NVIDIA Q2 FY2027 Earnings: Revenue More Than Doubles, but Margin Outlook Resets on Memory Costs

NVIDIA reported $96 billion in fiscal second-quarter revenue, more than double the prior year, with Data Center revenue reaching $89 billion as Blackwell deployments and NeoCloud capacity additions drove growth. Management also gave an unusual preliminary fiscal 2028 outlook for roughly 70% revenue growth, citing demand above that level but supply constraints across its ecosystem.

The key offset is profitability. NVIDIA expects gross margin to decline from 75% in Q2 to a low of 71% to 72% in Q4 as memory costs rise, before recovering modestly to 72% to 73% in fiscal 2028 after planned price increases take effect.

As Reported on the Call

Management said it also delivered record operating income and EPS.

The Bigger Story

NVIDIA’s quarter reinforced that AI infrastructure spending remains broadening beyond the largest cloud providers. While hyperscale revenue reached $49 billion, the company’s ACIE business, which includes NeoClouds, industrial customers and enterprises, grew faster at 25% sequentially to $40 billion.

That mix matters because NVIDIA is positioning itself less as a GPU vendor and more as a supplier of complete AI factory systems, including CPUs, GPUs, networking and software. Management argues that this broader platform expands its revenue opportunity per deployed gigawatt of data-center capacity and makes its infrastructure more useful across training, post-training and inference workloads.

The strategic opportunity is substantial, but it also increasingly relies on NVIDIA’s willingness to support the financing and buildout of the broader AI ecosystem. The company has invested nearly $50 billion in frontier AI labs, introduced revenue-sharing arrangements with NeoCloud operators, and is selectively providing credit enhancement for some infrastructure deployments. These actions may accelerate deployment, but they also make the quality and sustainability of demand more important for investors to monitor.

Revenue and Operating Performance

Data Center remained the central driver of NVIDIA’s growth. Revenue from the segment rose 18% sequentially to $89 billion, supported by Blackwell demand across hyperscalers and by capacity additions at NeoCloud providers.

Hyperscale revenue increased 13% sequentially to $49 billion. Management said growth reflected sustained Blackwell demand as new GPU capacity came online. It also cited cloud-industry backlog of more than $2 trillion and estimated that capital expenditure by the five largest hyperscalers could reach nearly $800 billion in 2026 and $1.3 trillion in 2027. Those industry figures were management commentary, rather than NVIDIA-reported operating metrics.

The faster-growing ACIE business produced $40 billion of revenue, up 25% sequentially and 138% year over year. Management attributed that growth to NeoCloud capacity additions serving enterprises, AI start-ups and sovereign customers, as well as hyperscalers supplementing their own internal buildouts with external capacity.

NVIDIA said NeoCloud partners are expected to exit the year with 8 GW of installed capacity, up from about 3 GW at the end of 2025. That figure is capacity, not NVIDIA revenue, but it illustrates the scale of infrastructure being built by regional cloud providers around the company’s platform.

Sovereign AI was another fast-growing component. Management said this business, primarily delivered through regional NeoClouds, increased 35% sequentially and more than tripled year over year. It highlighted deployments and partnerships across Japan, South Korea, Europe, India, Africa, Taiwan, Australia and Malaysia.

Networking also posted a record quarter, with revenue increasing 18% sequentially. Spectrum-X Ethernet revenue grew 2.6 times year over year, according to management. NVIDIA is using networking, CPUs and software to capture a larger portion of the total data-center system value rather than limiting its exposure to accelerators.

Profitability and Margins

NVIDIA maintained both GAAP and non-GAAP gross margin at 75% in Q2, broadly unchanged from the prior quarter because product mix was similar. Operating expenses increased faster than revenue on a sequential basis, with GAAP operating expenses up 10% and non-GAAP operating expenses up 11%, driven primarily by compute infrastructure costs and compensation and benefits.

