Nvidia China outlook rattles shares after record quarter

Nvidia delivered numbers that would be jaw-dropping for almost any company, yet the market reaction was chilly because the narrative has shifted from silicon to statecraft. The company reported fiscal second-quarter revenue of 46.7 billion dollars and guided to 54.0 billion dollars for the current quarter, but crucially it assumed no H20 shipments to China in that outlook. After the release, shares fell about 3.2 percent in extended trading, shaving roughly 110 billion dollars off a 4.4 trillion dollar valuation.

The data center engine still dominates. Nvidia’s data center revenue reached about 41.1 billion dollars in the quarter. Management said around half of that came from large cloud service providers, which hints at some budgeting caution at the margins from hyperscalers even as AI capex remains intense.

Guidance was solid on paper. Revenue of 54.0 billion dollars plus or minus 2 percent lands a touch above the LSEG consensus, with non-GAAP gross margin guided to 73.5 percent plus or minus 50 basis points. The company also authorized an additional 60 billion dollars for share repurchases, underscoring confidence in cash generation.

China remains the swing factor. Nvidia reiterated that it did not include any H20 shipments to China in its outlook, despite receiving some export licenses earlier in August. Management added that if geopolitics ease and orders materialize, H20 could add 2 to 5 billion dollars to third-quarter revenue. In the second quarter, one customer outside China bought about 650 million dollars of H20, which helped offset the China pause.

The Nvidia China outlook is now inseparable from Washington’s evolving export regime and Beijing’s procurement posture. Earlier this month, the White House opened the door to allow scaled versions of next-gen Blackwell chips into China, alongside an unprecedented proposal that Nvidia and AMD remit 15 percent of China chip revenue to the U.S. government. Jensen Huang has publicly signaled willingness to work within that construct, but mechanics and formal rules are not yet final. Until they are, Nvidia is prudently excluding China from forecasts.

That is why this quarter’s wobble feels less about demand and more about diplomacy. As one portfolio manager put it, Nvidia’s bottleneck is not silicon. It is the negotiation channel between two superpowers. Markets can accept manufacturing lead times and supply chain hiccups. They struggle to price binary policy switches that can unlock or foreclose billions of dollars in quarterly sales.

Management’s message was that structural demand remains enormous, with multiple growth vectors beyond U.S.–China trade dynamics. Colette Kress said “sovereign AI” programs, where governments build their own national AI infrastructure, are on track to deliver about 20 billion dollars of revenue this year. Nvidia also estimates that cloud and enterprise customers will spend roughly 600 billion dollars on AI in 2025 and that cumulative AI infrastructure investment could reach 3 to 4 trillion dollars by decade’s end. Those numbers support the idea that the current digestion by some hyperscalers is a pacing issue, not a thesis break.

This was the smallest post-print reaction since Nvidia’s AI-era re-rating began, which speaks to how much was already priced into the stock. Still, the cross-asset message is clear. When the company that anchors the AI buildout reminds investors that a key growth cylinder depends on fast-moving policy, risk premiums tick up. Peer moves reflected that cautious tone. AMD dipped in sympathy right after the release as traders rebalanced AI infrastructure exposure.

Buyback firepower softens the blow. The added 60 billion dollars of repurchase capacity gives Nvidia flexibility to lean against volatility and absorb issuance tied to employee compensation as it scales headcount for Blackwell and Rubin. Cash returns also frame the company’s view that free cash flow can sustain both investment and shareholder payouts through cycle turns that are driven more by regulation than by demand.

First, watch the rulemaking. Any formalization of the proposed 15 percent China revenue share and clarity on permissible chip configurations would give CFOs in China the certainty they need to place multi-quarter orders. Until then, Nvidia will likely keep China out of guidance and treat any licenses as optionality.

Second, monitor cloud capex pacing. The quarter confirmed that cloud providers still account for about half of data center revenue. If those budgets re-accelerate as the next wave of agentic and inference workloads scale, the recent hesitation will look transitory. If returns stay hard to quantify, spend could tilt toward efficiency upgrades and networking while compute ramps in waves.

Third, track sovereign AI wins. If the 20-billion-dollar annual run rate proves durable, that line can diversify the geographic and customer mix at a moment when concentration risk is front of mind for investors. That, combined with the Blackwell ramp and Rubin’s pipeline, supports a multi-year narrative even if the Nvidia China outlook stays constrained for several quarters.

Bottom line for investors: The thesis did not break. It just got a policy haircut. Nvidia’s print shows relentless end-demand for accelerated computing, a strong balance sheet, and more levers than any peer to convert AI capex into revenue. The near-term multiple will trade on headlines out of Washington and Beijing rather than wafers out of the fab. If those headlines turn constructive, the guidance math changes quickly. If they do not, the company still has a deep bench of global projects to lean on.