NVIDIA revenue doubles to $96.2bn as $108bn Q3 guide excludes China data center compute

NVIDIA Corporation (NASDAQ: NVDA) has delivered another acceleration in the global AI infrastructure cycle, reporting fiscal second-quarter 2027 revenue of US$96.22 billion, up 106% year over year and 18% sequentially, while Data Center revenue surged 117% to US$89.0 billion. GAAP operating income increased 124% to US$63.73 billion and net income rose 126% to US$59.69 billion, producing a quarterly net margin of approximately 62%. The company is now guiding to US$108 billion of third-quarter revenue, plus or minus 2%, despite explicitly assuming no Data Center compute revenue from China in that outlook.

The scale of the business has moved far beyond merely benefiting from higher GPU prices. Data Center represented approximately 92.5% of total second-quarter revenue, meaning NVIDIA’s financial performance is increasingly tied to whether hyperscalers, frontier AI laboratories, sovereign customers and enterprises continue expanding AI factories at extraordinary rates. The company simultaneously announced that Amazon Web Services plans to deploy another two million NVIDIA GPUs across its global infrastructure during 2027 and 2028, adding a concrete hyperscaler commitment to the demand signals embedded in the earnings report.
The quarter also shows why the next stage of NVIDIA’s growth is becoming increasingly capital-intensive throughout the ecosystem rather than merely inside NVIDIA itself. Vera Rubin has entered full production, major cloud providers are already running Rubin racks, and NVIDIA is working with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR on financing platforms intended to mobilize more than US$500 billion of third-party capital for AI infrastructure over time. The investment question is therefore shifting from whether AI demand exists toward whether the power, memory, networking, financing and physical data-center infrastructure required to satisfy that demand can expand quickly enough.

How much larger has NVIDIA become in only one year?

Second-quarter revenue increased from US$46.74 billion a year ago to US$96.22 billion, meaning NVIDIA added approximately US$49.48 billion of quarterly revenue in twelve months. That incremental revenue by itself is larger than the total quarterly sales of many of the world’s biggest semiconductor companies, illustrating how unusual the current AI investment cycle has become. Sequential growth also remained strong, with revenue increasing approximately US$14.61 billion from the US$81.62 billion generated in fiscal Q1.

Profitability expanded even faster than revenue on a year-over-year basis. GAAP operating income rose from US$28.44 billion to US$63.73 billion, while GAAP net income increased from US$26.42 billion to US$59.69 billion. NVIDIA consequently converted approximately 66 cents of every revenue dollar into operating income and about 62 cents into net income during the quarter, although the net result also benefited from US$7.77 billion of net gains from equity securities.

The equity gains explain why GAAP net income increased only 2% sequentially even though revenue rose 18% and operating income increased 19%. NVIDIA generated US$7.77 billion of net gains from equity securities in Q2, while the corresponding six-month figure reached US$23.71 billion, making investment valuations an increasingly visible component of reported earnings. Non-GAAP net income, which removes specified investment and other effects, increased 18% sequentially to US$53.95 billion and 118% year over year.

Why does the $108bn Q3 guidance matter more because China is excluded?

NVIDIA expects third-quarter revenue of US$108 billion, plus or minus 2%, which implies approximately 12.2% sequential growth at the midpoint from the US$96.22 billion reported in Q2. Achieving that rate would add another roughly US$11.8 billion of quarterly revenue and take the company above a US$100 billion quarterly run rate for the first time. The guidance becomes more consequential because NVIDIA says it assumes no Data Center compute revenue from China, meaning the forecast depends on demand elsewhere rather than a reopening of a strategically important but heavily restricted market.

China has historically represented a meaningful semiconductor market, but U.S. export controls have repeatedly restricted which advanced accelerators NVIDIA can sell there. Removing China Data Center compute from the guidance lowers one source of upside but also makes the outlook cleaner for investors trying to assess underlying demand from the United States, Europe, the Middle East, Asia-Pacific and sovereign AI programmes. If China revenue eventually returns under permitted configurations, that could create upside relative to guidance, but current investors do not need that assumption for NVIDIA to target US$108 billion next quarter.

The midpoint would also place Q3 revenue approximately 2.3 times the US$46.74 billion NVIDIA generated only one year before the latest quarter. Growth at that scale is difficult to sustain indefinitely because the comparison base rapidly becomes larger, yet management is continuing to describe demand as supply constrained rather than demand constrained. The tension investors now need to follow is therefore whether infrastructure availability becomes the primary ceiling on growth before customer economics begin limiting AI capital expenditure.

What does AWS’s two-million-GPU commitment reveal about hyperscaler demand?

AWS and NVIDIA announced immediately after the earnings release that Amazon’s cloud business plans to deploy two million additional NVIDIA GPUs across 2027 and 2028. The programme will include Blackwell Ultra, Rubin and Rubin Ultra accelerators and comes on top of AWS’s earlier plan, announced at GTC 2026, to add more than one million NVIDIA GPUs beginning in 2026. AWS said demand since that earlier announcement has exceeded its expectations.

The expansion means the incremental 2027-2028 deployment alone is twice the size of the more-than-one-million-GPU programme announced only months earlier. Neither company disclosed the dollar value of the new commitment, so multiplying two million units by public accelerator prices would be inappropriate because system configurations, networking, CPUs, memory, volume economics and commercial terms vary significantly. The strategically useful number is the unit commitment itself, which demonstrates that one hyperscaler is planning GPU capacity on a scale measured in millions rather than tens of thousands.

