Nvidia Chief Executive Jensen Huang said the company aims to sell twice as many chips next year as it does this year, citing expanding investment in artificial intelligence.
Huang made the comment to reporters on the 17th at the Scotland AI Summit, hosted by Britain's King Charles III at Dumfries House in Cumnock, Scotland, according to CNBC and other outlets. "I expect we will sell twice as many chips next year as this year," he said.
"AI is contributing enormously to many industries and economies," he said. "In nearly every country we operate in, people want to invest in AI."
The doubling Huang referred to is measured by the number of semiconductors sold, not by revenue. Last October, he said the company had shipped 6 million Blackwell graphics processing units to that point. Nvidia supplies not only data center GPUs but also central processing units, network switches, optical communication chips and Jetson chips for robots and vehicles.
In its second-quarter earnings announcement last month, covering May through July, Nvidia projected that fiscal 2028 revenue would rise about 70% from the previous year. Applying market estimates, that would put revenue at $673 billion, or about 930 trillion won. The company said customer demand forecasts point to greater growth potential next year, but that it expects revenue growth of around 70% given supply constraints.
Huang partly distanced himself from arguments that AI development should be slowed or subjected to broad new regulation on safety grounds. "AI is different from social media," he said. "Social media is a product, but AI is the technology behind every product and every technology, so what should be regulated is products, not AI technology." Still, he stressed that companies bear responsibility for rigorous pre-release testing, saying, "If a product is not safe enough, you should delay its launch and keep developing it."
Huang's remarks have drawn attention to how quickly Samsung Electronics and SK hynix will expand memory output. For Nvidia to increase supply of AI accelerators, it needs more than the high-bandwidth memory mounted on GPUs — it also requires greater supply of low-power DRAM for CPUs as well as general-purpose server DRAM and NAND flash.
With memory chips in short supply and prices continuing to climb, revenue and profit at Samsung Electronics and SK hynix are expected to grow further if their output expansion plans materialize.
Samsung Electronics is accelerating capacity expansion centered on its Pyeongtaek campus. It is currently building P5-1 and P5-2, and the industry expects each fab to reach monthly capacity of 300,000 wafers on a 300-millimeter basis, or 600,000 combined. SK hynix is expanding capacity around two hubs, Cheongju and Yongin. After bringing online M15X in Cheongju, which produces HBM, it moved up the start of the first clean room at Yongin's Y1 to February next year, more than six months earlier than planned.
