NVIDIA posted FY2027 Q2 results on Wednesday, Aug. 26, delivering what the report described as one of the strongest quarters ever seen in the semiconductor industry. For the quarter ended July 26, 2026, revenue reached $96.22 billion, up 106% from a year earlier and 18% from $81.61 billion in Q1. That also came in 4.5% above the analyst consensus of about $92.1 billion.
Data center revenue was $89.02 billion, up 117% year over year and 18% quarter over quarter. The segment now makes up about 93% of NVIDIA’s total revenue. By comparison, gaming, automotive, professional visualization and edge computing together contributed $7.2 billion.
Adjusted EPS was $2.22, ahead of the $2.09 consensus by about 6.2%. On a GAAP basis, EPS was $2.46, including $7.8 billion in gains from equity investments, with holdings that include Intel and SpaceX. GAAP net income reached $59.7 billion, up 126% from a year earlier. Gross margin, on both a GAAP and non-GAAP basis, held at 75%, flat from the previous quarter and 250 basis points above 72.5% a year ago.
NVIDIA guided Q3 revenue to $108 billion, plus or minus 2%, versus prior analyst expectations of roughly $104 billion. After the release, the stock rose about 4% in after-hours trading to around $218.
Founder and CEO Jensen Huang said in the company’s news release: “AI has reached the tipping point. It is doing valuable work. Every token is productive and generating revenue. Now, compute is revenue.”
The quarter in plain numbers
NVIDIA’s fiscal calendar runs ahead of the standard calendar year, so the company’s FY2027 Q2 report released in August 2026 covers the three months ended July 26, 2026. Financial media may call it NVIDIA’s August 2026 earnings report, while the company labels it FY2027 Q2. They refer to the same set of results.
At $96.22 billion, quarterly revenue was larger than NVIDIA’s full-year revenue for fiscal 2023, according to the source text. The company also returned $26 billion to shareholders during the quarter through buybacks and dividends, had $99 billion left under its repurchase authorization, and issued $25 billion of senior unsecured notes. Free cash flow for the quarter was listed at $21.1 billion.
Where the growth came from
The report ties NVIDIA’s 2026 growth story to one dominant force: major technology companies are racing to build AI infrastructure, and NVIDIA supplies the chips at the center of that build-out.
Alphabet, Amazon, Meta and Microsoft spent a combined $166 billion in capital expenditures in the quarter ended June 2026, up 87% year over year and 27% from the prior quarter. Over the 10 quarters since the start of 2024, their combined capex has increased 272%.
Chief Financial Officer Colette Kress said capital spending by the top five hyperscalers is expected to rise from about $800 billion in 2026 to about $1.3 trillion in 2027. A large share of that spending is flowing into NVIDIA’s data center business.
Within data center, hyperscalers contributed $48.7 billion in revenue, up 102% from a year earlier. Growth outside that group was even faster. AI cloud, industrial and enterprise customers, or ACIE, contributed $40.3 billion, up 138% year over year and 25% quarter over quarter.
The customer base is also broadening in concrete ways, according to the report. Amazon Web Services plans to deploy an additional 2 million NVIDIA GPUs. SpaceX said it will build its AI infrastructure entirely on NVIDIA. Government-led sovereign AI projects have also become a revenue source the company can no longer ignore.
$279 billion in supply commitments became a key signal
One of the most closely watched figures in the report was not a revenue line. Supply commitments surged from $119 billion at the end of Q1 to $279 billion at the end of Q2, more than doubling in a single quarter. The increase was tied mainly to memory procurement for the Vera Rubin platform.
That matters because it locks in a large amount of future revenue that has not yet shipped. In the report’s framing, this is the clearest evidence that the demand NVIDIA sees is not only a forecast, but an order book backed by commitments.
The next-generation Vera Rubin platform began volume shipments on schedule in Q3 and is expected to account for about 20% of Q3 data center revenue. Kress guided to revenue growth of about 70% in fiscal 2028 and said that pace reflects supply constraints. Huang put it this way: “Even though demand is far beyond 70%, our supply gives us the confidence to deliver 70%.”
Why gross margin, not revenue, became the main debate
Once investors had the full picture — revenue above expectations, EPS above expectations, a $108 billion Q3 guide and $279 billion in supply commitments — the stock moved higher. But the argument after hours centered on margins, not the top line.
NVIDIA guided Q3 gross margin to 74%, down 1 percentage point from the 75% it just delivered in Q2. At a revenue scale of $108 billion, a 1-point drop in gross margin means more than $1 billion less in quarterly gross profit. Kress then guided Q4 gross margin lower still, to 71% to 72%.
Management tied that pressure directly to memory costs. The company said memory cost increases had already exceeded prior expectations and are expected to be “even higher” next year. Most of the $279 billion in supply commitments are linked to memory procurement, and the price of high-bandwidth memory, or HBM, embedded in each NVIDIA GPU is still rising.
Earlier this month, NVIDIA had already told customers it would raise AI server prices by more than 15% to offset the jump in memory costs. But the source says cost inflation has outpaced what the company has been able to pass through, putting gross margin under pressure at scale.
The report links this directly to SK hynix. SK hynix posted a 72% operating margin in Q1 2026, in part because NVIDIA and its customers had little choice but to keep buying HBM at higher prices. What looks like a cost headwind for NVIDIA has become a revenue tailwind for memory suppliers.