The partnership extends beyond accelerators. AWS plans to bring NVIDIA Vera CPU infrastructure into its cloud, build U.S. government AI factories including 100,000 GPUs on secure AWS infrastructure, integrate NVIDIA networking with AWS Nitro and Elastic Fabric Adapter technologies and deploy NVIDIA physical-AI technology within Amazon Robotics. Those initiatives increase the number of layers in which NVIDIA technology can generate demand, extending the relationship from GPU supply into CPUs, networking, robotics and model infrastructure.

Is NVIDIA maintaining margins while revenue scales this quickly?

GAAP gross margin reached 75.0%, up 260 basis points from 72.4% a year earlier and essentially unchanged from 74.9% in Q1. Non-GAAP gross margin was also 75%, demonstrating that the massive revenue expansion has not required a major near-term sacrifice in product economics. NVIDIA expects Q3 gross margin around 74%, plus or minus 50 basis points, implying a modest sequential decline as newer systems and component costs affect the mix.

Operating expenses are increasing quickly but much more slowly than revenue. GAAP operating expenses rose 55% year over year to US$8.41 billion while revenue more than doubled, causing operating leverage to expand substantially. Research and development expense alone reached US$7.05 billion, up from US$4.29 billion, reflecting the cost of sustaining annual architecture transitions across GPUs, CPUs, networking, software and robotics platforms.

That relationship is central to NVIDIA’s current profitability. Spending another US$2.76 billion on quarterly R&D while adding almost US$49.5 billion of year-over-year revenue produces enormous operating leverage when gross margins remain around 75%. The longer-term question is whether increasingly complex AI systems, higher memory costs and greater infrastructure involvement eventually require NVIDIA to surrender part of that margin as the company moves from selling accelerators toward supplying entire AI-factory architectures.

What does NVIDIA’s balance sheet show about the cost of supporting this growth?

NVIDIA’s operating scale is increasingly visible in working capital. Accounts receivable rose to US$63.06 billion at July 26 from US$38.47 billion at the January fiscal year-end, an increase of approximately 64%, while inventory climbed to US$31.58 billion from US$21.40 billion, up roughly 48%. Those increases reflect the amount of capital required to support rapidly expanding customer shipments and next-generation product transitions even for a company with exceptional margins.

Operating cash flow remained substantial at US$24.08 billion during Q2 but was well below GAAP net income of US$59.69 billion because receivables, inventory and other working-capital movements absorbed considerable cash. Accounts receivable alone consumed approximately US$22.35 billion of quarterly operating cash flow, while inventory consumed another US$5.78 billion. The gap illustrates why revenue growth of this magnitude creates balance-sheet demands even when reported profitability is extremely high.

NVIDIA nevertheless returned approximately US$26 billion to shareholders through share repurchases and dividends during the quarter and still had approximately US$99 billion remaining under its repurchase authorization. The company will also pay a US$0.25 quarterly dividend in October. Capital returns at that scale indicate that management believes the company can fund product development and strategic investments while continuing to return substantial excess capital.

Why does Vera Rubin matter to NVIDIA’s next growth phase?

Vera Rubin has now moved into full production, with systems running at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius. That transition matters because NVIDIA’s revenue growth cannot depend indefinitely on Blackwell alone; each architecture must hand demand to the next generation without causing customers to delay purchases while waiting for newer technology.

The Rubin platform is broader than a GPU upgrade because it integrates Vera CPUs, Rubin GPUs, Spectrum networking and other components into an AI-factory architecture. NVIDIA has also put its Groq 3 LPX inference accelerator into full production and expanded software around agentic and physical AI, widening the workloads the company hopes to monetize. This increasingly system-level strategy gives NVIDIA opportunities to capture more value per AI deployment but simultaneously increases dependence on components, memory, networking and power infrastructure outside its direct control.

That is why the US$500 billion third-party infrastructure financing initiative is strategically connected to the earnings story. NVIDIA can produce accelerators only if customers have sufficient capital, powered data-center capacity and financing to install them, so facilitating infrastructure investment can indirectly expand the market for NVIDIA’s own products. The company is effectively helping to solve bottlenecks surrounding its customers because those bottlenecks have become one of the largest constraints on NVIDIA’s continued revenue growth.

What is the central question after another record NVIDIA quarter?

The latest numbers leave little doubt that AI infrastructure spending remains exceptionally strong. Revenue doubled, Data Center revenue more than doubled, margins stayed around 75%, AWS committed to another two million GPUs and NVIDIA is guiding to a US$108 billion quarter without relying on China Data Center compute. Those facts suggest the current cycle still has substantial momentum even after NVIDIA reached a revenue scale that would normally make triple-digit growth mathematically difficult.
The more important uncertainty has shifted toward physical and financial capacity. Accounts receivable and inventory are expanding rapidly, customers need enormous amounts of power and data-center infrastructure, next-generation systems depend on advanced memory and packaging, and NVIDIA is helping mobilize hundreds of billions of dollars of outside capital to keep AI factories moving. None of those constraints necessarily implies demand is weakening, but they increasingly determine how much of that demand can be converted into shipments within a particular quarter.

NVIDIA’s fiscal Q2 therefore represents more than another earnings beat. The company has reached a point where its growth depends on the ability of an entire infrastructure ecosystem to expand around it, while the two-million-GPU AWS commitment suggests hyperscalers are still planning for substantially more capacity rather than preparing for an AI investment slowdown.