Nvidia Steps Back From AI Lab Investments as OpenAI and Anthropic Eye IPOs

Jensen Huang delivered what sounded like a routine observation at the Morgan Stanley Technology, Media and Telecom Conference on March 4, 2026. Nvidia's recent investments in OpenAI and Anthropic, he said, would 'probably' be the company's last private stakes in AI labs. The reason offered was mechanical: initial public offerings close the window for private capital. But as TechCrunch noted, the statement 'raises more questions than it answers', and those questions deserve unpacking.

Nvidia has committed $40 billion to the two leading AI research labs. In OpenAI's $110 billion round, Nvidia invested $30 billion alongside Amazon's $50 billion and SoftBank's $30 billion. Nvidia separately invested $10 billion in Anthropic. These are substantial positions by any measure, yet they represent less than 19 percent of a single year's revenue for a company that generated $215.9 billion in fiscal 2026.

The Arithmetic of Strategic Withdrawal

The timing warrants attention. OpenAI is targeting an IPO valuation exceeding $840 billion in 2026. Anthropic, the research lab founded by former OpenAI executives and backed heavily by Amazon, is seeking a valuation above $360 billion. Combined, these two offerings would represent more than $1.2 trillion in market capitalisation, dwarfing most sectors of the Australian economy. Nvidia, having secured positions in both companies at earlier valuations, would appear well-placed to maintain or expand its holdings through public market participation.

Instead, Huang signalled retreat. The stated logic — that IPOs terminate the private investment phase — is accurate but incomplete. Companies regularly anchor their shareholder registers with strategic partners through IPO allocations and follow-on offerings. Microsoft, Amazon, and Google have maintained and expanded their AI lab stakes through multiple financing rounds. Nvidia's decision to step back suggests either a different assessment of value or a different set of constraints.

Jensen Huang discusses Nvidia's AI investments at the CNBC interview

Customer, Supplier, Competitor

Nvidia's relationship with the AI labs has always contained an inherent tension. OpenAI and Anthropic are among Nvidia's largest customers, purchasing the H100 and now Blackwell GPUs that powered Nvidia's revenue to $68.1 billion in Q4 FY2026 alone. At the same time, these labs are developing custom silicon and investigating alternatives to Nvidia's hardware dominance. Google's TPUs, Amazon's Trainium chips, and OpenAI's rumoured internal chip programme all represent existential threats to Nvidia's core business.

An investor in a company that is simultaneously a critical customer and an emerging competitor occupies an awkward position. The discomfort becomes more acute when those companies go public and their strategic decisions face quarterly scrutiny. Nvidia's withdrawal from active investment could reflect a preference for commercial clarity over the complications of cross-ownership, particularly as antitrust scrutiny of big tech intensifies globally.

That gap isn't about brand size — it's about who owns the relationship with the AI systems answering user queries.

— Jensen Huang, Morgan Stanley TMT Conference, March 4, 2026

The Pledge That Shrank

A detail worth noting: Nvidia originally pledged $100 billion to OpenAI's round before ultimately investing $30 billion. The $70 billion reduction has not been publicly explained. It could reflect prudent capital discipline, recognition that the original commitment was excessive for a non-core activity, or a negotiation outcome driven by other factors entirely. What it does confirm is that Nvidia's initial enthusiasm for holding equity in AI labs moderated substantially before the deal closed.

Historical parallels are instructive. Intel spent the 2000s and early 2010s making strategic investments in software companies, digital media ventures, and mobile chip designers, most of which delivered poor returns and distracted management attention. AMD, by contrast, stayed narrowly focused on its core semiconductor business during its turnaround under Lisa Su. Nvidia under Huang has historically resembled AMD's discipline more than Intel's sprawl. The decision to cap AI lab investments may represent a reassertion of that focus.

What the Market Is Pricing

If OpenAI achieves an $840 billion IPO valuation and Anthropic reaches $360 billion, Nvidia's $40 billion in combined investments would represent minority stakes in companies valued at 30 times its entry price or more, depending on the entry valuation terms. On paper, the returns could be extraordinary. In practice, public market valuations of this magnitude would make these among the most expensive companies in global markets, trading at multiples that assume decades of flawless execution.

Nvidia's caution may reflect a view that these valuations price perfection, and that the risk-reward calculus favours selling GPUs to the AI labs rather than owning equity in them. A chipmaker's margin structure is more predictable than a research lab's; Nvidia's gross margins exceeded 70 percent in fiscal 2026, while OpenAI reportedly burns billions annually and has yet to demonstrate sustainable profitability. Huang has built a $2.5 trillion company by focusing on hardware that every AI lab requires. The decision to avoid deeper entanglement with any single lab preserves optionality across a customer base that may look very different in five years.

Nvidia's Q4 FY2026 revenue of $68.1 billion would rank as the eighth-largest quarter ever posted by any technology company.