Analysis
Nvidia's earnings produced a $400 billion move in market value, and then the footnotes produced a second conversation. The company's disclosed future commitments now total roughly $366 billion: $279 billion in supply and capacity commitments, $29 billion in cloud service agreements, $25 billion in data center leases, $25 billion in equity investments and $8 billion in capex, Fortune reported. The supply and capacity line alone more than doubled from $119 billion a quarter earlier.
Beyond that sit further obligations: $56 billion in AI-cloud and third-party lease commitments, $108.5 billion in land and power guarantees, and more than $500 billion in memorandums of understanding for third-party capital.
The cash flow tell
Q2 revenue was $96 billion. Free cash flow was $21.3 billion, down from $48.5 billion the prior quarter. A more-than-halving of free cash flow in a quarter where revenue grew is the number worth sitting with. It reflects prepayments and inventory commitments made to lock supply -- rational behavior for a company that has spent three years supply-constrained, and also the mechanism by which a supplier converts flexibility into obligation.
What the analyst community is now saying
"Investors increasingly need to consider customer credit quality, leases, guarantees, revenue-sharing agreements and Nvidia's equity investments -- not just GPU shipments," said Charu Chanana, chief investment strategist at Saxo. That is a meaningful shift in how a semiconductor company is analyzed. Nvidia has historically been modeled as a product business: units, average selling price, gross margin. Chanana is describing something closer to a structured-credit analysis.
The context
Nvidia is not alone in this. A Wall Street Journal analysis found nine top tech companies carrying roughly $3 trillion of off-balance-sheet commitments tied to AI, against about $600 billion of reported capex over the same period. What separates Nvidia is direction: the hyperscalers are committing to buy compute, while Nvidia is committing to build and finance the capacity that produces it, and simultaneously taking equity in the customers who consume it. Marvell's cautious guidance triggered a sector selloff a day after Nvidia's beat, which is a reminder that sentiment in this group can reverse inside 24 hours.
The bear and bull cases, honestly
Bull: these commitments are how you win a supply-constrained market, TSMC capacity is genuinely scarce, and locking it is worth the cash-flow hit. Nvidia's 2028 revenue projection implies 70% growth, and if that lands, $279 billion of supply commitments looks conservative. Bear: commitments are contractual and demand is not. If AI capex growth decelerates even modestly, Nvidia owns the take-or-pay obligations while its customers own the option to slow down.
The line to watch next quarter is not revenue. It is whether that $279 billion grows again and whether free cash flow recovers toward the $48 billion mark.
Where this shows up next
The practical consequence for the rest of the market is that Nvidia's commitments now function as a demand forecast that everyone else plans against. TSMC schedules capacity around it, memory suppliers like SK Hynix -- which is building its first US facility in Indiana, targeted as a key memory production base by 2030 -- size fabs against it, and power developers underwrite projects on it. That makes the $279 billion figure load-bearing for a supply chain far wider than Nvidia's own income statement. It also means a downward revision would propagate through a dozen public companies before it showed up in Nvidia's revenue, which is roughly what the Marvell reaction this week previewed.