Nvidia reported fiscal 2027 second-quarter revenue of $96.22 billion, beating the midpoint of its own forecast by $5.2 billion, but its shares fell as much as 4% in after-hours trading after the company projected lower gross margins for the next quarter.
The results underscored the scale of demand for Nvidia’s AI hardware while exposing a more difficult question for the market: whether the costs of delivering increasingly complex rack-scale systems will begin to narrow the company’s exceptional profitability. Nvidia reported a 75.0% GAAP and non-GAAP gross margin for the quarter, then guided for 74.0% in the fiscal third quarter, below the 75% level expected by consensus estimates cited in the supplied material.
Revenue rose 106% from a year earlier and 18% from the prior quarter. Adjusted earnings per share reached $2.22, up 120% year over year, while non-GAAP operating profit increased 124% to $63.96 billion. Nvidia’s reported non-GAAP operating expenses were $8.23 billion, slightly below the $8.32 billion expectation referenced in the material.
Data center sales drive the quarter
The company’s data center business generated $89.0 billion in revenue, up 117% from a year earlier and 18% from the preceding quarter. That division accounted for roughly 92.5% of Nvidia’s total quarterly sales, based on the company’s reported figures, showing how decisively the business has become tied to AI infrastructure spending.
Hyperscale cloud customers produced $48.71 billion in revenue, exceeding the $43.55 billion expectation cited in the supplied material. Those sales were the main source of the quarterly outperformance, adding about $5.2 billion above expectations.
Revenue from AI cloud providers, industrial users and enterprise customers totaled $40.31 billion. That was below the $41.96 billion expectation included in the material, suggesting that the strongest demand remained concentrated among the largest cloud operators rather than being evenly distributed across corporate and specialized AI customers.
Compute and networking revenue reached $88.3 billion, compared with an $84.69 billion estimate. Edge computing revenue came in at $7.2 billion, up 27% from a year earlier and 13% from the prior quarter, exceeding the $6.61 billion estimate cited in the material.
Excluding $7.77 billion in equity investment gains, Nvidia reported non-GAAP net income of $53.95 billion. The company’s ability to produce that level of profit while absorbing the costs of a major product transition remains unusual even within the high-margin semiconductor sector. Yet the lower third-quarter margin target showed that the mix of products being delivered is changing.
Rack systems change the cost equation
Nvidia said its Vera Rubin platform is moving into full production more quickly, with rack systems running at CoreWeave and Google Cloud. The company also cited deployments at Microsoft Azure, Oracle Cloud Infrastructure and Nebius.
The shift from selling standalone graphics processing units to shipping larger integrated systems gives Nvidia more control over the architecture delivered to cloud customers. It also makes the company more exposed to the costs and supply constraints surrounding high-bandwidth memory, advanced chip packaging, substrates and the networking equipment needed to build complete AI racks.
Nvidia linked its fiscal third-quarter margin outlook to early-stage ramp costs, supply-chain expenses and a higher mix of full-rack systems. Those systems carry more components and logistical complexity than individual GPUs, which can alter the cost structure even when demand remains strong.
The margin forecast appears to have outweighed the headline revenue beat in the immediate market reaction. Nvidia’s valuation has been supported by both extraordinary sales growth and unusually high profitability. A modest decline in gross margin does not erase either advantage, but it places greater attention on whether the company can maintain its cost discipline as customers demand increasingly complete data center installations.
Capital commitments widen Nvidia’s role in AI infrastructure
Nvidia has also expanded its financial exposure across the AI supply chain through investments, partnerships and technology agreements.
The company disclosed a $5 billion investment in Intel, completed through a private placement of more than 214.7 million shares on Dec. 29, 2025, according to the supplied material. The transaction followed the U.S. government’s reported agreement to acquire about 10% of Intel for $8.9 billion.
Nvidia also committed up to $100 billion to OpenAI in September 2025, then increased its total OpenAI commitment to $250 billion in July 2026, according to the material. It invested $10 billion in xAI in January 2026.
In August, Nvidia completed a Poolside transaction that included a $6 billion license for the Model Factory AI model and a $1 billion investment at a $12 billion pre-money valuation, according to the supplied information. More than 100 Poolside employees received job offers connected with Nvidia’s Nemotron open-model project, while Poolside said it would continue operating independently under its three co-founders.
These arrangements extend Nvidia’s role beyond supplying chips. They connect the company more closely to the model developers, cloud providers and system builders that will need large volumes of computing capacity. They also raise the stakes if AI infrastructure demand slows, because spending across the ecosystem has increasingly been financed through large commitments and debt issuance.
Custom chips and infrastructure debt add pressure
The supplied material estimated that hyperscale cloud companies and Nvidia-related entities issued $225 billion in AI infrastructure bonds year to date in 2026, a 973.7% increase from the prior year. Such borrowing reflects confidence that AI computing demand will produce durable cash flows, but it also makes the economics of new data centers more sensitive to equipment delivery schedules, power availability and utilization rates.
Supply limits in high-bandwidth memory and advanced packaging remain central constraints. A shortage can restrict Nvidia’s ability to ship systems even with strong customer orders, while also increasing component costs during the Vera Rubin ramp.
Competition is also becoming more complex. The supplied material estimated that custom silicon would rise from 20.9% of the AI chip market in 2025 to 27.8% in 2026. Google’s TPU, Amazon’s Trainium and Meta’s MTIA are among the in-house accelerator programs designed to reduce dependence on external chip suppliers for selected workloads.
Broadcom’s AI-related revenue was cited at roughly $10.8 billion per quarter, reflecting the growing commercial opportunity for custom accelerators and networking hardware. These chips are not direct replacements for Nvidia across every training and inference task, but they give the largest cloud companies additional leverage and alternative capacity for workloads tailored to their own platforms.
Nvidia’s quarter showed that hyperscale AI spending remains powerful enough to lift revenue far above already elevated expectations. The weaker margin outlook indicates that converting that demand into profit will become more operationally demanding as the company moves deeper into full-system deliveries and as cloud customers pursue more of their own silicon strategies.
Explore how traditional markets meet digital assets—discover key insights in our TradFi and crypto integration guide today.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
