Nvidia forecasts revenue above estimates, announces $80 billion share buyback

Nvidia forecast Q2 revenue of $91 billion, plus or minus 2%, above the $86.84 billion consensus, and unveiled an $80 billion share buyback. First-quarter revenue came in at $81.62 billion versus $78.86 billion expected, with data center revenue of $75.2 billion and adjusted EPS of $1.87 versus $1.76 estimated. The company also raised its quarterly dividend to 25 cents from 1 cent, though shares were down 0.2% in extended trading.

Analysis

The key second-order signal is not the beat itself, but that management is effectively trying to re-price the stock from “cyclical AI capex beneficiary” to “durable cash compounding machine.” The buyback and dividend step-up matter because they reduce the market’s ability to dismiss NVDA as a pure reinvestment story; if free cash flow remains this strong, the equity now has a structural bid that can dampen drawdowns and pull forward multiple support over the next 2-3 quarters.

The bigger competitive read-through is that inference is becoming the battleground where Nvidia’s moat is least certain and where hyperscalers can actually scale custom silicon. That shifts the risk from near-term GPU shortage upside to medium-term margin erosion if cloud customers successfully substitute in lower-cost, workload-specific chips; the earliest evidence will be in order-book quality and mix, not headline revenue. If the company’s new CPU/system push gains traction, it can partially offset that by expanding wallet share, but it also broadens the attack surface against specialized inference vendors.

For the broader AI complex, this is modestly bullish for the supply chain that feeds NVDA’s ramps, especially advanced packaging, HBM, and networking, but it is a selective positive rather than a blanket risk-on signal. The clearest losers are the names with the most exposed inference substitution narrative: custom accelerator efforts at hyperscalers and merchant x86/AI compute vendors that rely on Nvidia’s pricing discipline staying intact. The market may be underestimating how quickly capex growth can remain high while returns on that capex compress, which is why the stock reaction can stay muted even on an apparent beat.

Contrarian view: consensus is treating this as confirmation that the AI spend cycle is intact, but the more important question is whether spend is merely being reallocated from training to inference and from merchant silicon to in-house designs. If that’s right, NVDA can keep growing while the industry’s profit pool narrows, which is classic late-cycle behavior. The main catalyst that could reverse the current enthusiasm is a few quarters of decelerating backlog visibility or explicit commentary that custom chips are taking share faster than feared.

Key Decisions for Investors

• Maintain a tactical long NVDA position into the next 4-8 weeks, but size it as a cash-return compounder rather than a momentum trade; upside remains supported by capital return and visibility, while downside is cushioned by buybacks.
• Express the more durable AI theme via a pair trade: long NVDA / short a basket of inference-adjacent challengers with the weakest moat and highest valuation sensitivity over 3-6 months; the risk/reward favors NVDA if substitution takes longer than expected.
• For a lower-beta way to play continued AI infrastructure spend, overweight suppliers tied to memory, packaging, and networking over merchant compute names; these should benefit even if NVDA’s share of wallet compresses.
• Consider buying downside protection on NVDA only if implied volatility remains muted post-earnings; the cleanest hedge is a 2-3 month put spread, targeting a scenario where the market starts pricing inference share loss rather than growth disappointment.
• Avoid chasing long exposure to custom silicon narratives in GOOGL and AMZN until there is evidence of meaningful deployment scale; the near-term trade may be headline-positive, but the 6-12 month economics remain unproven.