Earnings Recap — Q2 FY2027
CY Q3 2026 · Reported August 26, 2026 · Beat 7 of last 7 quarters
NVIDIA Corporation reported Q2 FY2027 revenue of $96.22B, a beat of 4.3% against consensus, and EPS of $2.22, a beat of 6.2%.
The Buildout is a website built on agentic AI tools, currently in beta, and could have factual errors.
NVIDIA's Q2 results and FY28 outlook reinforce the accelerating AI infrastructure buildout, with data center revenue nearly doubling year-over-year and management guiding to ~70% growth next year despite supply constraints. The company's expanding full-stack platform—spanning GPUs, CPUs, networking, and now Groq LPUs—is capturing a larger share of the data center TAM, with revenue per gigawatt rising to $40 billion for Vera Rubin. The announcement of over $500 billion in third-party financing platforms and multi-gigawatt site commitments with partners like AWS and SoftBank Energy signals a sustained multi-year buildout of AI factories, which will drive demand across the entire supply chain.
NVIDIA delivered record revenue of $96 billion, more than doubling year-over-year, with data center revenue of $89 billion up 18% sequentially. Hyperscale revenue grew 13% sequentially to $49 billion, while ACIE revenue grew 25% sequentially and 138% year-over-year to $40 billion, driven by NeoCloud capacity additions. Networking revenue grew 18% sequentially, with Spectrum-X Ethernet up 2.6x year-over-year. The company returned a record $26 billion to shareholders in the quarter. Management also noted that they shipped less than 1% of total data center revenue in Hopper 200 products to customers based in China in Q2, with no China data center compute revenue in the forward outlook.
Management provided an unprecedented full-year outlook, guiding to approximately 70% revenue growth in fiscal 2028, a supply-constrained number against demand growth that is expected to double. They expect supply to remain a bottleneck through at least the end of fiscal 2028. Gross margins are expected to decline to 74% in Q3, bottom at 71-72% in Q4, and settle at 72-73% in fiscal 2028, driven by extreme memory pricing conditions that are expected to worsen into next year before executed price increases take effect in Q1. Vera Rubin production shipments began earlier this month, with purchase orders from every major hyperscaler, AI cloud, and system OEM, and management expects it to be the fastest product ramp in NVIDIA's history. They also highlighted expanding partnerships, including a 2 million GPU deployment with AWS through Q2 FY29, a 4.25 GW Portsmouth campus with SoftBank Energy for OpenAI, and over $500 billion in third-party financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to support Frontier AI lab infrastructure build-outs.
“We expect to grow revenue by approximately 70% in fiscal 2028. This is a supply-constrained outlook.”
on FY28 revenue outlook
“We want to be direct about this rather than let it linger as an open question. Memory scarcity today is being driven in large part by the AI build-out itself and unlike a component that simply raises our cost with no offset benefit. Tighter memory supply is a symptom of the same demand surge that's driving our own growth.”
on Memory pricing and gross margin
“We recognize the scale of this support, and we know some will call this circular financing. We see it differently. We're going through a major computing platform shift, the creation of one of the most important technologies in human history and these are once in a generation companies.”
on Frontier AI lab investments
What gives you confidence to guide a full year out, and what's the gap between 70% growth and 100% demand growth? What is the key constraint?
Jensen highlighted that AI agents require 15-100x more compute than human-prompted AI, and that NVIDIA's full-stack platform addresses a broader market including sovereign AI, NeoClouds, and enterprises growing ~100% annually. He noted that supply chain constraints, including land, power, and shell, limit growth to 70%, and that NVIDIA has unprecedented visibility upstream and downstream, allowing them to guide a year ahead.
How do you see your inference market share evolving with agentic AI, given the growing TAM per generation and the addition of Groq 3 LPX?
Jensen explained that the AI lifecycle now spans data preparation, pre-training, post-training, and agentic inference, all of which benefit from NVIDIA's fungible rack-scale architecture. He noted that revenue opportunity per gigawatt has grown from $18B with Hopper to $25B with Blackwell to $40B with Vera Rubin, and that Groq 3 LPX adds high-interactivity inference capabilities, but the vast majority of data centers will use Vera Rubin and NVLink 72.
Is the sum of ecosystem investments over the next several years around $500 billion, and how are you balancing investments in frontier labs that are designing their own custom chips?
Colette clarified that the commitments are primarily supply commitments essential for ramping Vera Rubin, with the largest portions in the first three years. Jensen emphasized that NVIDIA builds a full-stack platform that runs every model and is used across all clouds, and that he expects frontier labs to remain customers and partners for a long time, expressing regret he didn't invest more and sooner.