Nvidia CEO Jensen Huang officially signaled that the chip giant is unlikely to invest further in OpenAI or Anthropic, marking a strategic pivot from “Kingmaker” to infrastructure dominant.
Speaking at the Morgan Stanley Technology, Media & Telecom Conference, Huang noted that the window for large-scale equity stakes in these “consequential companies” is closing as both move toward potential Initial Public Offerings (IPOs).
The Final Bets
Nvidia has effectively capped its equity exposure in the two leading AI labs, opting to honor existing commitments rather than pursuing the massive expansion plans previously rumored.
- The OpenAI “Scaling Back”: Nvidia finalized a $30 billion investment in OpenAI’s latest $110 billion funding round (which valued the company at $840 billion). Notably, Huang ruled out a previously discussed $100 billion partnership, calling it “not in the cards” as OpenAI prepares for an IPO as early as late 2026.
- The Anthropic “Final Stake”: Nvidia’s $10 billion investment in Anthropic—announced alongside Microsoft in late 2025—is also expected to be its last. Anthropic is reportedly exploring its own public listing following its massive $19 billion ARR milestone.
Why the Pullback?
Analysts suggest three primary reasons for Nvidia’s shift in strategy:
- IPO Readiness: As these companies mature into trillion-dollar entities, the opportunity for Nvidia to secure “early-stage” equity advantages has passed.
- Regulatory Heat: Direct equity stakes in your biggest customers have drawn intense antitrust scrutiny. By stepping back, Nvidia avoids being seen as “self-dealing” or creating a circular economy.
- Customer to Competitor: As OpenAI moves to develop its own AI chips with Broadcom and Anthropic aligns more closely with Amazon’s Trainium, Nvidia is shifting its capital into the “Arteries of AI”—the networking and optics that no one else can build.
The New Investment Frontier: Optics & Photonics
While pulling back from model labs, Nvidia is aggressively funneling that capital into AI infrastructure hardware. On March 2, 2026, the company announced:
| Investment | Amount | Purpose |
| Lumentum | $2 Billion | To scale U.S. manufacturing of advanced laser components. |
| Coherent | $2 Billion | To develop next-generation optical interconnects for AI factories. |
“Computing has fundamentally changed… software now runs on intelligence with tokens generated in real time. We are investing to ensure the ‘arteries’ of this infrastructure do not become the new scarcity.” — Jensen Huang
The “Vera Rubin” Connection
As part of the finalized $30 billion deal, OpenAI secured dedicated access to 3 gigawatts (GW) of inference capacity and 2 GW of training capacity on Nvidia’s next-generation Vera Rubin systems. This ensures that even without more equity, OpenAI remains tethered to Nvidia’s flagship hardware for the rest of the decade.
