At its annual GPU Technology Conference in San Jose, California, the semiconductor giant outlined an ambitious outlook for the artificial intelligence industry, estimating that the revenue opportunity for AI chips could reach at least $1 trillion by 2027. The projection highlights Nvidia’s growing focus on the rapidly expanding segment of AI known as inference computing, where systems generate answers and perform tasks in real time.
The announcement was delivered by Nvidia’s chief executive Jensen Huang during a keynote address at the four-day developer conference, which has evolved into one of the technology sector’s largest showcases for artificial intelligence innovation.
Speaking to an audience of more than 18,000 attendees inside a packed arena, Huang declared that the next phase of AI development is already underway.
“The inference inflection has arrived,” he said, signalling what he described as a shift from experimental AI development to large-scale deployment.
From Training AI to Running It in Real Time
For much of the past decade, Nvidia has dominated the market for chips used to train artificial intelligence models, largely through its powerful graphics processing units (GPUs). These processors helped companies such as OpenAI, Anthropic and Meta Platforms build the advanced AI systems now used by hundreds of millions of people.
But as those systems move from research labs into everyday products, the focus of the industry is shifting toward inference computing — the stage where trained AI models respond to user requests, generate content, and make decisions in real time.
That transition is opening the door to new competition for Nvidia. Several large technology companies are designing their own custom AI chips to run inference workloads more efficiently, including Meta and other cloud providers.
To maintain its leadership, Nvidia used the conference to reveal new hardware and partnerships designed to improve the speed and efficiency of real-time AI processing.
New Chips and Strategic Partnerships
Among the announcements was a new AI computing architecture built around Nvidia’s upcoming Vera Rubin processors. According to the company, the chips will handle the first stage of inference known as “prefill,” where a user’s request is converted from human language into digital tokens that AI systems can interpret.
The second stage — known as “decode,” where the AI system generates its response — will be powered by technology from startup Groq. Nvidia licensed Groq’s technology in a deal valued at roughly $17 billion last December.
Groq specializes in chips optimized for high-speed inference tasks, allowing AI systems to deliver answers faster and at lower cost. The collaboration reflects Nvidia’s broader effort to adapt its technology for the next phase of AI deployment.
Expanding the AI Roadmap
During the keynote, Huang also previewed another next-generation AI processor known as Feynman, named after renowned physicist Richard Feynman. The chip is expected to form part of Nvidia’s long-term roadmap for increasingly powerful AI infrastructure.
Beyond hardware, Huang emphasized the importance of Nvidia’s software ecosystem — particularly its widely used programming platform CUDA. Many analysts view the software environment as one of Nvidia’s strongest competitive advantages because it enables developers to build AI applications directly on the company’s chips.
“The installed base is what attracts developers who then create new algorithms that achieve breakthrough technologies,” Huang told the audience. “We are in every cloud. We're in every computer company. We serve just about every single industry.”
Market Reaction and Analyst Views
Investors responded cautiously to the company’s long-term projection. Nvidia shares briefly rose after Huang referenced the $1 trillion opportunity before ending the trading session up about 1.6%.
The estimate marks a major increase from the roughly $500 billion AI chip market opportunity projected for 2026 during the company’s previous earnings call.
According to Jacob Bourne, an analyst at Emarketer, the forecast signals Nvidia’s confidence that demand for AI infrastructure will remain strong despite growing competition.
“Huang mapping out a $1 trillion opportunity through 2027 underscores the durable demand for Nvidia’s AI infrastructure,” Bourne said. “It suggests the industry is moving beyond early experimentation into large-scale deployment.”
AI Infrastructure Becoming a Global Priority
The expansion of AI infrastructure is also becoming a strategic priority for governments and businesses around the world. Several countries are investing heavily in national AI computing capabilities powered by Nvidia’s chips, including major projects in the Middle East.
At the same time, the company continues to release open-source AI tools, positioning itself as a central platform in the broader global competition for artificial intelligence leadership — particularly between the United States and China.
Despite the rise of rival chips and custom hardware developed by some of its own customers, Nvidia remains deeply embedded in the global AI ecosystem. As companies shift from building AI models to delivering them at scale, the battle to power real-time artificial intelligence is likely to define the next stage of the technology race.