Jensen Huang says Nvidia can grow revenue 70% next year as AI infrastructure demand expands

Nvidia CEO Jensen Huang renewed his forecast that the company’s revenue could grow by 70% next year, arguing that Nvidia’s position across the AI supply chain gives it unusual visibility into future demand.

Speaking at the Goldman Sachs Communacopia + Technology conference on September 10, Huang said Nvidia’s role now extends well beyond selling individual processors. The company supplies systems, networking, software, and infrastructure used by AI labs, cloud providers, startups, and data-center operators. That breadth, he argued, supports continued growth even as major customers and competitors develop their own AI chips.

The forecast is a company-level expectation repeated by Huang, not an independently verified projection. TechCrunch reported that analysts expect Nvidia to finish its current fiscal year with roughly $400 billion in revenue; a 70% increase would imply approximately $680 billion the following year.

Nvidia’s bet on full-stack AI infrastructure

Huang’s central argument is that Nvidia should not be evaluated as a conventional chip supplier. He described modern Nvidia systems as large-scale computing platforms that combine processors, networking, memory, cooling, and other components. In his example, one system can include 36 Grace CPUs and 72 Blackwell GPUs connected through NVLink.

The scale of those systems helps explain Nvidia’s expanding average sale value. Huang said one integrated GPU system can cost about $8.5 million and contain roughly two million parts. He also said orders for the Grace-and-Blackwell computer system were growing 27% month over month, although that figure came from the CEO and was not supported in the source material by an independent sales breakdown.

For AI builders and enterprise buyers, the distinction matters. Demand is increasingly shifting from purchasing isolated accelerators to deploying complete clusters capable of training and serving models. That raises the value of integration and performance across the entire system, but it also increases customer dependence on one vendor’s hardware, software stack, and supply chain.

Why Huang expects demand to persist

Huang said Nvidia can see activity across nearly every layer of the AI market. He cited relationships with model developers including Anthropic, OpenAI, and Google, as well as customers using open-weight models. He also pointed to suppliers, original equipment manufacturers, cloud companies, so-called neoclouds, and AI-native startups.

According to Huang, Nvidia tracks data-center projects by monitoring land, available power, and partially constructed facilities around the world. That information, combined with reports from partners, gives the company a view of planned computing capacity before it appears in quarterly revenue.

This is the foundation of his bullish case: if more AI companies are raising capital, reserving power, building data centers, and training increasingly capable models, Nvidia expects to supply much of the equipment required to operate them. The company’s exposure is therefore spread across many customers and applications, even though the underlying demand remains concentrated in AI infrastructure.

The argument also reflects a competitive reality. Amazon, Microsoft, and Google are developing their own AI chips, while Anthropic and OpenAI are reportedly pursuing custom hardware efforts. Cerebras and startups such as Etched are targeting parts of the accelerator market. Those initiatives could reduce reliance on Nvidia over time, but Huang’s position is that Nvidia’s broad platform and established ecosystem remain difficult to replace quickly.

Evidence, claims, and the circular-deal question

The strongest growth and demand claims in this story are executive statements reported by TechCrunch. Nvidia’s 70% revenue outlook was first provided when the company reported another record revenue quarter, according to the report, and Huang repeated it at the Goldman Sachs event. The article did not provide a new formal financial filing or a customer-by-customer order schedule validating the forecast.

Huang also faced questions about Nvidia investments in companies that later purchase Nvidia products. Critics have described these arrangements as circular deals because capital supplied by Nvidia can help fund customers that then spend on Nvidia’s hardware. The structure has drawn comparisons with earlier technology infrastructure cycles in which suppliers supported demand among their own customers.

Huang rejected that characterization. He said Nvidia invests only after confirming that a company has genuine customer contracts and claimed to have seen approximately $100 billion in such contracts. He did not provide the source material with a detailed list of those contracts, their timing, or the portion attributable to Nvidia revenue.

That makes the issue important for investors and enterprise observers. Equity investments can strengthen an ecosystem and help promising AI companies secure infrastructure, but they can also make reported demand harder to interpret. Buyers and analysts will need to distinguish end-customer usage from spending supported by financing relationships.

What the forecast means for AI builders and enterprises

If Huang’s projection proves accurate, AI builders will likely continue facing a market in which access to advanced computing is as important as model design. Startups may need to secure capacity earlier, commit to longer infrastructure contracts, or work through cloud and neocloud providers rather than buying systems directly.

For enterprises, Nvidia’s expansion could bring more mature deployment options, including integrated systems based on Blackwell, Grace, and NVLink. It could also reinforce switching costs. Teams that optimize models, monitoring systems, and software for Nvidia’s platform may find it expensive to move to competing AI chips, even when alternatives offer lower prices or better performance for specific workloads.

The main risk is efficiency. TechCrunch noted that much of the current expansion is driven by AI-native companies raising large sums and spending heavily on their own AI usage. As those businesses mature, they may improve model efficiency, reduce token consumption, or consolidate workloads. If demand grows more slowly than the amount of infrastructure being built, Nvidia’s current visibility may not translate into the same level of revenue growth.

What to watch next

The next meaningful signals will come from Nvidia’s formal earnings guidance, data-center revenue, and commentary on order visibility. Investors should also watch whether the company’s reported growth is driven by new deployments or by a small number of exceptionally large customers.

Competitive evidence will matter as well. The adoption of custom chips from hyperscalers and AI labs, the performance of Cerebras and Etched, and the availability of software that makes alternative accelerators easier to deploy could test Nvidia’s platform advantage.

Finally, scrutiny of Nvidia’s investments should intensify. Clear disclosures about deal structures, customer contracts, and revenue linked to invested companies would help distinguish durable market demand from ecosystem financing.

Creati.ai perspective

Huang’s 70% forecast is best understood as a statement about Nvidia’s breadth rather than a guarantee that accelerator demand will remain unchanged. The company is positioned across chips, systems, networking, software, and infrastructure planning, giving it more ways to participate as AI spending expands.

But that breadth also makes the forecast harder to audit from the outside. The key question for builders and buyers is not whether Nvidia is deeply embedded in AI today; the evidence supports that conclusion. It is whether future workloads will require enough new capacity, at Nvidia-level prices, to sustain the same pace after customers become more efficient and competing hardware improves.