Nvidia’s $96 Billion Quarter Puts the Factory in Charge

Nvidia posted $96.2 billion in quarterly sales and said supply, not demand, will cap growth through January 2028. Data Center brought in $89.0 billion of that total, and the company guided to $108.0 billion, plus or minus 2%, for the current quarter. The about 70% rise it now expects in fiscal 2028 is the amount of product it thinks it can ship, while each new rack generation also sells more dollars per gigawatt, which is how a $1 trillion year remains on some desks.

Last week’s print, for the quarter ended July 26, more than doubled the year-ago total of $46.7 billion. It still left customers in line.

The Quarter That Doubled Sales and Still Rationed Chips

Nvidia’s second-quarter fiscal 2027 results show a company already running at a scale most industrial firms never see. Sales rose 106% from a year earlier and 18% from $81.6 billion in the prior quarter, a fourth straight quarter of faster growth at this size.

GAAP gross margin held at 75.0%. Operating income was $63.7 billion, up 124%. Net income was $59.7 billion, or $2.46 a diluted share, up 128% from $1.08.

Q2 FY27 GAAP RESULTS

Those profits arrived with a queue. Jensen Huang, Nvidia’s founder and CEO, said AI “has reached its inflection point” and that “compute is revenue,” with Vera Rubin now in full production for that buildout. The call behind the release was about what the plants could not yet do.

Hyperscale customers spent $48.7 billion, up 13% from the prior quarter and 102% from a year ago. The AI clouds, industrial, and enterprise book spent $40.3 billion, up 25% sequentially and 138% from a year ago. Edge Computing added $7.2 billion, up 27% year over year. Data Center plus Edge is the whole company now.

Why the 70 Percent Outlook Is a Supply Number

Colette Kress, chief financial officer, gave full-year color a year early, which Nvidia has not usually done. She told investors the about 70% rise for fiscal 2028 is a factory number, not a demand number, and that customer forecasts point to growth doubling next year.

We expect to grow revenue by approximately 70% in fiscal 2028. This is a supply-constrained outlook.

Colette Kress, chief financial officer, NVIDIA Q2 fiscal 2027 earnings call

Huang said the company has “never guided to a year in advance,” and that “even though our demand is much greater than 70%, our supply allows us to confidently deliver 70%.” If the plants were unconstrained, he said, potential growth would be much higher. The figure is what Nvidia is willing to promise so wafer, memory, and building partners can plan against a single supply-constrained fiscal 2028 outlook.

The current quarter still assumes no Data Center compute revenue from China. Sequential growth is expected to come mainly from the non-hyperscale book, with large-cloud demand set to pick up again in the fourth quarter as Vera Rubin supply grows. After a blowout against estimates already near $92 billion, the remaining fight on the tape is whether 70% is already owed, not whether anyone still wants the boxes.

Nvidia Locked $279 Billion of Memory and Capacity

The factory problem is already on the balance sheet. Supply and capacity commitments rose from $119 billion last quarter to $279 billion of supply commitments as of July 26, a $160 billion jump in 13 weeks, primarily for memory and the plants that turn it into systems.

SUPPLY AND CAPACITY BY FISCAL YEAR

Cloud service agreements of $29 billion, data center leases not yet started of $25 billion, equity investments of $25 billion, and capital spending of $8 billion bring total future commitments to $366 billion. Some supply pacts can still be moved before firm orders, and changes can cost extra. The direction is not subtle: Nvidia is paying up to stand first in the memory line.

In the same quarter it issued $25.0 billion of senior notes and ended with $56.6 billion in cash, cash equivalents, and marketable debt securities, plus $42.8 billion of marketable equity securities. It is also writing land, power, and shell guarantees, including a cap of $105 billion tied to SB Energy, so that some of the campuses that will buy the chips can get built. That is the other side of the queue. Part of the demand Nvidia is rationing sits on its own paper.

Inventories were $31.6 billion at quarter end, up from $21.4 billion on January 25, as the company staged Vera Rubin. Accounts receivable were $63.1 billion. It returned about $26.0 billion to shareholders and still had about $99.0 billion left on its buyback authorization, with a $0.25 dividend due October 1 to holders of record on September 10.

Vera Rubin Raises the Take Per Gigawatt

Unit caps matter less if each shipped megawatt is worth more. Kress put Hopper-era sales at about $18 billion per gigawatt, Blackwell at about $25 billion, and Vera Rubin at about $40 billion. The newer stack spans the Vera CPU, Rubin GPU, NVLink, InfiniBand or Ethernet, and the Groq LPU.

REVENUE PER GIGAWATT OF AI FACTORY

• Hopper: About $18 billion of Nvidia sales per gigawatt of deployed capacity.
• Blackwell: About $25 billion per gigawatt as full-rack systems took a larger share.
• Vera Rubin: About $40 billion per gigawatt, with the platform expected to be about 20% of Data Center sales in the current quarter.
• Efficiency claim: 30 times higher throughput per megawatt and 35 times lower token cost than Grace Blackwell Ultra, per Kress.

