Nvidia Corp CEO Jensen Huang (黃仁勳) unexpectedly unveiled the company’s next-generation graphics processing unit (GPU) architecture, codenamed Rubin, to be available in 2026, during a speech in Taipei on Sunday.
To go with the Rubin GPUs, Nvidia also plans to introduce a new-generation CPU called Vera in 2026, Huang said at the National Taiwan University Sports Center.
The company plans to launch an enhanced version of the upcoming Rubin GPUs, called Rubin Ultra, in 2027, he said.
Photo: Annabelle Chih, Bloomberg
The company’s Blackwell GPU architecture, announced in March to succeed the Hopper framework, are to hit the market this year followed by the Blackwell Ultra units next year, he said.
Rubin GPUs are likely to be built using Taiwan Semiconductor Manufacturing Co’s (TSMC, 台積電) most advanced 3-nanometer process technology, an upgrade from the 4-nanometer process technology used to produce the Blackwell B100 accelerators, TF International Securities Group Co (天風國際證券) analyst Kuo Ming-chi (郭明錤) wrote on X.
Nvidia would likely also adopt TSMC’s chip-on-wafer-on-substrate advanced packaging technology for the new Rubin GPUs, as it would with the B100s, Kuo said.
The Rubin GPUs would likely enter mass production in the fourth quarter of next year, he said.
“Our company is in a one-year rhythm,” Huang said. “On a one-year rhythm and we push everything, the technology limits whatever TSMC process technology pushes it to the absolute limits. Whoever’s packing technology pushes it to absolute limits. Whatever memory technology pushes it to absolute limits.”
During his speech, Huang also talked about digital humans, artificial intelligence (AI) PCs, robotics and industrial automation, and expected a larger swath of companies and government agencies to embrace AI.
In the future, "almost every interaction you have with the Internet or with a computer will likely have a generative AI running in the cloud somewhere," he said.
But as the amount of data that needs to be processed grows exponentially in the AI era, traditional computing methods cannot keep up and it’s only through Nvidia’s style of accelerated computing that firms can cut back the costs, Huang said.
He therefore outlined how some top tech companies such as Taiwan's Foxconn Technology Group (富士康科技集團) — the world's largest contract electronics manufacturer — and German industrial giant Siemens AG are using Nvidia's platforms to develop AI-powered autonomous robots.
In addition, he announced that Asustek Computer Inc (華碩), Gigabyte Technology Co (技嘉), Ingrasys Inc (鴻佰), Inventec Corp (英業達), Wiwynn Corp (緯穎), Pegatron Corp (和碩), Cloud Technology Inc (雲達) and Super Micro Computer Inc would deliver cloud, on-site, embedded and edge AI systems using Nvidia GPUs and networking chips.
“Taiwan is the home of our [Nvidia’s] treasured partners. This is, in fact, where everything Nvidia does begins. Our partners and ourselves take it to the world. Taiwan and our partnership has created the world’s AI infrastructure,” Huang said.
Additional reporting by agencies
The New Taiwan dollar is on the verge of overtaking the yuan as Asia’s best carry-trade target given its lower risk of interest-rate and currency volatility. A strategy of borrowing the New Taiwan dollar to invest in higher-yielding alternatives has generated the second-highest return over the past month among Asian currencies behind the yuan, based on the Sharpe ratio that measures risk-adjusted relative returns. The New Taiwan dollar may soon replace its Chinese peer as the region’s favored carry trade tool, analysts say, citing Beijing’s efforts to support the yuan that can create wild swings in borrowing costs. In contrast,
Nvidia Corp’s demand for advanced packaging from Taiwan Semiconductor Manufacturing Co (TSMC, 台積電) remains strong though the kind of technology it needs is changing, Nvidia CEO Jensen Huang (黃仁勳) said yesterday, after he was asked whether the company was cutting orders. Nvidia’s most advanced artificial intelligence (AI) chip, Blackwell, consists of multiple chips glued together using a complex chip-on-wafer-on-substrate (CoWoS) advanced packaging technology offered by TSMC, Nvidia’s main contract chipmaker. “As we move into Blackwell, we will use largely CoWoS-L. Of course, we’re still manufacturing Hopper, and Hopper will use CowoS-S. We will also transition the CoWoS-S capacity to CoWos-L,” Huang said
Nvidia Corp CEO Jensen Huang (黃仁勳) is expected to miss the inauguration of US president-elect Donald Trump on Monday, bucking a trend among high-profile US technology leaders. Huang is visiting East Asia this week, as he typically does around the time of the Lunar New Year, a person familiar with the situation said. He has never previously attended a US presidential inauguration, said the person, who asked not to be identified, because the plans have not been announced. That makes Nvidia an exception among the most valuable technology companies, most of which are sending cofounders or CEOs to the event. That includes
INDUSTRY LEADER: TSMC aims to continue outperforming the industry’s growth and makes 2025 another strong growth year, chairman and CEO C.C. Wei says Taiwan Semiconductor Manufacturing Co (TSMC, 台積電), a major chip supplier to Nvidia Corp and Apple Inc, yesterday said it aims to grow revenue by about 25 percent this year, driven by robust demand for artificial intelligence (AI) chips. That means TSMC would continue to outpace the foundry industry’s 10 percent annual growth this year based on the chipmaker’s estimate. The chipmaker expects revenue from AI-related chips to double this year, extending a three-fold increase last year. The growth would quicken over the next five years at a compound annual growth rate of 45 percent, fueled by strong demand for the high-performance computing
RSS