Nvidia has beaten Wall Street’s earnings estimate in each of the past six quarters, by an average of roughly 5.6%. Its shares have declined the following session all six times. That record, more than any figure on the income statement, frames what is at stake when the company reports fiscal second-quarter results.
The published consensus calls for revenue of $91.7 billion and adjusted earnings of $2.07 a share. Almost no one on the buy side is underwriting those numbers. Whisper expectations cluster between $93 billion and $96 billion, roughly 4% above the printed figure, with data-center revenue estimates having widened from $80 billion to as high as $96 billion in a matter of weeks — an unusually large dispersion for a company this heavily covered.
The gap exists because Nvidia’s management has guided conservatively for four straight quarters, delivering revenue above the top end of its own range each time. In the fiscal first quarter, the company guided to $78 billion give or take 2% and printed $81.61 billion. Investors have learned to treat the midpoint as a floor rather than a forecast.
What they have not learned to trust is the reaction. The pattern in the tape suggests the market is now trading the guide, not the quarter — and that a beat unaccompanied by a materially higher outlook has become a selling event by default. Breaking the streak likely requires revenue above $93 billion paired with a third-quarter forecast north of $105 billion. An in-line result with a roughly $98 billion guide would, on recent form, be met with a decline of 2% to 4%.
Underneath that sits an unusual valuation setup. The shares have fallen about 15% from their May highs on rate concerns and regulatory scrutiny, even as analysts revised earnings estimates up 6% over the past 30 days. Roughly 34 to 35 analysts raised estimates in that window against two or three cuts. Some 95% maintain buy ratings, and the average price target sits near $309. The stock trades around $201 — roughly 24% below even the lowest target on the Street.
That decoupling is the crux of the print. Either the market is pricing a risk the models are not, or the models have not caught up to the price.
The risk in question is China. Nvidia faces Senate scrutiny over chips reaching Chinese buyers through third-party diversion, an issue consensus estimates do not currently reflect. Management’s ability to rebut the allegations cleanly — and to quantify how much revenue is genuinely exposed — may matter more to the multiple than the revenue line itself.
Three disclosures carry outsized weight. The first is the Vera CPU, a new segment where confirmation of a $20 billion revenue opportunity this year would shift the narrative from accelerator vendor to full platform and force upward revisions to fiscal 2028 estimates. The second is qualification status for HBM4 memory, where a three-supplier qualification would remove a bottleneck ahead of the Rubin architecture ramp; friction there would threaten the 2026 schedule. The third is sovereign AI revenue, running at about $9 billion on a trailing basis, which serves as the clearest available buffer against concentration in both hyperscaler capital spending and China.
Networking is the quieter tell. At roughly $18 billion consensus against a $20 billion-plus buy-side bar, the segment functions as evidence of ecosystem lock-in against merchant silicon and Ethernet-based competition — the structural argument for the multiple rather than the cyclical one.
Gross margin has held near 75% through successive architecture transitions, and rising memory costs are the main threat to that. Should it slip alongside a revenue figure below $89 billion, the downside scenario is not a few percentage points but a re-rating.
