Nvidia chief executive Jensen Huang has repeated his expectation that the company can increase revenue by 70% next year. With analysts projecting about $400 billion for the current fiscal year, that rate would imply roughly $680 billion in the following year.
Huang presented the case at Goldman Sachs’ Communacopia and Technology conference. His forecast rests on Nvidia’s position across AI laboratories, cloud providers, data-centre builders and component suppliers. It is management guidance, not a guaranteed result, and arrives as competitors develop their own accelerators.
Nvidia Now Sells Complete Computing Systems
Huang argued that investors still misunderstand Nvidia as a chip vendor rooted in gaming graphics. He described a modern high-end system linked with NVLink as an $8.5 million product containing roughly 2 million parts and requiring 250,000 kilowatts. TechCrunch reports his figures without independently detailing the configuration or power measurement.
One system combining 36 Grace CPUs with 72 Blackwell GPUs is experiencing 27% month-to-month order growth, according to Huang. He did not provide the order base, customer mix or duration behind that rate, so it cannot be projected indefinitely.
The larger point is that Nvidia sells integrated infrastructure rather than isolated processors. That broadens the value of each order and ties the company to networking, memory, power and data-centre construction. It also makes revenue dependent on customers completing unusually large and capital-intensive projects.
Partners Give Huang a View of Demand
Huang said Nvidia can support models from Anthropic, OpenAI and Google as well as open-weight systems. He described the company as a foundational platform that works with laboratories, original-equipment manufacturers, cloud services, specialized AI clouds and startups.
Those relationships give Nvidia information about planned projects. Huang said the company tracks land, power availability and unfinished data-centre buildings worldwide through reports from partners. He uses that visibility to support confidence in future demand, although the source does not show the underlying pipeline or how Nvidia discounts delayed and cancelled projects.
Competition is growing from major cloud companies designing chips, AI laboratories developing their own hardware, newly public Cerebras and accelerator startups such as Etched. Nvidia’s compatibility across customers may protect demand in the near term, but buyers also have a financial incentive to reduce dependence on one supplier.
Investment Deals Add a Circularity Debate
Nvidia has invested in companies that subsequently purchase its systems, prompting questions about whether capital is circulating back as sales. Huang rejected that characterization, arguing that Nvidia invests a relatively small amount and receives much larger orders. He said companies must already have genuine customer contracts before Nvidia provides capital and claimed to have reviewed $100 billion of such contracts.
The source does not disclose those agreements, investment terms or the share of Nvidia revenue connected with portfolio companies. Customer contracts may reduce risk without eliminating it: a startup’s ability to pay still depends on financing, customer retention and the economics of AI services.
Huang also acknowledged that much current growth comes from AI-native startups raising large sums and spending heavily on computation. As the sector matures, customers may become more efficient in their use of infrastructure and model tokens. That could weaken the assumption that today’s spending intensity continues.
The 70% outlook combines real recent growth with an exceptionally bullish view of project visibility. Nvidia’s reach across the AI supply chain gives management unusually broad demand signals, but it also concentrates exposure to the same buildout. Reaching approximately $680 billion would require customers to finance, power and deploy systems at a scale that remains subject to competition, construction schedules, energy availability and changing economics. The forecast will ultimately be tested by completed sales rather than the breadth of Nvidia’s partner conversations.
