Single-quarter revenue tops $100 billion—does Nvidia still need Jensen Huang?
- Core Viewpoint: Nvidia's first earnings report after Jensen Huang's departure comprehensively beat expectations, with quarterly revenue surpassing $100 billion for the first time. AI computing demand continues to spread across diverse scenarios including inference, agents, and sovereign AI, but the transition from founder to professional manager brings governance and strategic uncertainty.
- Key Elements:
- Fiscal 2027 Q3 revenue reached $115.8 billion, up 103% year-over-year; data center revenue was $107.6 billion, up 110% year-over-year; net income was $65.3 billion, up 115% year-over-year—all exceeding market expectations.
- Q4 revenue guidance is $130 billion, and the fiscal 2028 growth guidance was raised from 70% to 75%; cloud industry backlog reached $2.5 trillion, an increase of $500 billion in a single quarter, with demand far outstripping supply.
- Demand structure is diversifying: frontier model labs are expected to contribute approximately 15% of data center revenue; sovereign AI revenue more than tripled year-over-year; Agentic AI accounts for roughly 80% of OpenRouter's weekly token workload, driving CPU and networking demand.
- Gross margin this quarter was 74.8%, benefiting from an improved Blackwell Ultra product mix; however, Q4 gross margin guidance dropped to 71%-72%, reflecting continued supply chain cost pressures including memory.
- The next-generation platform Vera Rubin contributed nearly $20 billion in revenue in its first quarter of mass production, making it the fastest-ramping platform in history; networking revenue exceeded $21 billion, and the CPU business is expected to more than double next fiscal year.
- The board unanimously accepted Jensen Huang's resignation, has initiated a formal CEO search, and established a new National Security and Export Controls Oversight Committee. The successor selection and strategic continuity constitute the primary uncertainties.
Original | Odaily@OdailyChina)
Author | Golem (@web3_golem)

On November 17, after U.S. market hours, Nvidia released its first earnings report since Jensen Huang's departure, overall beating market expectations. Nvidia has officially entered the era of quarterly revenue exceeding $100 billion.
The report shows Nvidia's FY2027 Q3 revenue came in at $115.8 billion, above the market expectation of $109.1 billion, up 103% year-over-year and 20% quarter-over-quarter; data center business remained the core revenue source, reaching $107.6 billion, up 110% year-over-year and 21% quarter-over-quarter; Q3 net income was $65.3 billion, above the market expectation of $60.406 billion, up 115% year-over-year and 9% quarter-over-quarter; GAAP and non-GAAP gross margins were 74.8% and 74.9% respectively, while GAAP and non-GAAP diluted earnings per share were $2.70 and $2.68 respectively, both above market expectations.

Nvidia expects FY2027 Q4 revenue of $130 billion, above the market expectation of $125.2 billion. In addition, interim CEO and CFO Colette Kress also gave a higher growth forecast for FY2028 revenue on the investor call, raising the growth rate from 70% to approximately 75%.The reason Nvidia can provide such high future revenue guidance lies precisely in what Colette Kress repeatedly emphasized on the investor call: "Demand is far outpacing supply."
Affected by the earnings report and the call, Nvidia rose more than 4% after hours. This earnings report alleviated the market's anxiety over Jensen Huang's departure. On the investor call, Nvidia's executive team did not respond to any questions regarding Jensen Huang's resignation or the ongoing investigation.
Demand Is Spreading to More Players Across the AI Industry
Judging from the earnings data Nvidia provided this time, there are still no signs of demand cooling.
Interim CEO and Chief Financial Officer Colette Kress revealed on the call that the cloud computing industry's backlog has reached $2.5 trillion, up from $2 trillion three months ago. An increase of $500 billion in just one quarter indicates that cloud service providers still have substantial infrastructure buildout demand that has yet to be fulfilled over the coming years.
Moreover, this demand is not only coming from traditional hyperscale cloud providers. Nvidia said current demand covers hyperscalers, Neoclouds, frontier AI labs, enterprises, and governments around the world, and the customer base is becoming increasingly diversified.
Nvidia expects that major frontier model labs will contribute about 15% of its data center revenue this year. Take Anthropic as an example: as of the end of October, its annualized revenue run rate reached $81 billion. Model companies' own commercialization capabilities are strengthening, which also means they have more funds to continue investing in compute infrastructure.
These purchased GPUs have not been left idle as the market has questioned. Nvidia said the compute resources it provides to major hyperscale cloud providers and Neoclouds remain at 100% utilization, and customers are still using deployed GPUs as fully as possible.
Nvidia's demand expansion is no longer limited to model training.
Nvidia mentioned on the call that weekly token usage on OpenRouter grew 35x year-over-year, and 3.5x in the past three months alone, with Agentic AI workloads already accounting for about 80%. As AI agents begin to autonomously call tools and execute tasks, model inference is no longer a simple question-and-answer exchange but may run continuously for longer periods, further increasing demand for GPUs, CPUs, and networking infrastructure.
At the same time, sovereign AI and enterprise AI are also becoming new growth sources. Nvidia revealed that Q3 sovereign AI revenue more than tripled year-over-year, physical AI revenue nearly doubled year-over-year, and full-year physical AI revenue is expected to exceed $25 billion. Enterprise-side financial services performed particularly well, with on-premise deployment revenue approaching $5 billion over the past 12 months, more than doubling year-over-year.
So it is no longer just hyperscale cloud providers buying GPUs; AI compute demand is also spreading to more industries and application scenarios. Model training, inference, agents, enterprise deployment, sovereign AI, and even robotics are all becoming new compute consumers and will all become Nvidia's customers.
