Jensen Huang Predicts Nvidia Will Grow 70% Next Year as AI Boom Continues
TL;DR
- Jensen Huang says Nvidia can grow roughly 70% next year as Blackwell and Rubin AI chips sell out alongside a rebound in gaming and enterprise AI adoption.
- Huang detailed an "every pie" strategy to capture spending across cloud data centers, sovereign AI factories, robotics, gaming, and enterprise software.
- He forcefully rejected claims of circular deals with OpenAI and other startups, calling Nvidia's investments normal, open-market demand for computing.
The 70% Forecast That Shook Wall Street
Jensen Huang is not tempering expectations. Speaking this week ahead of Nvidia's next earnings cycle, the Nvidia CEO laid out why he expects the chipmaker to grow an astounding 70% next year, a pace almost unheard of for a company already generating well over $200 billion in annual revenue.
Huang's argument is simple: AI demand is not slowing, it is broadening. Cloud giants are still racing to build capacity, sovereign nations are now building their own AI factories, and a second wave from enterprises, gamers, and robotics developers is just starting to hit.
If he is right, Nvidia would add more than $140 billion in new annual sales in a single year, cementing its position as the central supplier of the AI era.
Why Data Centers Are Still Starving For GPUs
The bulk of that growth, Huang said, will still come from data centers. Despite two years of massive purchases of Hopper and Blackwell systems, hyperscalers like Microsoft, Amazon, Google and Meta remain supply-constrained.
Huang pointed to three drivers. First, the shift from training to reasoning. New reasoning models from OpenAI, Anthropic, Google DeepMind and xAI require 10 to 100 times more compute per query than older chatbots, forcing data centers to massively expand inference capacity.
Second, the Blackwell Ultra ramp and the upcoming Rubin platform. Huang said Blackwell is fully sold out and Rubin, paired with the Vera CPU and next-generation networking, is already seeing record pre-orders for late 2026 and 2027 deployments.
Third, the $500 billion Stargate project with OpenAI, Oracle and SoftBank, plus sovereign deals in Saudi Arabia, the UAE, Japan, the UK and France. Huang said sovereign AI alone could be a multi-hundred-billion-dollar market as countries treat computing as national infrastructure.
Gaming And Enterprise Join The Party
While data centers get the headlines, Huang stressed that gaming and enterprise are set to become major growth engines again next year.
On gaming, Nvidia's GeForce RTX 50-series, launched earlier this year, is seeing its strongest adoption in years as gamers upgrade for AI-powered DLSS 4, neural rendering and on-device AI assistants. Huang said gaming revenue, which had stagnated during the post-crypto slump, is now reaccelerating.
On enterprise, the story is agentic AI. With Nvidia NIM microservices, NeMo, Omniverse and its RTX-powered workstations, Huang said thousands of companies in finance, healthcare, manufacturing and retail are moving from AI pilots to full production. He described 2026 as the year of enterprise agents and 2027 as the year they scale across every department.
Combined, Huang argued, gaming plus professional visualization plus automotive and robotics could add tens of billions in new sales on top of data center growth.
Inside The Eat Every Pie Strategy
At the heart of Huang's pitch is what analysts are calling his every pie strategy. Rather than picking one winner in AI, Nvidia wants a slice of every pie.
That means selling H100, Blackwell and Rubin GPUs to every cloud, building custom silicon with Broadcom rivals through NVLink Fusion, supplying RTX chips for PCs and workstations, providing Drive Thor for cars and robotaxis, Jetson Thor for humanoid robots, and licensing CUDA, Omniverse and DGX Cloud software on top.
Huang told investors: We don't need to pick the winning model, the winning startup, or the winning country. Every model needs computing. Every country needs infrastructure. Wherever computing happens, we want to be there.
It is an expensive strategy that has Nvidia investing aggressively in fabs with TSMC and Samsung, expanding networking with Spectrum-X and InfiniBand, and pouring billions into the U.S. supply chain, but Huang says it insulates Nvidia if any single segment cools.
Fighting Back Against Circular Deal Claims
Huang also used the appearance to mount his strongest defense yet against claims of circular deals on Wall Street.
Critics have argued Nvidia is artificially inflating demand by investing billions in customers like OpenAI, Anthropic, xAI, CoreWeave and Nebius, who then turn around and buy Nvidia chips. Nvidia's planned investment of up to $100 billion in OpenAI and multi-billion dollar stakes in Intel and other partners have fueled those concerns and drawn scrutiny from regulators.
Huang dismissed the idea as fundamentally misunderstanding computing. He said Nvidia does not require anyone to buy its chips, does not offer discounts tied to investments, and invests for the same reason other tech giants do: to accelerate the ecosystem.
Nobody is forced to buy anything, Huang said. These are great companies building the future. They need computing, and we are the world's best supplier of computing. If we invest, it is because we believe in them, not because we need to create demand. Demand is already overwhelming.
He added that the vast majority of Nvidia's revenue comes from long-standing, cash-rich hyperscalers paying full price, not from startups it has funded.
What Comes Next For Nvidia
Wall Street remains divided. Bulls say Huang's 70% target is credible given backlog, Rubin pricing power, and inference demand. Bears warn of supply chain bottlenecks, export controls on China, power constraints, and the sheer law of large numbers.
For now, Huang is betting that the AI boom is still in its second inning. With Rubin on deck for 2026, Rubin Ultra for 2027, a revived gaming cycle, and AI factories spreading from Texas to Tokyo to Riyadh, he sees no sign of a slowdown.
As he put it: We are at the start of a new industrial revolution. Computing is becoming intelligence. And intelligence will be built, like electricity before it, everywhere.
Get All The Latest Updates Delivered Straight To Your Inbox For Free!