NVIDIA's $53B AI Investment Spree: Boom or Bubble?

NVIDIA has invested $53 billion across 170 deals from 2020 to 2025, backing AI startups that overwhelmingly spend that capital on NVIDIA GPUs. The company also pledged up to $100 billion to OpenAI in September 2025 — but five months later, no contract had been signed and no money had changed hands. Critics argue the chipmaker is financing its own demand in a closed loop that echoes the vendor-financing bubble that destroyed telecom in 2000.

The numbers are staggering even by NVIDIA's standards. In a single year through late 2025, the company deployed $23.7 billion through 59 deals. Its venture arm, NVentures, alone closed 30 deals in 2025 and roughly 20 more in the first five months of 2026. The pace is accelerating, not slowing — which is precisely what makes analysts nervous.

Key Takeaways

• NVIDIA invested $53 billion across 170 deals from 2020 to 2025, with $23.7 billion in a single year through late 2025.
• The $100 billion OpenAI commitment announced in September 2025 remained unsigned with zero capital deployed five months later.
• Most of the OpenAI money was reportedly earmarked for leasing NVIDIA chips — a textbook example of circular financing.
• Deal velocity is rising: 67 venture deals in full-year 2025, up from 54 in 2024, with NVentures adding ~20 more by mid-2026.
• The FTC and European regulators are intensifying scrutiny of these interlinked AI capital flows.
• The structure mirrors the Lucent/Nortel vendor-financing chains that inflated and then destroyed the late-1990s telecom sector.

How the Closed Loop Works

The mechanics are simple. NVIDIA writes a check to an AI startup. That startup spends the money on NVIDIA GPUs. NVIDIA books the GPU sale as revenue. CEO Jensen Huang frames these as strategic, arm's-length minority stakes in durable infrastructure. But when most of the capital flows right back to NVIDIA's top line, the distinction between "investment" and "customer financing" gets thin.

Analysts mapping money flows between NVIDIA, OpenAI, Oracle, CoreWeave, Microsoft, and Anthropic have produced what insiders call a "spaghetti diagram" — a web of interlocking capital commitments that forms a suspiciously closed financial loop. NVIDIA backs startups. Startups lease chips. Chip revenue funds more investments. As one analyst put it: "Circular artificial intelligence agreements set off alarm bells among investors."

Is This the Telecom Bubble Again?

Yes and no. In the late 1990s, Lucent and Nortel lent regional telecom carriers money to buy their networking equipment. The carriers booked the gear as infrastructure buildout; the equipment makers booked it as revenue. When carriers ran out of cash and demand evaporated, the lenders discovered they'd financed a phantom market. NVIDIA's structure inverts the arithmetic — it's an equity investor, not a lender — but the risk is identical. If AI startups' revenue doesn't materialize or they can't raise fresh capital, NVIDIA loses both the chip sales and the investment.

The comparison gained traction after NVIDIA pledged $5 billion to Intel — a struggling legacy chip competitor — in the same week it announced the $100 billion OpenAI commitment. Two opposite ends of the semiconductor industry, same news cycle, blurred into one statement of dominance. AMD has followed a similar playbook with its $5 billion Anthropic stake, suggesting vendor financing may be becoming an industry norm rather than an NVIDIA-specific gamble.

The $100 Billion That Hasn't Moved

The OpenAI headline is the starkest example of what might be called "announcement capital." In September 2025, NVIDIA committed up to $100 billion to OpenAI. Five months later: no signed contract, no money deployed. Whether the delay reflects regulatory review, structural renegotiation, or cold feet, it illustrates how much of the AI investment boom exists as press releases rather than wire transfers.

And the details that did leak weren't reassuring for skeptics. Reports indicated most of the OpenAI money would be used to lease NVIDIA chips — meaning the "investment" was really prepaid GPU revenue wearing a venture-capital costume. That's not inherently fraudulent, but it does complicate NVIDIA's revenue growth narrative.

Why Bulls Still Have a Case

There's a legitimate counter-argument. NVIDIA is a cash-rich dominant incumbent deploying capital during a genuine platform shift. Intel and Microsoft both made aggressive venture investments during prior technology transitions. Minority stakes lock in customer relationships, secure chip allocation commitments, and provide equity upside if AI's trajectory holds. NVIDIA's 9.3% stake in Nebius and its broader portfolio give it visibility across every layer of the AI stack — from foundation models to cloud infrastructure to enterprise applications.

Huang's argument is straightforward: these companies are building real infrastructure that will generate real revenue, and NVIDIA is positioned as both supplier and shareholder. If the AI boom is durable, the circular capital flow is just a feature of being the dominant platform — the way Google's ad revenue funded the Android ecosystem, or AWS's margins subsidized Amazon's retail business.

What to Watch Next

The signal that separates a real boom from a circular one is contract execution. Track whether NVIDIA's $100 billion OpenAI commitment actually converts into a signed deal with defined payment schedules — or whether it quietly shrinks or restructures. Separately, watch FTC and European regulatory actions; both are reportedly intensifying scrutiny of these interlocking capital flows. And keep an eye on whether AI startups in NVIDIA's portfolio can generate independent revenue that doesn't ultimately trace back to an NVIDIA check. If they can, Huang is right and this is smart platform investing. If they can't, we've seen this movie before — it ended in 2000, and the credits weren't pretty.

Educational content only — not financial advice.