Supply Chain Continues to Expand, Gross Margin Beats Expectations
The market had previously been concerned that advanced process nodes, HBM memory, and advanced packaging capacity such as CoWoS could become bottlenecks for continued Blackwell and Vera Rubin ramp-ups. In particular, with memory prices rising, the market worried that even with ample orders, Nvidia might sacrifice profits due to rising supply chain costs.
On this call, Colette Kress specifically addressed supply chain issues. She said Nvidia is working with partners to expand capacity in key areas such as wafers, substrates, and advanced packaging. On substrates, the company reached new agreements this quarter with Kinsus and Samsung Electro-Mechanics; on advanced packaging, it reached a new agreement with SPIL. These arrangements build on the capacity expansion plans announced in September with Amkor, KYEC, and Unimicron.
Beyond supply chain issues, gross margin was also a major focus of this earnings report.
Previously, as high-bandwidth memory such as HBM rose in price, the market had been worried that Nvidia's cost pressure would continue to grow. After all, products like Blackwell are not simply GPU chip sales but involve memory, networking equipment, advanced packaging, and complete server systems. Cost increases in any one link could affect final profits.
But this quarter Nvidia's GAAP and non-GAAP gross margins reached 74.8% and 74.9% respectively, both above the previous guidance of around 74%, mainly benefiting from improved mix driven by Blackwell Ultra, as product structure optimization offset part of the supply chain cost pressure.
However, gross margin is not entirely free of pressure. Nvidia expects FY2027 Q4 gross margin to decline to 71%–72%, forming a cyclical bottom, before recovering to the 72%–73% range in FY2028.
This means the impact of rising memory and other supply chain costs will continue to show, and at least in the short term Nvidia will still need to pay some price for product iteration and supply expansion. But judging from management's assessment, this pressure is temporary, and the company does not see a trend of long-term gross margin erosion—though it still chose to give the market a preemptive warning.
Vera Rubin Nearly $20 Billion in First Quarter, CPU Revenue to More Than Double Next Fiscal Year
While the supply chain continues to expand, Nvidia's next-generation products have also begun contributing revenue—namely Vera Rubin.
This quarter, the Vera Rubin platform generated nearly $20 billion in revenue in its first quarter of mass shipment, making it the fastest-ramping product platform in Nvidia's history.
In terms of delivery progress, Vera Rubin NVL72 began mass shipments in August this year, and just three months later, related racks have already entered the data centers of major hyperscale cloud providers and Neoclouds. Nvidia revealed that it took hyperscale cloud providers about only one month from receiving Vera Rubin to putting it into production, half the time required for Grace Blackwell.
Vera Rubin is not just a GPU upgrade. With the rise of Agentic AI, AI systems need more CPUs to handle task scheduling, tool invocation, and data processing, and CPUs are becoming a new compute bottleneck. Nvidia said Agentic AI is driving CPU demand to significantly exceed supply, and the company's own Vera CPU has thus gained adoption from more customers.
This quarter, CoreWeave, Meta, Nscale, and OpenAI joined the Vera CPU customer lineup, after AWS, Oracle Cloud Infrastructure, and SpaceX AI had previously announced adoption of related products.Colette Kress expects Nvidia's CPU business revenue to more than double in FY2028.
Nvidia's networking business is also expanding in tandem. This quarter Nvidia's networking business revenue exceeded $21 billion, indicating that AI data center investment is no longer just about purchasing GPUs but requires simultaneous buildout of high-speed interconnect networks, server racks, and complete cluster systems.
Nvidia also introduced the DSX POD reference architecture, hoping to pre-integrate and validate entire AI factories and then deliver them to customers in modular form. According to management, this approach can shorten data center deployment cycles from months to weeks. For a market where compute demand far exceeds supply, shortening deployment time itself means generating revenue faster.
Looking at these businesses together, Nvidia's role as the "chief architect of the AI industry" is becoming increasingly evident. What it sells to customers is no longer just GPUs but AI infrastructure spanning GPUs, CPUs, networking, software, and complete data center architecture.
Nvidia is turning single-chip competition into competition over complete AI infrastructure, which is also why Nvidia is not overly worried about large model companies developing their own ASIC chips.
Will Nvidia No Longer Need Jensen Huang in the Future?
On November 17, Jensen Huang issued his first public statement since resigning on social media, saying that since his departure he has received calls from employees, investors, partners, and customers, and that if the board believes his return would be in the best interests of shareholders, employees, the company, and the United States, he is willing to return as CEO.
Hours later, board chairman Harvey Jones said on the call:The board unanimously accepted the resignation, has no intention of reinstating him, and has engaged Heidrick & Struggles to lead a formal CEO search, with both internal and external candidates under consideration, and will appoint after finding a suitable candidate.The board also newly established a National Security and Export Controls Oversight Committee, whose members include some independent directors, senior external advisors, and legal counsel.
Judging from the earnings report itself, cloud industry backlog is still increasing, existing compute is running at full capacity, supply chain expansion is progressing, Vera Rubin contributed nearly $20 billion in its first quarter, next-quarter guidance is $130 billion, and the FY2028 revenue growth guidance was raised from 70% to 75%. Everything seems to indicate that Nvidia's future is bright.
But the black swan event of Jensen Huang's departure has shifted some investors' attention regarding the FY2027 Q3 earnings report away from gross margin and supply chain bottlenecks toward corporate governance, which also increases uncertainty about Nvidia's future performance.
After all, the company's achievements to date have all been accomplished under Jensen Huang's leadership, and for the next quarter and FY2028 beyond his control, it is unknown whether his successor can continue supply chain-related cooperation and whether the company might change Nvidia's long-term strategic direction in order to be accountable for short-term stock prices.
Historically, the only case of a company of such scale maintaining sustained success after transitioning from founder mode to professional manager mode is Apple. And Nvidia's next important task is to find its own "Cook."


