Nvidia just dropped another monster quarter, and buried in the filing is a number that should grab every AI operator’s attention: the company’s stakes in private startups nearly doubled in three months, jumping from $22 billion to $43 billion. According to TechCrunch AI, Nvidia poured $18.5 billion into private equity purchases between January and April alone, compared to just $649 million the previous quarter. That’s a 28x acceleration in startup checks.
The headline numbers were already eye-watering. Revenue hit $81.6 billion for the quarter ending April 26, up 20% sequentially. Data center revenue set a record at $75.2 billion. The board authorized another $80 billion in share buybacks. Next quarter’s guidance: $91 billion, which would be 12% growth, signaling the company itself expects the curve to bend.
What stands out
The private equity story is the real news. Nvidia isn’t just selling shovels anymore. It’s quietly becoming one of the largest AI venture investors on the planet, and that $43 billion figure doesn’t even include the headline deals everyone’s been talking about.
The excluded pieces:
- The $30 billion OpenAI commitment announced in February (structure still undisclosed)
- Public stakes in Corning and IREN
- A “significant” buildout for Anthropic, which Jensen Huang said had “largely zero” coverage before this
- Future commitments that haven’t closed yet
Add those up and Nvidia’s exposure to the AI ecosystem it powers is closer to $100 billion than $43 billion.
Why this matters
Nvidia is now the customer, the supplier, and the investor in a closed loop. When Nvidia backs OpenAI or Anthropic, those companies turn around and buy more Nvidia chips. The capital flows in a circle, and Nvidia sits at the center collecting margin on every lap. CFO Colette Kress put it plainly on the call: “Our Blackwell architecture is everywhere, adopted and deployed by every major hyperscaler, every cloud provider, and every major model maker.”
This is the same playbook Microsoft ran with OpenAI, except Nvidia is doing it across the entire industry simultaneously. Regulators in the US and EU have been circling these compute-for-equity arrangements for months. Expect that scrutiny to intensify.
The China question
H200 chips have been approved for US export to China, but Kress confirmed Nvidia has generated zero revenue from that channel and remains “uncertain whether any imports will be allowed.” Translation: the geopolitical overhang isn’t resolved, and Nvidia is hitting these numbers without China meaningfully contributing.
What to watch next
Three things AI practitioners and founders should track:
- The Anthropic buildout. Huang said capacity coming online “this year and next year is going to be quite significant.” That’s fresh competitive pressure on OpenAI’s compute moat.
- Slowing growth. A 12% quarter-over-quarter forecast is still extraordinary, but it’s the first sign Nvidia sees a ceiling forming. Watch whether enterprise demand picks up the slack as hyperscaler capex normalizes.
- Concentration risk. If you’re a startup competing with a Nvidia portfolio company, you’re now competing with your chip supplier’s balance sheet.
The broader picture: Nvidia isn’t just riding the AI wave anymore. It’s actively financing the surfers. Full breakdown at the original TechCrunch AI report.