20 May 2026. Jensen Huang does not believe in picking winners. “There are so many great, amazing foundation model companies, and we try to invest in all of them,” Nvidia’s founder and CEO said during an April podcast appearance. “We don’t pick winners. We need to support everyone.”
The numbers behind that philosophy are extraordinary. Nvidia has committed more than $90 billion in acquisitions, investments, and strategic deals in 2026 alone — a figure that places it among the most active dealmakers of any company, in any sector, in the world this year. The pace has accelerated so dramatically that Nvidia’s non-marketable equity securities — its private company investments — swelled from $3.39 billion at the start of fiscal 2025 to $22.25 billion by the end of January 2026. In the months since, the commitments have continued to accumulate at a pace that has left analysts scrambling to model the implications.
What Nvidia Is Buying and Why
Join The European Business Briefing
New subscribers this quarter are entered into a draw to win a Rolex Submariner. Join 40,000+ founders, investors and executives who read EBM every day.
SubscribeThe investment spree spans the entire AI value chain. Corning, the glass and optical fibre manufacturer, received a commitment of up to $3.2 billion — securing the supply of the photonic infrastructure that connects data centres at the speeds AI workloads require. Data centre operator IREN received a right-to-invest agreement worth up to $2.1 billion, with IREN committing to deploy up to 5 gigawatts of Nvidia’s DSX infrastructure design in its facilities worldwide. OpenAI and Anthropic — the two most prominent large model companies and significant buyers of Nvidia’s GPU hardware — are both in the portfolio. So are dozens of earlier-stage AI startups.
The pattern is consistent. Nvidia is investing across chips, optical communications, data centres, and large language models — a full-industry-chain matrix that mirrors the structure of the AI buildout itself. Every company Nvidia invests in is, to varying degrees, a customer for Nvidia’s hardware. And every dollar of Nvidia investment that flows into those companies flows back through the system as demand for Nvidia chips.
The Circular Financing Question
This is where the debate begins. Goldman Sachs, Wedbush Securities, and Mizuho analysts have all raised the same concern, likening Nvidia’s investment model to the vendor financing that characterised the dot-com bubble — where technology companies provided capital to customers, who used that capital to purchase the technology company’s products, which inflated revenue figures that justified valuations that required more vendor financing to sustain.
Matthew Bryson at Wedbush called it “squarely the circular investment theme.” The logic is simple and uncomfortable: if Nvidia is investing billions into companies that then spend those billions on Nvidia GPUs, the revenue growth Nvidia reports is partly self-generated. That does not make the GPUs less real or the AI applications less valuable — but it does raise questions about how sustainable the demand profile is at current pricing, and whether the investment gains Nvidia reports are independently generated or an artefact of its own dealmaking.
Huang dismissed the characterisation directly. “Ridiculous,” he told Bloomberg, insisting that Nvidia’s investments represent a small fraction of the total capital these companies require and are a vote of long-term confidence in generational businesses, not financial engineering. The point has merit — Nvidia generated $97 billion in free cash flow in its last fiscal year, and the companies it invests in have independent funding sources far larger than any Nvidia cheque. The SoftBank model of making massive conviction bets on transformational technology is built on similar logic — and has produced similar scepticism alongside similar returns.
The Intel Trade as Proof of Concept
The most striking evidence in Huang’s favour is the Intel position. Nvidia’s $5 billion bet on Intel — taken when Intel was under severe pressure and trading at distressed valuations — is now worth over $25 billion. That is a return that has nothing to do with GPU sales or circular financing. It is a straight equity investment that has compounded spectacularly as Intel’s recovery story gained credibility. As a proof of concept for Nvidia’s capacity to generate independent investment returns, it is difficult to argue with.
The broader portfolio has also performed. Nvidia’s non-marketable equity securities generated significant unrealised gains in the most recent quarter, contributing meaningfully to net other income. The financial case for the investment programme is real — the question is whether the returns are being generated by genuine market insight or by Nvidia’s own market power creating self-fulfilling demand.
What This Means for European AI
The $90 billion commitment is concentrated almost entirely in American companies and American infrastructure. European AI startups, European data centre operators, and European semiconductor manufacturers are largely absent from Nvidia’s dealmaking map. The EU’s own AI investment programme — attempting to close the gap with US hyperscalers through the Draghi competitiveness framework — operates at a fraction of the scale Nvidia alone is committing in a single year.
That asymmetry has consequences. The NextEra-Dominion deal that restructured American power infrastructure around AI demand and the SpaceX IPO that crystallised $20 billion in hedge fund returns are expressions of the same underlying dynamic: American capital, at American scale, building American AI infrastructure with American chips. Europe is watching a race being run on a track it has not yet fully entered.
Jensen Huang is spending $90 billion to make sure that wherever AI goes next, Nvidia’s ecosystem is already there waiting for it. Whether that is visionary capital allocation or the most sophisticated vendor financing scheme in corporate history may ultimately depend on whether the AI boom sustains the valuations that justify it.
Related Analysis