Production shipments started in early August. Kress said Nvidia already has purchase orders from every major hyperscaler, AI cloud, and system OEM, and that Rubin should be the fastest product ramp in the company’s history. Networking had another record, up 18% sequentially, with Spectrum-X Ethernet at 2.6 times the year-ago level. Trailing-twelve-month Grace CPU revenue has exceeded $5 billion, and the early view is that CPU sales more than double in fiscal 2028 against about $20 billion of server-CPU demand Nvidia says it can see.

That mix shift is the quiet way around a memory shortage. Fewer extra wafers can still produce a much larger bill if the rack is a full AI factory rather than a tray of GPUs.

Who Is Already in Line for the Next Racks

The buyers waiting are named, funded, and in several cases already running Rubin racks. Kress said Nvidia compute is fully used across every cloud the company serves, and that more GPU capacity has been turning into more cloud revenue and wider margins for those customers.

WHO ALREADY HAS A PLACE IN LINE

• AWS expansion: An additional 2 million GPUs starting this quarter through the second quarter of fiscal 2029, along with Vera CPUs, some integrated with Rubin.
• Racks already live: CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure, and Nebius are among the partners already running Rubin systems.
• Neocloud build: Regional AI clouds are expected to exit the year with 8 gigawatts of installed capacity, up from about 3 gigawatts at the end of 2025.
• Spending behind them: Kress put the cloud industry backlog at greater than $2 trillion, with capex by the top five hyperscalers expected at nearly $800 billion in 2026 and $1.3 trillion in 2027.

The AI clouds, industrial, and enterprise bucket is on track to be about half of Data Center. Sovereign AI revenue rose 35% sequentially and more than tripled from a year ago, often through regional clouds that can take land and power a country would not hand a foreign hyperscaler. Global venture funding in AI topped $400 billion in the first half of 2026, above the $265 billion raised in all of 2025, and Kress said roughly 70% of that money is spent on compute.

Groq 3 LPX, the first rack-scale LPU system from last year’s Groq tie-up, is in full production, with volume shipments due later this quarter and Nebius as the first adopter. Huang’s software map still runs through Nvidia’s platform, from closed models at the large labs to open models. The allocation fight is among buyers who already standardized on that stack.

Leopold’s $1 Trillion Math Still Needs the Factory

Simon Leopold, an analyst at Raymond James, raised his price target 46% to $515 from $352 after the print and wrote that $1 trillion of sales in fiscal 2029 “seems possible.” FactSet’s fiscal 2029 consensus sits just under $750 billion, a gap of about $250 billion a year.

THE PATH AGAINST THE STREET

From about $686 billion, Nvidia would need another 46% in fiscal 2029 to reach $1 trillion. That is slower than the 106% the company just posted, and slower than the doubling customers say they want. Leopold did not map the extra wafers. Nvidia said yields and capacity should improve, and that 70% is what it can currently back.

A $108.0 billion quarter is already more than the $60.9 billion Nvidia sold in the entire fiscal year that ended in January 2024. First-half fiscal 2027 revenue was $177.8 billion. The run rate is not the constraint. The bill of materials is.

Price Increases Arrive After Margins Dip

Memory is also hitting the income statement. Kress said third-quarter gross margin should be 74.0%, plus or minus 50 basis points, then bottom in the fourth quarter in the 71% to 72% range. Fiscal 2028 is expected to settle at 72% to 73% once price increases take effect in the first quarter.

GROSS MARGIN PATH KRESS LAID OUT

1. Q2 FY27: GAAP and non-GAAP gross margin both land at 75.0%, helped by Blackwell Ultra mix.
2. Q3 FY27: Both measures guided to 74.0%, plus or minus 50 basis points, on higher memory cost.
3. Q4 FY27: Margins expected to bottom in the 71% to 72% range.
4. Fiscal 2028: Margins expected to settle at 72% to 73% after first-quarter price increases.

That path trades a few points of margin for a larger ticket and a locked supply queue. Non-GAAP earnings of $2.22 a share, up 120%, are the cleaner run-rate figure; GAAP net income of $59.7 billion includes large gains on equity securities, which is why profit rose only 2% sequentially on a GAAP basis. Free cash flow was $21.3 billion. GAAP operating expenses are seen at about $9.2 billion in the current quarter, with non-GAAP at about $9.0 billion.

Buyers are still lining up. The open variable through January 2028 is whether memory plants, packaging lines, and the richer rack mix can convert that queue into Leopold’s fiscal 2029 number without Nvidia having to keep writing ever-larger checks to stand first in line. Kress said supply will remain a bottleneck at least through the end of fiscal 2028, and that is now the date that sits on the $96.2 billion quarter.

Disclaimer: This article is news reporting and analysis of Nvidia’s public filings, earnings remarks, and named-analyst notes. It is informational only and does not constitute investment advice, a recommendation to buy or sell Nvidia or any other security, or a prediction of the share price. Readers should consult a licensed financial adviser or investment professional who can weigh their own objectives and risk tolerance before acting on any figure or scenario mentioned here, including third-party revenue projections. Revenue, margin, commitment, and guidance figures reflect company statements and data providers as of September 2, 2026, and may change with later filings, supply updates, or shifts in customer demand.