Research · Earnings preview · November 3, 2026

NVIDIA Q3 FY2027 Preview: Checkpoints Before November 17

NVIDIA reports third-quarter fiscal 2027 results on Tuesday, November 17, 2026, with results announced at approximately 1:20 p.m. PT and a conference call at 2 p.m. PT (5 p.m. ET). This preview sets six checkpoints from the second-quarter record — the guidance bar, Data Center concentration, the margin step-down, the supply commitment schedule, the Vera Rubin ramp, and what this quarter cannot settle — so the release can be read against a fixed baseline instead of a headline. BlitzLab's score for NVIDIA is unchanged by this preview: 87 out of 100, grade A, as of 2026Q4. See the NVIDIA company page for the full scorecard.

When, and what this preview is

NVIDIA announced that it will host a conference call on Tuesday, November 17, at 2 p.m. PT to discuss financial results for the third quarter of fiscal year 2027, which ended October 25, 2026. Results will be publicly announced at approximately 1:20 p.m. PT, and written commentary from the chief financial officer will be posted to the investor relations website immediately after that announcement, ahead of the call. The call will be webcast live in listen-only mode. That announcement is the only forward-looking date used here; everything else in this preview is a second-quarter fact, a stated guidance figure, or arithmetic drawn from them.

This preview makes no price forecast, no earnings estimate, and no recommendation of any kind. It uses only company releases and SEC filings as sources. Where a figure below comes from the second-quarter release, it is labeled as company-reported; where it comes from the Form 10-Q, the filing is named. Guidance figures are the company's own outlook as stated with the second-quarter results, not BlitzLab's expectations.

The baseline: Q2 FY2027 in one table

The second quarter, ended July 26, 2026, is the quarter that defines this preview: revenue more than doubled year over year, Data Center supplied more than nine-tenths of it, and margins held at 75.0% on both a GAAP and a non-GAAP basis. The third-quarter guidance that came with it is a step up in revenue and a step down in margin — both directions matter, and neither alone describes the quarter to come.

Q2 FY2027 measure (quarter ended July 26, 2026)FigureChange
Total revenue$96,221 million+106% year over year; +18% quarter over quarter
Data Center revenue$89.0 billion+117% year over year; +18% quarter over quarter; 92.5% of total revenue
Edge Computing revenue$7.2 billion+27% year over year; +13% quarter over quarter
GAAP gross margin75.0%74.9% in Q1 FY2027; 72.4% a year earlier
Non-GAAP gross margin75.0%75.0% in Q1 FY2027; 72.5% a year earlier
GAAP operating income$63,734 million+124% year over year
GAAP net income / diluted earnings per share$59,688 million / $2.46Non-GAAP diluted earnings per share $2.22
Capital returned in the quarterApproximately $26.0 billion (repurchases and dividends)Approximately $99.0 billion remaining under the repurchase authorization at quarter-end
Q3 FY2027 guidance (stated with Q2 results)Revenue $108.0 billion, plus or minus 2%; GAAP and non-GAAP gross margin 74.0%, plus or minus 50 basis pointsGuidance assumes no Data Center compute revenue from China; GAAP operating expenses approximately $9.2 billion, non-GAAP approximately $9.0 billion
Supply and capacity commitments (Form 10-Q, Note 10)$279 billion as of July 26, 2026$119 billion as of April 26, 2026; schedule: $92 billion in the remainder of fiscal 2027, $87 billion in fiscal 2028, $88 billion in fiscal 2029, $12 billion thereafter

Two Q2 details matter for how Q3 should be read. First, the guidance bar is arithmetic, not a forecast: $108.0 billion plus or minus 2% spans $105.84 billion to $110.16 billion, and the midpoint is about 12% above Q2 revenue of $96.2 billion. Second, the margin guidance is a step down, not a collapse: 74.0% against a Q2 print of 75.0% on both bases. A result inside both ranges would be a quarter that grew and gave back one point of margin at the same time — the two movements should be reported together.

Checkpoint 1: the $108 billion bar, and the China assumption inside it

The single most quoted number in the release will be total revenue against the $108.0 billion guidance midpoint, with the plus-or-minus-2% range of $105.84 billion to $110.16 billion as the band. Read the band, not just the midpoint: a result above the midpoint but below the top of the range is still inside guidance, and a result below the midpoint but above the bottom of the range is too. The guidance carries an explicit assumption that deserves equal billing: the company stated it is not assuming any Data Center compute revenue from China in its outlook. Q3 will therefore show whether that zero assumption still frames the quarter, or whether the company changes what it assumes — a disclosure change that matters more for the scorecard's China item than any single revenue print.

Checkpoint 2: Data Center concentration, and the smaller line

Data Center revenue was $89.0 billion in Q2, up 117% year over year and 92.5% of total revenue; Edge Computing was $7.2 billion, up 27% year over year. For Q3, check both lines and the share between them. If Data Center grows but its share rises further, concentration deepens even in a strong quarter; if Edge grows faster than Data Center, the mix broadens without the headline needing to accelerate. Either pattern is informative, and neither is visible from the total alone. The scorecard's Moat case rests on the platform, not on one quarter's mix — but the mix is where a platform story first shows strain, because a narrowing base has fewer places to hide a slowdown.

