Nvidia Pulls Back from OpenAI and Anthropic, Jensen Huang’s Explanation Sparks New

Nvidia’s strategic retreat from further equity stakes in OpenAI and Anthropic marks a sharp pivot from the “all‑in” narrative that dominated the chip‑maker’s public messaging last year. At the Morgan Stanley Technology, Media and Telecom conference on Wednesday, CEO Jensen Huang told investors the firm’s recent investments are likely to be its last, because the two AI startups are expected to go public later this year and “the opportunity to invest closes” once they list (TechCrunch). The comment underscores a shift from the earlier rhetoric of a $100 billion commitment to OpenAI, a figure that never materialized in cash but was instead framed as a reciprocal purchase‑of‑chips arrangement (Reuters).

Huang’s explanation, while straightforward, raises questions about the economics of Nvidia’s “ecosystem” play. In a transcript of the company’s fourth‑quarter earnings call, Huang said all Nvidia investments are “focused very squarely, strategically on expanding and deepening our ecosystem reach” (TechCrunch). The stakes in OpenAI and Anthropic have already delivered that reach: both firms now run the majority of their workloads on Nvidia’s H100 GPUs, and the partnership has effectively locked the hardware vendor into a dominant position in the emerging generative‑AI market. Yet the financial mechanics of the deal appear to be a wash. MIT Sloan professor Michael Cusumano, cited by the Financial Times, described the original $100 billion pledge as “kind of a wash” because OpenAI agreed to buy an equivalent amount of Nvidia chips, creating a circular investment that adds little net cash to Nvidia’s balance sheet (TechCrunch).

The most recent tranche of Nvidia’s capital infusion into OpenAI further illustrates the scaling back. While the company announced a $30 billion contribution to OpenAI’s $110 billion financing round in February—a sizable sum—it fell short of the earlier, more ambiguous “up to $100 billion” ceiling (TechCrunch). The reduction suggests that Nvidia is calibrating its exposure now that the AI firms are nearing IPO, preferring to reap the upside from chip sales rather than continue pouring equity into companies that will soon be publicly traded and potentially less dependent on Nvidia’s hardware subsidies.

Analysts see the move as a natural evolution of Nvidia’s business model rather than a retreat. By the end of 2025, Nvidia expects its data‑center revenue to exceed $30 billion, driven largely by AI workloads that are already locked in through these ecosystem deals (Reuters). The company’s cash‑flow generation from selling GPUs to OpenAI and Anthropic is likely to dwarf any incremental returns from equity stakes, especially given the high margins on hardware. Moreover, the impending IPOs could introduce market discipline that reduces the need for Nvidia’s “soft‑money” support, allowing the chip maker to focus on next‑generation products such as the upcoming Blackwell architecture.

The broader implication for the AI startup financing landscape is that the era of deep, multi‑year equity partnerships between chip makers and AI firms may be winding down. As OpenAI and Anthropic transition to public markets, they will gain access to a broader pool of capital, diminishing the strategic leverage that Nvidia once wielded through its cash commitments. For investors, the key takeaway is that Nvidia’s upside now hinges more on its ability to supply the next wave of AI hardware than on holding equity in its customers, a reality that aligns with the company’s own statements about “expanding and deepening our ecosystem reach” (TechCrunch).