The more consequential development was the reset to the gross-margin outlook. Management cited “extreme pricing conditions” in memory and said component-cost inflation had exceeded prior expectations.

For Q3, NVIDIA expects GAAP and non-GAAP gross margin of 74%, plus or minus 50 basis points. Management expects margins to bottom at 71% to 72% in Q4, then settle at 72% to 73% in fiscal 2028 as price increases take effect in the first quarter.

This is not a minor modeling issue. A three- to four-point decline from Q2’s 75% margin would reduce the degree of operating leverage that investors have come to expect from NVIDIA’s revenue growth. The company’s ability to pass through higher memory costs without disrupting demand or competitive positioning will be a key test.

Management noted that shipments of Hopper 200 products to China represented less than 1% of total Data Center revenue in Q2 and were dilutive to corporate gross margin. Given geopolitical uncertainty, it included no China Data Center compute revenue in its forward outlook.

Demand, Customers and Pricing

Management’s demand message was unequivocally strong: it said customer forecasts point to NVIDIA’s revenue doubling in fiscal 2028, while its own preliminary 70% growth outlook reflects supply limitations rather than demand constraints.

The company said its compute is fully utilized across the clouds it serves. It attributed demand to increasing use of agentic AI, which requires more compute for reasoning, planning and tool use than conventional user-prompted AI workloads. This is management’s view of the market, but it is consistent with the company’s expanding product scope across training and inference.

NVIDIA also announced an expansion of its AWS partnership. AWS plans to deploy an additional 2 million NVIDIA GPUs starting in the current quarter through the second quarter of fiscal 2029, according to management. AWS is also expected to deploy Vera CPUs and use NVIDIA’s Nemotron models on Amazon Bedrock and SageMaker. The planned deployment is a significant demand signal, though timing and revenue recognition were not disclosed.

Vera Rubin began production shipments earlier in August, and management said it had purchase orders from every major hyperscaler, AI cloud and system OEM. NVIDIA expects Vera Rubin to be its fastest product ramp. It expects the platform to represent roughly 20% of Data Center revenue in Q3.

Management also said its Grace CPU business exceeded $5 billion in trailing-12-month revenue and expects CPU revenue to more than double in fiscal 2028, based on customer demand and improving supply. Groq 3 LPX, its rack-scale low-latency inference system, is in production and is expected to ship in volume later in the quarter to early adopters, with Nebius identified as the first customer.

What is NVIDIA’s guidance and outlook?

For Q3, NVIDIA expects revenue of $108 billion, plus or minus 2%, with ACIE expected to provide the primary source of sequential Data Center growth. Management expects hyperscale growth to reaccelerate in Q4 and into fiscal 2028 as Vera Rubin supply increases.

More notably, management gave preliminary fiscal 2028 guidance for approximately 70% year-over-year revenue growth. It said supply is expected to remain a bottleneck at least through the end of fiscal 2028.

That forecast provides more visibility than NVIDIA has typically offered, but it is also dependent on an unusually complex supply chain. Management described constraints across the entire ecosystem, including components, manufacturing capacity and the physical infrastructure needed to deploy AI systems, such as land, power, cooling and data-center shells.

Management’s presentation of revenue opportunity per gigawatt was not fully consistent. In prepared remarks, it said the opportunity had increased from roughly $18 billion with Hopper to $25 billion with Blackwell and $40 billion with Vera Rubin. In Q&A, Jensen Huang cited higher figures, including an estimate that Vera Rubin could represent about $60 billion per gigawatt. Investors should treat these figures as directional management illustrations rather than a precise revenue framework.

Balance Sheet, Cash Flow and Capital

Inventory rose to $32 billion as NVIDIA prepared for the Vera Rubin launch. Days sales outstanding increased to 60 days, reflecting extended payment terms for certain large purchases by investment-grade customers that will be shipped over multiple quarters.