Checkpoint 3: margin against 74.0%, from a 75.0% base

GAAP and non-GAAP gross margins were both 75.0% in Q2, against 74.9% GAAP in Q1 and 72.4% GAAP a year earlier. Guidance for Q3 is 74.0% on both bases, plus or minus 50 basis points — a range of 73.5% to 74.5%. A Q3 margin inside that range would confirm the guided step-down and nothing more; a margin above the range would mean the step-down did not arrive on schedule; a margin below it would mean it arrived early or larger than guided. Watch the GAAP and non-GAAP prints separately: in Q2 they coincided at 75.0%, so any gap between them in Q3 is itself new information about what sits between the two measures. Margin, not revenue alone, is what converts scale into the economics the scorecard credits — a revenue beat with a margin miss and a revenue miss with a margin hold are different quarters.

Checkpoint 4: the $279 billion commitment schedule

The Form 10-Q for the quarter ended July 26, 2026, reported supply and capacity commitments of $279 billion, up from $119 billion as of April 26, 2026, described in the filing as primarily for data center infrastructure systems, principally memory and manufacturing facilities. The schedule is heavily front-loaded: $92 billion falls in the remainder of fiscal 2027, $87 billion in fiscal 2028, and $88 billion in fiscal 2029, with only $12 billion after that. The filing also states that, in certain instances, these agreements may be cancelable, rescheduled, or adjustable before firm orders are placed, and that changes may result in additional costs — so the $279 billion is a commitment schedule, not cash already spent and not an unconditional liability.

For Q3, the question is the direction of that schedule in the next 10-Q: continued growth means the company is still buying ahead of demand; a stall or decline would be the earliest filing-level signal that its own demand conviction has softened, before revenue itself turns. Do not add the commitment schedule to separately disclosed guarantees to manufacture a single total — the filing presents them in separate tables, and this preview does not combine them. What matters is the trend of the schedule itself, and whether the front-loaded shape ($92 billion, then $87 billion, then $88 billion) holds or shifts.

Checkpoint 5: the Vera Rubin ramp, from “in full production” to delivered revenue

With the second-quarter results, the company announced that the Vera Rubin platform is ramping into full production, with racks running at partners including CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure, and Nebius. Q3 is therefore the first quarter in which a production ramp can be read as delivered product rather than as an announcement: check whether Data Center growth is described as Rubin-led, Blackwell-continuing, or mixed, and whether the platform split inside Data Center moves. A ramp that shows up only in commentary and not in the revenue mix would be a thinner result than the announcement suggested; a ramp visible in the mix would confirm that the architecture transition is shipping, not just starting. This checkpoint is about evidence of delivery, not about which architecture deserves credit — the scorecard credits platform plus speed, and speed is measured in shipments.

Checkpoint 6: what Q3 cannot settle

Several score items cannot move on a single quarterly release, and Q3 should not be graded as if they could. The China export-control position is already framed as a zero assumption in guidance; one quarter does not resolve a structural market question. Regulatory probes, where no violation determination has been reported, are not settled by a revenue print. The structural competition from hyperscaler in-house silicon — customers building their own accelerators — plays out over product generations, not quarters, and its test is developer and platform adoption, not one quarter's share. The commitment schedule in Checkpoint 4 is a multi-year position; its character changes when its trend changes, not when a single quarter passes. Treating any of these as a Q3 pass-or-fail test would over-read the quarter.

Score context: 87 out of 100, and what would have to change

BlitzLab scored NVIDIA 87 out of 100 (grade A) as of 2026Q4: Speed 23 out of 25, Moat 52 out of 60, Creative Destruction 15 out of 15, and Demerits of minus 3, for 23 + 52 + 15 − 3 = 87. That is 22 points above the 65-point threshold, and the highest score among measured companies. The gates passed: G1 on sustained growth with second and third curves visible, and G2 as a borderline pass on estimate — the scorecard itself calls this the paradox of scale, where a gate once passed comfortably now passes narrowly because the revenue base is so large.

Q3 is the nearest observation point for that score, but a rescore is a separate step from a preview: it requires the released filing, not the preview's checkpoints. Nothing in this article changes the score, the methodology, or the grade. If Q3 lands inside the guided ranges, holds margin near 74.0%, and shows the Rubin ramp in delivered revenue, the Speed and Moat inputs will have fresh evidence at rescore time; if revenue misses the guided band, margin falls below it, or the commitment schedule stalls, the same inputs will have fresh evidence in the other direction. Either way, the evidence, not the preview, will decide.

What to watch on the call, in one list

1. Total revenue against the $108.0 billion guidance band ($105.84 billion to $110.16 billion), and whether the no-China-Data-Center-compute assumption is restated or changed.
2. Data Center revenue and its share of the total, against the Q2 base of $89.0 billion and 92.5%, with Edge Computing alongside it.
3. GAAP and non-GAAP gross margin separately, against the 74.0% guidance range (73.5% to 74.5%) and the Q2 base of 75.0% on both bases.
4. The supply and capacity commitment schedule in the next 10-Q: growth, stall, or decline from $279 billion, and any change in its front-loaded shape.
5. Evidence that the Vera Rubin ramp is delivered revenue, not only announced production — and what it displaces or extends.
6. Operating expenses against the guided approximately $9.2 billion GAAP and $9.0 billion non-GAAP, as the check on whether growth is being bought with cost.

Sources