Jenson Huang’s massive AI prediction has a small problem

Quick read

• NVDA’s data center revenue rose 92% to $75B, but it has $119B in supply obligations to enterprise buyers who have yet to prove AI ROI.

• Uber can’t link token spending to product gains, Microsoft canceled AI licenses for budget overruns and Meta shut down its internal AI application leaderboard.

• Don’t wait: The analyst who called Nvidia in 2010 revealed his top 10 AI stocks. View the full list now for free.

Jensen Huang has big announcements. In NVIDIA’s most recent earnings call, the CEO told investors that “the creation of AI factories, the largest infrastructure expansion in human history, is accelerating at an extraordinary pace.” The numbers behind that statement are real: Nvidia (NASDAQ:NVDA) posted quarterly revenue of $81.61 billion, with the data center alone contributing $75.25 billion, up 92% year over year.

Small problem? Customers who pay those factories are starting to ask tough questions about what they’re getting back.

The supply side seems bulletproof

Huang’s supply-chain framing is optimistic. Speaking alongside the SK Group chairman last weekend, he flagged chronic chip shortages and extensive deepening cooperation with Korean memory suppliers. NVIDIA’s own balance sheet supports the boom: Total distribution-related liabilities rose to $119.0 billion, up from $95.2 billion a quarter ago.

Demand-side commitments are equally ambitious. OpenAI has committed to at least 10 gigawatts of NVIDIA systems, Meta signed a multi-year deal covering millions of Blackwell and Rubin GPUs, and CoreWeave is targeting 5-gigawatt AI factories by 2030.

Small problem: ROI math is getting worse

Three recent disclosures suggest that institutional buyers are struggling to translate token spending into measurable returns.

First, Uber COO Andrew Macdonald said at a May 25 conference that a clear line cannot be drawn between rising AI token spending and scalable consumer product improvements. Second, Microsoft revoked internal cloud code licenses in its Experiences and Devices division effective June 30, citing budget overruns from the token-based bill. Third, Meta shut down Cladeonomics, its internal AI token leaderboard, in April after it found employees engaged in gaming.

Don’t wait: In 2010, an analyst called NVIDIA revealed his top 10 AI stocks. View the full list now for free.

The discipline behind AI demand is getting tougher, and overall spending continues to grow. Reddit sentiment on NVDA has turned mixed with a widely circulated post suggesting that H200 GPU rental prices dropped 38% in the second half of May, an early signal that compute capacity may be outpacing monetized workloads.

Where the tension resolves

NVIDIA shares closed at $205.10 on June 5, down 8% since the May 20 earnings release, and the stock is up 47% over the past year. Polymarket traders have a 59% probability NVDA pays around $200 for June.

Huang’s prediction may still be grossly correct. Token volumes continue to climb, hyperscaler capex continues to expand, and Q2 revenue guidance of $91.0 billion indicates continued acceleration. The question is whether corporate CFOs will write checks at the pace necessary to justify $119 billion in supply commitments while their own ROI dashboards are still reading. Endless.

Look at the next two revenue cycles. If the hyperscalar capex commentary holds firm and token growth continues, the supply-demand handshake remains intact. If corporate pilots stall, the biggest infrastructure expansion in human history faces the oldest question in business: What’s the return?

Don’t wait: In 2010, an analyst called NVIDIA revealed his top 10 AI stocks. View the full list now for free.