The increase in receivables terms is worth watching. Extended payment terms can support major deployments, but they can also increase working-capital needs and raise the importance of customer credit quality as NVIDIA scales sales to a wider set of AI infrastructure providers.

NVIDIA returned $26 billion to shareholders in Q2, including $20 billion in share repurchases and $6 billion through dividends. Management said it has returned 60% of free cash flow year to date, above its stated plan to return at least 50%, and intends to increase returns of excess free cash flow after strategic uses.

Those strategic uses are becoming material. NVIDIA said it has invested nearly $50 billion in frontier AI labs and has partnered with six infrastructure capital providers to establish financing platforms intended to raise more than $500 billion of third-party capital. It also said it will provide selective credit enhancement for nearly 2 GW of compute for one frontier lab and has supported revenue guarantees for some NeoCloud projects.

Management characterized these arrangements as ecosystem development rather than circular financing, arguing that its infrastructure is fungible and can be redeployed. Still, these commitments connect a growing share of future demand to the financing capacity and business performance of AI labs and NeoCloud operators.

What Investors Should Watch

1. Vera Rubin Ramp and Supply Availability

Vera Rubin is expected to account for about 20% of Data Center revenue in Q3 and is central to NVIDIA’s fiscal 2028 outlook. Investors should watch whether production and supply-chain expansion support the company’s stated 70% growth target without significant delivery delays.

2. Gross-Margin Trough and Price Increases

Management expects gross margin to fall to 71% to 72% in Q4 before recovering to 72% to 73% in fiscal 2028. Evidence that price increases offset memory inflation would support the profitability outlook; a deeper or more prolonged margin decline would weaken it.

3. ACIE and NeoCloud Revenue Conversion

ACIE grew faster than hyperscale revenue in Q2 and is expected to drive Q3 growth. The key question is whether capacity additions at NeoClouds and sovereign customers continue translating into high-quality hardware revenue and, eventually, the recurring revenue-sharing streams management has outlined.

4. Frontier-Lab Financing Exposure

Management expects AI labs receiving balance-sheet support to account for roughly a quarter of its business next year. Investors should monitor the scale, terms and credit exposure of future commitments, particularly as NVIDIA supports customers that may not yet have the financing profile of major hyperscalers.

5. Working Capital and Customer Terms

Inventory and days sales outstanding both increased in Q2. Continued increases could be appropriate during a major product transition, but they would also indicate greater capital intensity and potentially more exposure to customer payment timing.

AlgovestIQ Takeaway

NVIDIA’s Q2 results showed that AI infrastructure demand remains stronger and broader than the company can currently supply. Data Center revenue grew 18% sequentially to $89 billion, ACIE accelerated faster than hyperscale revenue, and management now sees sufficient visibility to project roughly 70% fiscal 2028 revenue growth.

The investment case has also evolved. NVIDIA is increasingly selling a full data-center platform, expanding its exposure to networking, CPUs, software and low-latency inference. This broader system position could support continued growth even as customers diversify their compute architectures or develop specialized chips for certain workloads.

However, the near-term earnings story is no longer solely about revenue growth. Memory inflation is pushing gross margins down, and management’s expected 71% to 72% Q4 margin range marks a meaningful reset from the 75% reported in Q2. The recovery depends in part on planned pricing actions, which have yet to be tested in reported results.

The other unresolved issue is the growing role NVIDIA is playing in enabling its own ecosystem’s infrastructure financing. Investments, credit enhancements and revenue guarantees may help unlock demand that otherwise would be delayed by financing or data-center development constraints. But they also increase the importance of deployment execution, customer credit quality and the durability of AI-lab economics.

The central question for investors is whether NVIDIA can convert extraordinary demand visibility into sustained revenue growth while preserving most of its margin structure and limiting balance-sheet risk. Faster Vera Rubin supply, successful pricing actions and continued ACIE conversion would strengthen that case. Further margin pressure, rising working-capital needs or expanding financing commitments would warrant closer scrutiny.