NVIDIA Earnings Preview: Date, Estimates and What to Expect

NVIDIA reports second-quarter fiscal 2027 results on Wednesday, 26 August 2026, with the conference call at 2:00 p.m. Pacific / 5:00 p.m. Eastern (NVIDIA Newsroom, 29 July 2026). The quarter closed on 26 July 2026. One number decides the reaction: data-center revenue against the company's own $91.0bn total-revenue guide, given on 20 May 2026.

Everything else on the call — Rubin shipments, China licences, the new financing deals — matters mainly for how it changes the trajectory of that line.

What NVIDIA guided and what the street expects

NVIDIA set the bar itself at the Q1 FY2027 call. Sell-side aggregates sit fractionally above it, which is the normal pattern and is why an in-line print is usually read as a mild disappointment.

Metric	Company guide (20 May 2026)	Street (mid-Aug 2026)
Revenue	$91.0bn ±2%	~$91.8–91.9bn
Non-GAAP EPS	not guided	~$2.06–2.08
Non-GAAP gross margin	~75.0% ±50bp	in line with guide
Data-center revenue	not guided separately	no reliable single consensus

Consensus figures here are from published sell-side aggregations (Hudson Labs preview, August 2026; tickerleague estimate page, accessed 19 August 2026). We could not verify a data-center-specific consensus from LSEG, FactSet or Visible Alpha, so we are not quoting one — treat any precise data-center "estimate" you see repeated in coverage with scepticism.

The base the quarter is built on

Q1 FY2027, reported 20 May 2026, was the strongest quarter the company has ever printed:

• Revenue $81.6bn, up 85% year over year, against roughly $79.1bn expected.
• Data-center revenue $75.2bn, up 92% year over year — a record, and ~92% of the total.
• EPS of $1.87 versus about $1.77 expected.
• An additional $80bn buyback authorisation and a dividend lifted to $0.25 a share.

Growing 85% off a base that large is what makes the comparison arithmetic unforgiving. Hitting the $91.0bn guide requires roughly $9.4bn of sequential growth — more absolute dollars in one quarter than NVIDIA's entire data-center segment produced in most of fiscal 2024.

What has changed since the last print

Hyperscaler capex went up, not down. The four largest cloud buyers spent roughly $172bn on capital expenditure in Q2 2026, up about 29% sequentially and 77% year over year (Platformonomics capex scoreboard, 31 July 2026). Alphabet raised its 2026 capex plan at the end of July, and Microsoft — FY26 Q4 revenue of $90.0bn, up 18% — signalled capex rising again in the new fiscal year (CNBC, 28–29 July 2026). Combined 2026 cloud capex is being reported near $600bn (The Register, 4 August 2026).

Rubin is shipping. The Vera Rubin platform ramped into full production on 31 May 2026 (NVIDIA Newsroom) and systems were confirmed shipping to CoreWeave, Google Cloud, Azure, Oracle Cloud, Nebius and OpenAI by late July 2026. A clean transition matters because product changeovers are historically where NVIDIA's gross margin wobbles.

China is still a rounding error. A senior US Commerce official testified on 14 July 2026 that only a "very small quantity" of H200 chips had shipped to China and Hong Kong (CNBC). Licences exist; volume does not. Any guidance that assumes a China recovery is guidance built on a policy decision that has not been made.

NVIDIA has started financing its own demand. On 10 August 2026 it announced platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to mobilise more than $500bn of third-party capital for AI compute buyers. A week later, on 17 August, it committed up to $105bn of financing and compute support for an OpenAI data centre in Ohio, plus $1.5bn of equity into SB Energy (Bloomberg, CNBC, 17 August 2026). This is the single most important strategic change of the quarter, and the part of the call we will be listening to hardest.

Our scenario weights for the print

These are our own weights, not a consensus survey, and they describe the first-session reaction rather than a fair-value target. NVDA closed at $219.74 on 18 August 2026, inside a 52-week range of $164.07–$236.54, for a market capitalisation near $5.3–5.4tn.

Scenario	What it looks like	Weight	Plausible first-session move
Beat and raise	Revenue above ~$93bn, next-quarter guide comfortably above $100bn, margin held at ~75%	40%	+4% to +8%
Beat, in-line guide	Revenue ~$92bn, next-quarter guide roughly where the street already is	40%	-4% to +2%
Margin or China disappointment	Guide met but gross margin below ~74%, or supply/licence commentary turns cautious	15%	-6% to -10%
Genuine miss	Revenue below the $89.2bn low end of the guide	5%	-12% or worse

The 40% we assign to "beat, in-line guide, stock down" is deliberately high, because that is what actually keeps happening.

How the stock has actually traded after recent reports

Report	Date	Result	Stock reaction
Q2 FY2026	27 Aug 2025	Beat, data centre +56%	Closed ~3% lower next session
Q3 FY2026	19 Nov 2025	Beat	Up ~5% after hours, closed ~3% lower
Q4 FY2026	25 Feb 2026	Beat	Fell ~5% on capex-sustainability worries
Q1 FY2027	20 May 2026	Large beat	Muted; reported moves conflict across sources

Four consecutive beats; not one of them produced a durable positive session. That is the base rate a buyer of the print is fighting, and it is why we would not treat "will NVIDIA beat?" as the interesting question. It almost certainly will. Whether a beat is enough is a different question entirely — and it is the same dynamic we described in our NVIDIA stock forecast for 2026.

What we are listening for on the call

1. Data-center revenue and its sequential dollar growth. The percentage is flattering; the incremental dollars tell you whether the ramp is still accelerating.
2. Gross margin against the ~75% guide. A Rubin-driven dip below 74% would be read as a cost problem, not a demand problem — but it would still be sold.
3. How the financing deals are accounted for. Vendor-adjacent financing at $500bn scale invites the question of how much reported demand is being underwritten by the seller. Management needs a clear answer.
4. Any China assumption inside the next-quarter guide. If a number is embedded, it is a policy bet.
5. Supply commentary and lead times for Rubin, which set the ceiling on the next two quarters regardless of demand.

What invalidates the constructive view

Two things would change our reading, and neither is about a single quarter. First, a hyperscaler explicitly guiding capex down for 2027 — the opposite of what all four did in July. Second, sustained non-GAAP gross margin below 72%, which would say NVIDIA is buying share rather than being handed it. Absent those, a soft reaction to a good print is a re-rating of expectations, not of the business.

If you are positioning around the event rather than the company, remember the read-across is broader than one ticker: the whole AI-adjacent complex, including the quantum computing names that reported in July, trades off this print, and index-level volatility is where most of the risk shows up. Our indices signals hub tracks the S&P 500 and Nasdaq setups into the event.

This article is analysis, not investment advice. Earnings events carry elevated single-session risk and options-implied moves regularly exceed 6%. Nothing here is a recommendation to buy or sell any security. If you want to trade the reaction rather than read about it afterwards, you can open a trading account with our partner broker — capital at risk.