NVIDIA Accelerates Next-Gen 'Rubin' AI GPU Launch, Featuring Advanced 3nm Process and HBM4 Technology

Nvidia Blackwell chip successor "Rubin" moves forward by six months, uses 3nm manufacturing and HBM4

In brief: The immense demand for Nvidia's AI data center GPUs has sent the company into the stratosphere over the last few years, putting it in seesaw competition with Apple for the title of the most valuable firm on Earth. As Nvidia emerges from the production problems that delayed the Blackwell GPU launch, new reports indicate that its successor is significantly ahead of schedule. According to the Taiwan Economic Daily, Nvidia's Rubin series AI graphics chips will be available in the second half of 2025, half a year earlier than previously reported. Although details regarding performance improvements are slim, the new GPUs will feature more advanced memory and a newer semiconductor process node than Blackwell, which is set to launch soon. Prior roadmaps from Nvidia pegged the new architecture, named after American astronomer Vera Rubin, for release sometime in 2026. However, the chipmaker, manufacturing partner TSMC, and other supply chain participants have accelerated the lineup's development and implementation. A Morgan Stanley Semiconductor analyst told the Economic Daily that Rubin represents Nvidia's shift toward TSMC's 3nm N3 process node, which could significantly improve performance. The new chips will also upgrade to HBM4 memory, substantially boosting memory bandwidth. If the reports prove accurate, this could feel like a reversal of the delays that pushed Blackwell's implementation into the first quarter of 2025. The GB200 series, now entering mass production, encountered late-stage design flaws and currently suffers from overheating issues. Despite the setbacks, Blackwell chips are sold out through late 2025, as tech giants like Microsoft, Google, and Meta have spent billions to acquire them. The delays occurred partly due to challenges with TSMC's CoWoS-L packaging technology. The manufacturer is expected to significantly expand CoWoS production with Rubin, though its possible effects on the Blackwell successor's rollout remain unclear. The impact on GB300, previously called Blackwell Ultra, is another mystery. The upgraded lineup is set for a late 2025 debut, coinciding with Rubin's new release window. Prior reports indicate that Nvidia is considering moving to a socketed design for GB300, which might compromise performance in exchange for streamlining the production and upgrade pipelines. However, the lineup's introduction of FP4 is expected to boost demand considerably. Meanwhile, competitors like AMD and Intel are preparing to answer GB300 and Rubin with upcoming architectures like MI300 and Falcon Shores, also based on N3.

NVIDIA's next-gen Rubin AI GPU could be pushed up 6 months ahead of schedule with HBM4

NVIDIA's next-generation Rubin AI GPU architecture release rumored to be pulled up by 6 months, TSMC 3nm process expected, with ultra-fast next-gen HBM4 memory. The new Rubin AI GPU architecture is the successor to the Blackwell GPU architecture, which is being used in the current fleet of B200 and GB200 chips, as well as the future GB300 series AI GPU that we're hearing more and more about lately. In a new report from UDN, we're hearing that NVIDIA is already working with supply chain partners in Taiwan on the Rubin AI GPU architecture and its new R100-powered AI servers. Rubin was originally scheduled for 2026, but sources of UDN say that the company has launched the development of Rubin early, so that the AI boom can continue from one AI GPU chip to another (Blackwell to Rubin, and so on). TSMC is a key partner of the triangular alliance between NVIDIA + SK hynix + TSMC, with the company expected to expand its CoWoS advanced packaging capacity in 2026 to handle the large Rubin chip demand. TSMC plans to increase CoWoS production capacity to 80,000 pieces per month by Q4 2025 in preparation of Rubin.

NVIDIA's Next-Gen "Rubin" Architecture Is Now Rumored To Be Released By "Six Months" Ahead of Schedule

NVIDIA's "next-gen" Rubin architecture is now rumored to be available "six months" ahead of schedule as Team Green also looks to capture the upcoming AI markets. There's no stopping NVIDIA now, given that the firm is committed more than ever to maintaining dominance in the AI markets. Even though the Blackwell architecture has only seen a few weeks of the official launch, the industry is already talking about the next-gen Rubin AI lineup, which shows that NVIDIA is indeed more determined than ever to impact the industry. In a report by Taiwan Economic Daily, Team Green is now said to be releasing the Rubin architecture ahead of schedule, accelerating cadence between architectures. The Rubin lineup, initially scheduled for 2026, has now been pushed back to mid-2025. This gives NVIDIA the opportunity to advance much more quickly than its competitors, ultimately allowing the firm to maintain its dominance over the markets. For those unaware, NVIDIA is currently in a "one-year" cadence plan, which means that the company will release a new generation after each year, which is why Ampere, Hopper, and Blackwell are all separated by year intervals; however, Rubin, this might change. Reports don't mention a specific reason for why NVIDIA's Rubin will be launched early on, apart from categorizing it as a business move. However, if we look at the supply chain, Rubin is expected to feature TSMC's 3nm process, along with the cutting-edge HBM4 memory chips. Both of these components are either in their mass production stages or about to enter them. Given that NVIDIA has all the "tools" for Rubin available, the company may not feel it right to release Rubin in 2026. Previously, we reported on the debut of NVIDIA's "Blackwell Ultra" lineup or the "B300" series, where it was claimed that Team Green plans to release it by mid-2025. So, by this, we might expect Blackwell Ultra and Rubin to drop in similar timelines, although the release strategy isn't specific for now since, with Rubin, NVIDIA might potentially overshadow the B300 series. We expect further updates from the company soon, probably around CES 2025.

Nvidia Next gen Rubin GPU Release based on Enhanced TSMC 3nm Process and HBM4 Technology

Nvidia has moved up the release date for its Rubin GPU, now scheduled to launch in the latter half of 2025, six months earlier than the previously planned 2026. The Rubin architecture, succeeding Nvidia's Blackwell series, is engineered to provide advanced artificial intelligence (AI) computing performance. This GPU will feature Taiwan Semiconductor Manufacturing Company's (TSMC) latest 3-nanometer (nm) process technology and High Bandwidth Memory 4 (HBM4), which are expected to significantly improve computational efficiency and memory bandwidth necessary for complex AI tasks. To support the early launch, Nvidia is working closely with supply chain partners, including TSMC, to accelerate the development of the Rubin platform. This collaboration aims to meet the increasing demand for high-performance AI hardware and help Nvidia maintain its leadership position in the AI market. The use of TSMC's 3nm process technology is set to enhance the GPU's performance and energy efficiency, while the integration of HBM4 memory will provide the high data transfer rates required for demanding AI applications. In anticipation of the higher production requirements for the Rubin GPU, TSMC is expanding its Chip on Wafer on Substrate (CoWoS) advanced packaging facilities. The company plans to increase its monthly production capacity to 80,000 units by the fourth quarter of 2025. Source: ithome