Nvidia just delivered one of the most profitable quarters in corporate history, and investors sold the stock anyway.
Revenue hit $96.2B in fiscal Q2 2027, up 106% from a year ago, while adjusted earnings of $2.22 per share beat expectations. The quarter was powered by another surge in data center demand.
The hyperscaler risk is growing
Nvidia's hyperscaler dependence is the single biggest structural question hanging over the stock. Amazon, Alphabet, Microsoft, Meta, and SpaceX account for much of its GPU demand.
In Q1, hyperscaler revenue reached $37.9B, while AI Clouds, Industrial and Enterprise brought in $37.5B. The newer segment grew 31% quarter-over-quarter compared with 12% for hyperscalers, pointing to some diversification, though the dependence remains significant.
The hyperscalers themselves are starting to buckle under their own spending. Amazon and Alphabet both turned cash flow negative in Q2.
Meta's cash generation shrank by more than 90% from a year earlier, and both of Elon Musk's public companies reported negative free cash flow. When your five biggest customers are collectively burning cash to buy your product, the growth rate has a ceiling.
Nvidia knows this. The company recently announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to mobilize over $500B in third-party capital for AI infrastructure.
The idea is to make GPUs function like real estate: an investable asset class that smaller companies can finance at lower rates. Few details have been released beyond a memorandum of understanding.
Margins are moving the wrong way
Nvidia’s gross margin held at 75% in Q2, but Q3 guidance came in below Wall Street’s expectations. That matters because Nvidia has already raised server prices to offset higher memory costs, yet margins are still expected to tighten as cost pressure builds.
Supply commitments make this more concrete. Nvidia's future supply commitments more than doubled in a single quarter, jumping to $279B from $119B.
The company has $92B committed for the rest of this fiscal year, followed by $87B in FY28 and $88B in FY29. These are locked-in obligations to memory suppliers rather than costs already incurred.
The commitments show confidence in future demand, but they also leave Nvidia more exposed if orders slow.
Nvidia's lead gets tested
The Q3 revenue guide of $108B, plus or minus 2%, came in above the Street's consensus. Jensen Huang's $1T revenue target from Blackwell and Vera Rubin through 2027 remains the bull case anchor.
Vera Rubin is now in full production, while China sales contributed less than 1% of data center revenue in Q2 and none are assumed in the Q3 outlook, removing one clear downside risk.
The bigger risk to Nvidia may come from customers learning to do more with fewer chips. AI firms currently use only a fraction of their GPUs’ theoretical capacity, while DeepSeek showed how better software can sharply reduce hardware needs.
Nvidia has spent $6B on licenses from Poolside AI as it pushes further into that part of the stack. If efficiency keeps improving, the pressure could eventually show up in hardware demand and margins.
Nvidia still has a formidable lead, but expectations are doing a lot of the heavy lifting. Its AI Clouds, Industrial and Enterprise (ACIE) business needs to keep growing, margins have to recover, and hyperscaler spending can’t lose momentum.