Nvidia expects to earn at least $1 trillion from its Blackwell and Rubin AI chips

Nvidia has sharply raised expectations for its AI business. At the GTC conference in San Jose, CEO Jensen Huang said the Blackwell and Rubin lines should bring Nvidia at least $1 trillion by the end of 2027.

A trillion as the benchmark

For a semiconductor maker, this is not just a big number but a public claim on the scale of the entire AI infrastructure market. Huang announced the forecast from the stage at GTC, Nvidia's main annual event, where the company typically presents its roadmap for data centers, accelerators and the software around them. If the target is met, it would be one of the largest revenue streams ever tied to a single line of computing platforms.

The time horizon matters too: Nvidia is not talking about a distant future, but about the period through the end of 2027. For the industry, such statements serve as a guide to future purchasing, supply-chain utilization and the pace of compute expansion.

A bet on two platforms

The statement named two products at once: Blackwell and Rubin. That matters because Nvidia sells the market not a single chip but a sequence of generations, around which customers build budgets and data center expansion plans. Blackwell is the immediate engine of the new wave of AI systems; Rubin is the next major step on the same roadmap.

When the CEO ties both platforms into one forecast, he is effectively telling investors and customers that demand should survive a generational change rather than end with one successful launch. In other words, Nvidia is packaging the current AI frenzy into a longer commercial story.

• Blackwell is presented as Nvidia's nearest sales driver
• Rubin is already included in the company's long-term expectations
• The forecast horizon is not an abstract future but the end of 2027
• Nvidia is betting on a multi-year cycle of AI infrastructure purchases

This approach differs from short-term hype around individual GPUs. Nvidia is showing that it sees lasting demand not only from experimental AI teams but also from large buyers willing to plan infrastructure several years ahead. For the ecosystem, that means the race for compute capacity, memory, networking and energy continues, because accelerators alone cannot cover a volume like this. It takes a whole stack of servers, networking equipment, cooling and software.

A signal to the whole market

Huang's statement matters not only to Nvidia shareholders. It sets the tone for the entire supply chain: server makers, cloud platforms, data center operators and enterprise customers budgeting for generative AI. The more confidently Nvidia talks about a trillion dollars, the more the market builds further growth in AI capital spending into its models.

At the same time, it raises the bar: if demand slows, each subsequent quarterly report will be compared not just with a strong baseline but with a publicly stated, highly ambitious target.

In effect, Nvidia is trying to secure a role not just as a beneficiary of the AI boom but as the company around which the whole computing economy for modern models is built. That adds pressure on competitors, who must respond not only on performance but also on the scale of available supply, the ecosystem and the speed of new product releases. Against this backdrop, even cautious corporate customers get a signal that the window for slow adoption is narrowing. If the largest infrastructure supplier is this confident about demand, putting off a rebuild of AI strategy becomes harder.

What it means

Jensen Huang's forecast is a bet that AI infrastructure investment will remain enormous for at least two more years. If Nvidia is right, the market will see not a brief spike but a long data center refresh cycle, in which winners include not only chip makers but the whole layer of services, clouds and enterprise AI products around them.

For business, it is one more signal that the question is no longer whether there will be a compute shortage, but who will manage to claim a place in this new value chain.