Nvidia Expects 70% Growth Next Year, Huang Explains

Nvidia founder and CEO Jensen Huang spoke at the Goldman Sachs Communacopia + Technology conference held on September 10, 2026. He reiterated his outlook that revenue next year, corresponding to fiscal 2027, could increase 70% year over year. According to reporting by Julie Bort of TechCrunch AI, he expressed strong confidence in sustaining record growth. It was a repeat of the outlook first presented at the August earnings announcement. Some in the market have also voiced concerns about a slowdown in the company’s growth. In response, he cited high visibility into demand.

The company’s revenue for the current fiscal year is analyzed to reach about $400 billion. If 70% growth is achieved, next fiscal year would amount to about $680 billion. The focus is on how long demand for AI computing infrastructure will last. He emphasized that his company is embedded throughout the ecosystem. He made clear his position that it is not merely a semiconductor supplier.

Reaffirms 70% Revenue Growth Outlook for Next Year

Huang again referred to next year’s revenue growth outlook at the conference. He said the company could grow 70% year over year. He added that he is confident it can be achieved. This outlook was first disclosed at the quarterly earnings in August. At that time, the company had also posted record quarterly revenue. Attention focused on the fact that it expects acceleration, not a slowdown.

He explained that current sales trends are also strong. He was referring to a system combining 36 Grace CPUs and 72 Blackwell GPUs. He said sales of this product continue to grow 27% month over month. A single product is sustaining double-digit monthly growth. It shows that capital investment in AI infrastructure remains strong. He argued there are no signs of a short-term pullback in demand.

“I think we could grow 70% year over year. We’re confident about that,”

This remark symbolizes a bullish business outlook. It is seen as aimed at dispelling investor anxiety. Expansion of the AI infrastructure market is a precondition. Whether supply constraints can be resolved holds the key.

The Reality of an $8.5 Million GPU

Huang said misunderstandings about his company’s products persist. He said many people think Nvidia just builds semiconductors. But he explained that current products are different from conventional consumer graphics devices. He described them as systems so large they require airplanes to ship.

“Most people think Nvidia builds a chip. I mean, you need airplanes to ship what we build,”

According to him, current GPUs reach $8.5 million per unit. They are connected by NVLink and comprise 2 million parts. He said power consumption reaches 250,000 kilowatts. They are huge computing systems shipped at a scale of thousands of units. The structure is different from the era when it sold $399 products. The company supplies the entire computer together with the interconnect network.

This argument is also a rebuttal to criticism of high gross margins. His position is that the product is not a standalone chip but a distributed computing infrastructure. It is offered in a form that includes everything from development to deployment support. The aim is to justify the price by the scale of the system. Customers pay for computing capacity and uptime. He argued that comparing the cost of a single component does not reflect reality.

Its endpoint product strategy is also moving forward in parallel. As reported in Nvidia to Roll Out RTX Spark Long-Term, N2X and N3X Already Planned, it is also pursuing the continued rollout of small-scale computing machines. It aims to capture the market with both large-scale systems and small-scale devices. The aim is seen as winning over developers. Lock-in across the entire platform is progressing.

The Supply-Chain View That Captures All Demand

Huang argued he can foresee future demand. He cited as the reason the depth of its integration into the entire ecosystem. He said his company runs every model. That includes models from Anthropic, OpenAI, and Google. He said open-weight models also run on his company’s platform.

“Nvidia runs every model. Every single lab can use us,”

He positioned his company as the foundational infrastructure of the AI ecosystem. He expressed the view that it underpins the entire AI industry. On the supply side, it works with memory manufacturers. On the demand side, he said it tracks everything from data center projects to startups. He said it tracks land, power, and buildings around the world. By buildings, he meant data center buildings before computers are installed.

He explained that reports gather from its partner network. That includes information from emerging cloud operators and equipment manufacturers. He said large cloud providers and AI-focused companies are also reporting sources. He argued that because it works with every layer, it can see the full picture. Backlog orders and projects under construction feed into demand forecasts. Visibility across the entire supply chain underpins the bullish outlook.

Competition over infrastructure is intensifying in other regions as well. Agentic Infra Chosen by Four Companies, Infinigence’s Strategy reports on moves by different players to select infrastructure. Moves to avoid dependence on a specific vendor are also spreading. Even so, consolidation around Nvidia continues for now. High visibility itself has become a competitive advantage.

Rebuttal to Circular-Dealing Criticism and

the Competitive Landscape

Huang’s remarks raised questions about circular deals. Circular deals refer to a structure in which companies Nvidia has invested in purchase its products. There is criticism that revenue is inflated without real demand. He stressed his position that these are not such deals. He expressed the view that sales are based on real demand. He did not go into detailed financial breakdowns.

The competitive environment is becoming tougher. Large cloud providers Amazon, Microsoft, and Google are advancing in-house development. AI labs Anthropic and OpenAI are also exploring their own chips. Startups such as the newly listed Cerebras and Etched are also closing in. It faces pressure from every direction. Even so, he argued its near-term advantage remains unshaken.

The openness of instruction sets and software stacks is also an issue. Tenstorrent Releases GCC Patch for RISC-V Core ‘Ascalon XG’ shows the progress of the camp built around open instruction sets. Nvidia is countering with proprietary interconnect technology and software platforms. High switching costs encourage customers to keep using it. Competition extends beyond chip performance to the entire ecosystem. Accumulated developer assets have become a barrier to entry.

Sustainability and Challenges of Data

Center Demand

Capital investment in AI data centers continues on a global scale. Securing power has become the biggest constraint. Land, power, and buildings are all needed at the same time. Huang said he tracks every gigawatt-scale project. He used the expression that he literally knows every project on Earth. The remark flaunts its supply-chain management capabilities.

The challenge lies in the quality and sustainability of demand. If capital investment becomes excessive, a correction will come. Constraints on power grids and cooling water are also becoming apparent. Geopolitical fragmentation could also affect supply chains. Tighter export controls limit sales regions. Even so, he expressed confidence in demand through next year.

The impact on engineers is also significant. As optimization for a specific vendor’s platform advances, switching costs rise. At the same time, open models and compatibility layers are being developed. Procurement departments seek to secure multiple supply sources. Operations departments demand improved power efficiency and uptime. Evaluating total cost of ownership is important in product selection.

Editorial Opinion

In terms of short-term impact, we see continued capital investment as the focus over the next three to six months. We assess that Huang’s 70% revenue-growth remark will have the effect of encouraging orders from customers and suppliers. Supply and demand for memory devices, power equipment, and cooling systems could tighten further. We see rivals’ custom chips remaining at the prototype stage, with no shift to mass production.

From a long-term perspective, we see the question over one to three years as whether infrastructure diversification will advance. We assess in-house development by large cloud providers and AI labs as moves aimed more at cost reduction than performance. If open instruction sets and compatible software layers mature, dependence on a specific vendor will decline. If power constraints persist, compute efficiency will become the axis of market evaluation.

As a question from the editorial team, we believe the transparency of demand visibility is in question. While the breadth of its reporting network is a strength, can outsiders verify the distinction between real demand and demand from investees? Is disclosure sufficient to support the claim that these are not circular deals? By what indicators should investors and engineers judge real demand?

References

• “Jensen Huang explains why Nvidia will grow an astounding 70% next year”, by Julie Bort — TechCrunch AI, 2026-09-10T21:51:59.000Z (ARR)
• Source URL: https://techcrunch.com/2026/09/10/jensen-huang-explains-why-nvidia-will-grow-an-astounding-70-next-year/