NVIDIA delivered another exceptional quarter with triple-digit revenue growth. The company recorded robust performance across almost every segment and subsegment, sending its share price up roughly 10% within a few days of the earnings release.
| Segmental Revenue | Q2 2026 (US$ million) | Q2 2027 (US$ million) | Percentage change |
| Data Centre | 41,096 | 89,023 | +116.6% |
| Hyperscale | 24,168 | 48,710 | +101.5% |
| AI Clouds, Industrial, & Enterprise (ACIE) | 16,928 | 40,313 | +138.1% |
| Edge Computing | 5,647 | 7,198 | +27.5% |
| Total | 46,743 | 96,221 | +105.9% |
Revenue growth accelerated for four consecutive quarters, driven by surging AI demand as global players race to build out data centres, with negligible revenue contribution from China. Strong revenue growth in the data centre segment was supported by robust agentic AI demand. The hyperscale subsegment was driven by sustained Blackwell demand. ACIE revenue grew 25.2% sequentially, driven by AI Clouds/neoclouds as well as hyperscalers that rent excess compute capacity from them.
Hyperscaler demand remains robust as planned capital expenditures among the top five hyperscalers approach US$800 billion in 2026 and US$1.3 trillion in 2027. Non-hyperscaler growth, namely NVIDIA’s ACIE (AI Clouds, Industrial, and Enterprise) subsegment, is expected to reach half of data centre segmental revenue, up from the existing 45.3%.
NVIDIA returned US$26 billion to shareholders during the quarter including US$20 billion through share repurchases and US$6 billion through dividend payments. Year-to-date, the company has returned 60% of its free cash flow to shareholders, outpacing its 50% target plan.
Revenue in Q3 FY2027 is expected to reach around US$108 billion, driven primarily by the ACIE subsegment. Revenue growth from hyperscalers is anticipated to re-accelerate in Q4 FY2027 and FY2028 as the company supplies more Vera Rubin. Vera Rubin alone is expected to account for about 20% of data centre revenue in Q3 FY2027.
Revenue in FY2028 is anticipated to rise 70% year-on-year, driven by agentic AI reaching what CEO Jensen Huang termed a major inflection point as well as price increases. This guidance is well ahead of Wall Street expectations and rather conservative amid ongoing supply constraints and excluding any revenue from China. If all demand is met, revenue is expected to more than double in FY2028. The supply constraints are expected to persist until January 2028. Gross margin will decline sequentially from 75.0% to around 74% in Q3 FY2027 due to higher-than-expected memory prices. The margin is expected to drop further to between 71% and 72% in Q4 FY2027, before rebounding to between 72% and 73% in FY2028.
Data Centre
Launched in 2021, Grace CPU’s trailing twelve-month revenue surpassed US$5 billion. Its successor Vera CPU is currently in production with initial shipments to Oracle Cloud, SpaceX AI, and AWS and is expected to tap into a total addressable market of US$20 billion. NVIDIA’s CPU revenue is expected to more than double in FY2028.
Vera Rubin delivers 30x higher throughput per megawatt and 35x lower token cost relative to Grace Blackwell Ultra, addressing data centre power constraints and driving down AI inference costs for customers.
Revenue from the networking business grew 18% sequentially, driven by the 2.6x growth in Spectrum-X.
ACIE
Encompassing sovereign regional neoclouds, enterprises, industrials, and air-gapped data centres, this subsegment is often overlooked, as noted by CEO Jensen Huang. Growth across this diverse customer base is supported by expanding global venture capital investment in AI. Delivered primarily through regional neoclouds, sovereign AI revenue grew 35% quarter-on-quarter and more than tripled year-on-year in Q2 FY2027.
NVIDIA introduced a revenue-sharing structure to help its neocloud customers scale. NVIDIA provides minimum revenue guarantees to secure project financing for these facilities. In exchange, NVIDIA takes a cut of the extra sales earned, creating a recurring revenue stream alongside initial hardware sales.
Frontier AI labs have spent nearly US$50 billion on infrastructure investments and remain limited by compute availability. NVIDIA is partnering with infrastructure capital providers to raise over US$500 billion for these labs, which are projected to generate about a quarter of NVIDIA’s business next year.
Key analyst questions
Huang noted the ongoing increase in total data centre buildout costs. NVIDIA continues to capture a larger share of the total expenditure per gigawatt, rising from US$18 billion for Hopper to US$25 billion for Grace Blackwell and US$40 billion for Vera Rubin.
Huang also expressed regret over not investing more and earlier in key partners like Anthropic and OpenAI. These investments will eventually generate more demand for NVIDIA hardware as compute drives revenue.
Huang dismissed concerns that open-source models threaten growth as demand for both open and closed models is increasing and most of these models run on NVIDIA infrastructure.
The fifth perspective
Even at its scale, NVIDIA’s growth trajectory remains exceptionally strong. The company sits at the centre of global AI infrastructure spend across hyperscalers, neoclouds, and frontier AI labs. As the transition to full-rack platforms increases NVIDIA’s revenue capture per gigawatt of compute deployed, the company remains the core beneficiary of the ongoing global AI race.
