Nvidia is pulling back from investments in OpenAI and Anthropic. Jensen Huang's reasoning sounds plausible: as soon as both companies go public, the investment window closes. Logical, clean, done.
Except that's not quite right.
#The official explanation isn't enough
Nvidia already earns absurdly much money with the chips OpenAI and Anthropic run on. So why go even deeper? At first glance, the withdrawal makes sense.
But minority stakes in the most powerful AI labs in the world aren't pure financial decisions. They're strategic. Nvidia spokespeople, after Huang's statements, pointed to a quote from the last earnings call: all investments are aimed at expanding and deepening the ecosystem. That sounds like an answer, but it isn't.
Because if the ecosystem argument applies, why does it suddenly no longer apply?
#The shovel seller needs no favourite mine
There's an old image from the gold rush: whoever sells the shovels doesn't have to know who finds gold. He earns regardless.
Nvidia is exactly that — the infrastructure behind the AI boom. H100, H200, Blackwell: whoever seriously trains AI today buys from Nvidia. OpenAI buys there. Anthropic buys there. Google buys there. Microsoft buys there.
As long as that stays this way, neutrality is the most profitable position. As a co-owner of OpenAI or Anthropic, the relationship becomes more complicated — not simpler.
#The real problem: own chips
Here lies the core. Microsoft, Google and Amazon are all building their own AI chips. TPUs, Trainium, Maia — the big tech corporations want to become more independent of Nvidia.
Why shouldn't OpenAI and Anthropic do the same? Both are growing fast, both have massive compute costs, both have the incentive to reduce dependence on a single supplier.
If Nvidia is a co-owner of OpenAI and OpenAI simultaneously develops its own hardware, an uncomfortable tension arises. You're supplier and investor at the same time — and eventually these roles collide.
The smartest move then is: out before it gets uncomfortable.
#What Jensen Huang doesn't say
MIT professor Michael Cusumano described Nvidia's original investment of up to 100 billion dollars in OpenAI as "kind of a wash" — essentially a circular transaction. Nvidia invests in OpenAI, OpenAI buys Nvidia chips, the money flows back.
That's not a bad deal. But it's also no argument for deeper stakes. It's an argument for keeping the customer relationship clean.
And that's exactly what Nvidia is now doing. The withdrawal is no sign of disinterest. It's a signal that Nvidia has understood the rules of the game.
#Strategic neutrality as competitive advantage
Nvidia simply can't afford to take sides. Not because Jensen Huang doesn't have a favourite. But because every public taking of sides alienates customers.
Imagine Nvidia holds a significant stake in OpenAI — and simultaneously Google is negotiating the next chip deal. Do you think that makes the negotiation easier?
Neutrality isn't a compromise here. It's the strategy.
#What you can take from this
This principle applies far beyond Nvidia. As a designer, developer or service provider, you know the pattern: as soon as you get too deep into a client's business, you lose flexibility.
Three thoughts on that:
Supplier or partner? Both roles have their place — but don't mix them unconsciously. Whoever sells infrastructure should think carefully about when a stake makes sense and when it weakens their own position.
Neutrality has a price, but also a value. You forgo upside when you don't invest. But you keep freedom of action. For Nvidia, the second point clearly outweighs right now.
Communicate withdrawals. Nvidia officially says little. That generates speculation. When you withdraw from a partnership, clear communication is worth it — even if you don't explain everything.
Nvidia's withdrawal is no closure. It's a repositioning. And that only makes sense when you understand which game is really being played here.
