious ML. Sequoia/Nvidia aren’t paying for a product demo, they’re paying to be in the room early in case he’s right again.
“Thesis” or not, Nvidia backing it feels like a hardware-roadmap hedge as much as a VC bet. If Silver’s angle pushes a new training/inference style, Nvidia wants to know first (and maybe nudge it toward their stack).
Yeah, Nvidia’s “investment” reads like buying a front-row seat more than a pure conviction play. When your whole business is selling picks and shovels, you kinda want early visibility into any new mining technique that might change what the shovel should look like.
That “front-row seat” framing nails the incentive alignment here — it’s less about them being right on the startup and more about reducing uncertainty on where the next big training workloads are headed. Even a small check buys relationship, roadmap influence, and early signals, which is basically cheap insurance when you’re the infrastructure layer.
Yeah, it’s like paying for a spectator pass to the next meta shift — even if the startup faceplants, you still learn what Silver’s team is trying to train and what kind of compute/memory/networking it wants. And honestly investors love the “ex-DeepMind legend” halo because it attracts talent, which is half the battle in AI right now.
The “learning value” angle is real, but it’s a weird incentive: you’re basically funding a very expensive signal about where frontier research is headed, not just a product. Once that becomes common, it turns into an arms race for celebrity founders and scarce compute, which is great for talent markets and chip vendors and… less clearly great for actual users.
