Jensen Huang has been making the case that artificial intelligence represents a multitrillion-dollar industry, one with what he describes as “very clear and good return on investment.”

The scale of the bet

Huang has projected somewhere between $3 trillion and $4 trillion in global AI infrastructure spending by the end of the decade. During his CES 2025 keynote on January 7, Huang framed AI agents as a “multi-trillion-dollar opportunity,” with agentic AI at its center.

Agentic AI can plan, reason, and execute multi-step tasks on its own, as distinct from traditional AI that simply responds to prompts.

Nvidia’s position in the food chain

By March 2026, Huang reported a $1 trillion order backlog for AI-related solutions. The company is projecting at least $1 trillion in cumulative revenue from AI chips alone through 2027.

Huang has also identified robotics as a critical growth area alongside agentic AI. Nvidia’s Omniverse and Isaac platforms provide simulation environments where robotic systems can be trained before deployment.

The competitive landscape is heating up

As of September 2025, Huang identified OpenAI as being on track to become the next “multi-trillion-dollar hyperscale company.”

AMD continues to push its MI300 series chips into AI workloads. Google’s custom TPUs power its own AI infrastructure. Amazon has Trainium.

Nvidia’s moat, however, isn’t just hardware. It’s CUDA, the software ecosystem that developers have been building on for over a decade.

What this means for investors

A $1 trillion order backlog and a projected $3-4 trillion infrastructure buildout define the demand side of the equation. Huang’s emphasis on “very clear and good return on investment” addresses the concern that companies could spend trillions on AI infrastructure without seeing proportional productivity gains.

Google, Amazon, and Microsoft are all investing heavily in their own AI chips as alternatives to Nvidia’s hardware, though none have matched Nvidia’s performance yet.