On June 24 local time, the world’s most valuable public company, NVIDIA, held its 2026 Annual General Meeting. During the event, founder and CEO Jensen Huang dropped a bombshell answer regarding the ROI on AI investments:“It’s already figured out.”
In his keynote, Huang painted a vivid picture: AI data centers are no longer just warehouses for chips; they are “factories manufacturing tokens.” Think about it—tokens can be transformed into code, answers, designs, actions, and services. That means every single token is literally a unit of profit. Sure, NVIDIA systems might not be the cheapest to buy upfront, but they churn out the lowest-cost tokens per unit and boast the highest throughput—and consequently, the highest revenue. In plain English? Your customers aren’t just buying a rack of servers; they’re investing in an income-generating AI factory.
Huang explained that this logic is exactly why NVIDIA’s financials have been absolutely insane. For 2025, the company saw full-year revenue skyrocket by 65% to hit $216 billion, with operating income up 60% to $130 billion. Diluted earnings per share jumped 67% to $4.90. Operating cash flow reached $103 billion, and they handed back $41 billion to shareholders. To top it off, data center revenue surged 68% to $194 billion.
Huang made it clear: “Driven by our confidence in sustainable market growth and free cash generation, we plan to return 50% or more of our free cash flow to shareholders over the next few years, with stock buybacks and dividends set to keep climbing.”
This year, 2026, is shaping up to be a watershed moment for both NVIDIA and the entire computing industry. We are building the bedrock for a new era. Huang emphasized that AI isn’t just a model; it represents a fundamental shift in how we compute. For the past 60 years, computers were primarily used to retrieve, store, and send information. Now, thanks to AI, computing is being reinvented as “generative intelligence.”
According to Huang, the tech industry hits a reset button every 10 to 15 years: mainframes gave way to PCs, which led to the internet, then cloud, then mobile cloud. But this reset is bigger than all previous ones combined. With AI, computers can now understand, reason, plan, use tools, and actually get useful work done. They aren’t just tools anymore; in the AI age, they are intelligent assistants. Consequently, data centers are evolving from mere “tool sheds” into massive AI factories staffed by digital assistants—the true infrastructure of digital intelligence.
Two years ago, generative AI grabbed global attention with ChatGPT writing essays, drawing art, and answering questions. Then, reasoning AI learned to think through problems. Now, agent AI has arrived. These agents can wield tools, access memory, write code, call other agents, run tests, and keep working until the job is done.
“This wave of construction will be measured in decades, similar to building power grids, transportation networks, and the internet. We believe this will become the largest infrastructure project in human history,” Huang stated. Tokens are the basic unit of intelligence, manufactured in these new-style data centers—AI factories—and monetized through commercialization. More compute power equals more tokens, which equals more revenue.
As the engine for NVIDIA’s next growth phase, the Vera Rubin architecture is now in full-scale mass production. According to Huang, the new-generation AI chip, Blackwell, has significantly expanded NVIDIA’s footprint across diverse customer segments, including hyperscale cloud providers, cloud service providers, AI labs, industrial enterprises, and sovereign nations. Model developers and hyperscalers have already deployed hundreds of thousands of Blackwell GPUs. Meanwhile, AI factory construction is expanding rapidly across major industries. Companies like Capital One, Hyundai Motor Group, Jane Street, and Eli Lilly are all scaling up their NVIDIA infrastructure to deploy AI.
Huang revealed that this surge has pushed NVIDIA’s international revenue up by more than three times, exceeding $30 billion. Nearly 40 countries/regions, representing a combined $50 trillion GDP, are currently building AI factories powered by NVIDIA infrastructure.
“AI infrastructure is no longer experimental; it’s in production. AI isn’t just a model; it’s a new industry. Imagine it as a five-layer cake: Energy, Chips & Systems, Infrastructure, Models, and Applications. Traditional data centers stored and served files. AI factories manufacture tokens. Those tokens turn into code, answers, designs, actions, and services.” Huang explained.
NVIDIA drives AI factory construction in a unique way thanks to its full-stack, end-to-end co-designed infrastructure and a massive ecosystem of partners supporting these projects. “Agent AI is accelerating infrastructure investment because it’s the first time AI is truly doing real work and creating tangible economic value. As organizations seek to manufacture intelligence at scale, demand for AI factories will far exceed today’s cloud computing needs, extending to enterprises, sovereign nations, and regional AI clouds.”
Looking ahead, Huang believes the next major growth stage is Physical AI, and we’ve barely scratched the surface. Powering Physical AI requires a fresh round of infrastructure investment. Here, NVIDIA offers AI infrastructure, Omniverse simulation and digital twin platforms, open models, the Jetson Thor embedded computing platform, and software stacks for large-scale development and deployment of Physical AI. “Over time, AI will move from the digital world into robotaxis, humanoid robots, and industrial systems, enabling these machines to perceive, reason, and act autonomously in the physical world.”
Closing out the day on June 24 local time, NVIDIA shares closed at $199, down a tiny 0.52%, with a total market cap standing at a staggering $4.82 trillion.
