Nvidia disclosed during its fiscal Q2 2027 earnings call that CEO Jensen Huang signed commitments totaling $18 billion for investments in AI model developers and infrastructure partners through the remainder of the fiscal year. The disclosure named August as the month Huang finalized the bulk of the deals. The commitments are structured as a mix of equity stakes, convertible instruments, and prepaid compute credits—tools the company has used since 2023 to lock in demand for its H100 and H200 clusters.
The $18 billion figure represents roughly 14 percent of Nvidia's trailing twelve-month free cash flow and sits on top of the $12 billion the company deployed in similar arrangements during fiscal 2026. CoreWeave, one of Nvidia's largest infrastructure customers and a vehicle for its GPU-as-a-service model, announced within hours of the earnings call that it plans to raise $3 billion through convertible debt and up to $2.92 billion through equity. Nvidia owns approximately 5 percent of CoreWeave after a $100 million investment in 2023 and a follow-on in early 2025. The timing is not coincidental—CoreWeave's debt will price in September, and proceeds are earmarked for data center expansion in Texas and Norway, both sites where Nvidia chips are the exclusive accelerator.
The capital deployment matters because Nvidia is now the primary liquidity source for the firms that buy its hardware. The model resembles Intel's venture arm in the 1990s, except the sums are an order of magnitude larger and the investees are also the customers. This creates a feedback loop: Nvidia funds the buyers, the buyers order more chips, and the revenue growth justifies the capital deployment to public-market investors. The risk is concentration. If one counterparty stumbles—say, a model maker whose inference economics break at scale—Nvidia holds both the receivable and the equity stub. The company's CFO noted on the call that 87 percent of the $18 illon is allocated to infrastructure partners, not pure-play model developers, which suggests Huang is hedging toward the infrastructure layer where utilization is more predictable.
Operators and allocators should track CoreWeave's convertible pricing in mid-September and whether the structure includes Nvidia participation or a sweetener tied to chip delivery. The second item is Nvidia's fiscal Q3 guidance for deferred revenue, which will clarify how much of the $18 billion is recognized upfront versus amortized. The third is any 13F filings in mid-November that show whether Nvidia took new stakes in private AI labs during the August window. Huang has historically used direct equity when he wants board influence—convertibles when he does not.
The $18 billion is not patient capital. It is a structured demand-creation engine with eighteen-month payback windows and hardware lock-ins embedded in the term sheets.