The math is simple, and that is exactly why investors care. If NVIDIA’s revenue rises from $96 billion to $108 billion while gross margin slips from 75% to 74%, gross profit still increases, from about $72 billion to about $79.9 billion. But investors who modeled $108 billion of revenue at a 75% gross margin were expecting about $81 billion in gross profit. A gap of more than $1 billion per quarter adds up quickly over several quarters.
The report also explains gross margin in practical terms. In Q2, NVIDIA kept $0.75 from every $1 of revenue after paying direct costs for components, manufacturing and assembly. If that figure falls to $0.74 in Q3 and to $0.71 to $0.72 in Q4, a few cents per dollar turn into tens of billions at NVIDIA’s size. The source notes that a 75% gross margin is rare for a hardware company: Apple is at 49%, while most chipmakers are in the 50% range.
What the numbers say about the business structure
Step back from a single quarter, and several structural points become clear. NVIDIA is now heavily concentrated in AI infrastructure. Data center makes up about 93% of total revenue, which means the company’s earnings engine is, in practical terms, tied to a single dominant spending theme.
Customer concentration is still real, but the mix is improving. Hyperscalers remain the biggest buyers, contributing $48.7 billion. ACIE, however, contributed $40.3 billion and grew faster, a sign that the customer base is becoming more diversified.
The report also makes a direct point about what is limiting growth: supply, not demand. Huang said demand without supply constraints is “far beyond” the roughly 70% growth guide. If orders are already ahead of capacity, any disruption in the supply chain would hit revenue recognition quickly.
That is why the jump in supply commitments from $119 billion to $279 billion stands out. It is not simply a management talking point. It is money committed to real suppliers, mainly around memory.
Shareholder returns also stayed large. NVIDIA returned $26 billion during the quarter through buybacks and dividends.
“Compute is revenue” was the line investors noticed
The report singles out Huang’s phrase “compute is revenue” as a new and deliberate formulation. Its meaning, in context, is that companies are no longer spending on AI infrastructure only because they expect it to matter one day; the infrastructure is already producing measurable economic returns.
If that assessment is correct, it becomes one of the strongest structural arguments for sustained demand. If it proves early, and if the AI infrastructure built between 2024 and 2026 takes longer than expected to generate economic returns, then the capital spending cycle behind NVIDIA’s extraordinary growth could face a turning point later on.
How this quarter connects to broader market themes
The FY2027 Q2 results tie back to several major themes discussed across the market. The $166 billion in quarterly hyperscaler capex is the same force behind strength in AI-linked stocks, the memory supercycle, SK hynix’s record profitability and the HBM price gains now pushing down NVIDIA’s gross margin.
The source also points to China. Export controls have reduced China data center compute revenue to less than 1% of NVIDIA’s data center revenue, while creating room for CXMT to serve domestic AI infrastructure demand.
The $279 billion in memory-related supply commitments also connects directly to the pricing power of SK hynix, Samsung and Micron. At this scale, NVIDIA is one of the main demand drivers behind HBM pricing.
Margin trends are also discussed through a valuation lens. In an environment where risk-free rates are in the 3.5% to 3.75% range, investors want high and stable margins to justify premium valuations. When margins contract, even from 75% to 74%, the math behind those valuations changes.
What to watch next
- The Q3 earnings report in late November 2026. The most important metric may not be revenue, but whether gross margin stabilizes at 74% or continues toward 71% to 72%.
- The Vera Rubin production ramp. The platform is expected to represent about 20% of Q3 data center revenue. A faster ramp from Q4 into fiscal 2028 could offset margin pressure, while delays would worsen it.
- Memory costs. HBM pricing from SK hynix, Samsung and Micron feeds directly into NVIDIA’s product costs. Any softening would help gross margin and would also affect memory companies now benefiting from elevated pricing.
- China policy. Any change in U.S. AI chip export controls to China would have an immediate effect. The current guide assumes zero contribution from that market.
- Evidence of AI monetization. After Huang’s “compute is revenue” comment, investors will be watching for measurable proof that AI infrastructure is producing economic returns.
- The forecast for $1.3 trillion in 2027 capex from the top five hyperscalers. Kress said spending is expected to rise from about $800 billion in 2026 to about $1.3 trillion in 2027. If that holds, NVIDIA’s supply-constrained growth path could extend into fiscal 2028 and beyond.
A very strong quarter, with the market focused elsewhere
By most objective measures, this was a powerful earnings report: revenue more than doubled, data center revenue rose 117%, the Q3 guide was about $4 billion above consensus, supply commitments jumped from $119 billion to $279 billion, and the stock gained about 4% after hours.
Yet the market’s real focus has shifted to margin trajectory. Gross margin moved from 75% in Q2 to a 74% Q3 guide and a projected 71% to 72% in Q4, reflecting management’s view that memory costs will keep rising.
The data in the source is current as of Aug. 27, 2026. Listed materials include NVIDIA’s official news release, SEC Form 8-K dated Aug. 26, 2026, GlobeNewswire, AlphaStreet, CNBC, 24/7 Wall St., Tom’s Hardware, BigGo Finance, FinanceFeeds, Benzinga, Kiplinger, RexShares, Robinhood market data, CNN Markets, NextPlatform and Shacknews.