Nvidia CEO Huang: AI tokens now profitable, signaling shift to inference-driven revenue
In brief
- Jensen Huang: AI tokens now profitable, spurring model makers to boost output
- Nvidia data center revenue hit $75.2B, up 92% YoY from AI infrastructure demand
- AI inference flipped from cost center to revenue-generating business line
- Huang frames shift as emerging 'AI factory' economy driven by compute capacity
Tokens Flipped From Cost to Revenue
Nvidia reported quarterly revenue of $81.6 billion for Q1 FY2027, an 85% increase compared to the same period a year earlier. The data center segment brought in $75.2 billion, a 92% year-over-year increase, reflecting explosive demand for GPU infrastructure powering AI workloads globally.
Huang described the current demand environment as having "gone parabolic." This surge stems from a fundamental economic shift. Model makers are racing to produce more tokens because the unit economics have flipped—tokens now generate revenue rather than drain it.
"Tokens are now profitable. So model makers are in a race to produce more." — Jensen Huang, Nvidia CEO
The AI Factory Economy
Huang has framed this pivot as part of a broader shift toward what he calls the "AI factory" economy. In this model, compute capacity itself becomes a form of revenue generation. Companies no longer view inference as a necessary overhead; instead, they're building business models around token production and consumption.
At Nvidia's GTC event in March 2026, Huang suggested that companies could allocate AI token budgets to their engineers, potentially worth about half of an engineer's salary. This framing signals how deeply token economics are embedding themselves into corporate planning.
Huang also declared during the earnings call that agentic AI has arrived, underscoring his conviction that autonomous AI systems will drive the next wave of compute demand. The shift toward profitable tokens and agentic systems creates a self-reinforcing cycle: more agents require more inference, which generates more token revenue, which justifies further investment in GPU infrastructure.
It's worth noting that the tokens Huang discussed are purely computational units that live on GPU clusters, not on-chain. They carry no connection to blockchain technology or cryptocurrency.
Frequently asked questions
What does Huang mean by 'tokens' being profitable?
Huang refers to AI output tokens—units of text, code, or reasoning that language models generate. These are computational units living on GPU clusters, not blockchain tokens. He means that producing these tokens now generates revenue for companies, rather than being purely a cost of business.
How does the AI factory economy work?
In the AI factory model, compute capacity itself becomes a revenue-generating asset. Companies build business models around token production and consumption. Instead of viewing inference as overhead, they monetize the outputs their AI systems generate, creating a cycle where token revenue justifies further investment in infrastructure.
Why is Nvidia's data center revenue growing so fast?
Nvidia's data center segment grew 92% year-over-year to $75.2 billion, driven by explosive global demand for GPU infrastructure powering AI workloads. As companies race to build AI systems and monetize token production, they're investing heavily in the compute infrastructure Nvidia provides.
