NVIDIA’s latest earnings blew past expectations, and for a moment the market reacted as if the AI wave could only move in one direction. But once investors dug into the filing, the shine wore off quickly. The deeper numbers reveal a widening gap between the story being sold and the fundamentals underneath.
The company’s accounts receivable soared to 33.4 billion dollars, with days sales outstanding climbing to 53 days—far above NVIDIA’s historical average and noticeably higher than peers like AMD, Intel, Micron and TSMC. When demand is genuinely overwhelming, customers typically accelerate payments. Here, the opposite is happening.
Meanwhile, NVIDIA’s inventory ballooned 32 percent quarter-over-quarter. If supply is tight and everything is supposedly “spoken for,” inventory should shrink—not expand. Add to that the steady slide in GPU spot pricing across global compute marketplaces, and the narrative of hardware “shortages” becomes harder to defend.
Cash conversion slipped to 75 percent, well below an industry norm of roughly 100 percent. Even so, the company pushed nearly 10 billion dollars into share buybacks—an aggressive move that raises questions about how much of the reported profit is translating into actual, spendable cash.
But the structural risk goes beyond one company’s balance sheet. Much of today’s AI boom is powered not by organic customer demand, but by a circular flow of capital: NVIDIA backs AI startups → startups spend heavily on cloud → cloud providers purchase NVIDIA hardware → NVIDIA books the revenue → investors reward the cycle. It’s a loop where money rotates, but returns remain elusive.
That’s why major players—Thiel, SoftBank, Burry—have quietly stepped back. They’re not anti-AI. They’re skeptical of an economic engine running on promises instead of payback.
As global liquidity tightens, particularly in places like Japan, the bigger question emerges: what happens when AI spending cools and the invoices come due?
For data-center builders, policymakers, and energy developers, the message is clear: sustainable AI requires sustainable infrastructure—and that includes reliable, long-term power like geothermal.
