Nvidia Rubin Delayed? Implications

Earlier today, Barrons and other sources published articles about an explosive note from Fubon Financial of Taiwan about Nvidia (NVDA) having to go to drawing board with its next generation Rubin GPU due to competitive pressure from Advanced Micro Devices (AMD) MI450. Some choice quotes from Fubon:

“The first version of Rubin was already taped out in late June, but Nvidia is now redesigning the chip to better match AMD’s upcoming MI450.”

“We think the next tape out schedule will be in late September of October, and based on the tapeout schedule, the Rubin volume will be limited in 2026.”

“AMD and Broadcom will be the fastest growing CoWoS customers for TSMC. AMD will use CoWoS for both CPU and GPU in 2026. Broadcom will benefit from multiple projects’ production schedule.”

Nvidia spokesperson refuted the story by saying “Rubin is on track.”

We cannot be 100% sure that the Fubon report is true without further verification, but the report should be given a special consideration since it is coming Fubon, a very strong financial name from Taiwan – well plugged into the supply chain. Also, a leak with this much potential impact will not get through the research house without rigorous checking.

Nvidia’s response of saying “Rubin is on track” should also be taken with a pinch of salt. Note that Nvidia did not acknowledge and even denied Blackwell delays until it was obvious. To this day, three quarters after the originally anticipated launch time, Blackwell GB200 yields are considered subpar, and companies have struggled to ramp GB200. As a result of GB200 problems, and from its learning lessons, GB300 plans were also changed leading to a delay to that program.

Not just Nvidia, rarely do semiconductor companies admit to program slippage until it gets to a stage where it cannot be denied and even then, there will be some kind of spin on the delay. Beyond The Hype’s experience in the semiconductor industry is that programs area always on track – many times a new track with different schedules and goals.

Based on how Nvidia has been scrambling with Blackwell as it makes efforts to stay ahead of AMD suggests that the chances of the leak being true are high. Of course, reporters and other supply chain analysts will chase the rumor and will have their own takes in the days and weeks to come.

Over the next several days and possibly weeks AMD and Nvidia stocks could be volatile. Investors need to be prepared.

Note: This article uses MI400 to refer to the entire family of MI400 class products including MI450 referred to by Fubon.

AMD Is Out Executing Nvidia

What the market does not seem to be comprehending yet is that AMD as an execution machine is in a very different plane than its peers including Intel (INTC) and Nvidia. Consider the timelines of launch (first revenue) of each generation of AMD and Nvidia data center GPU products.

Nvidia generational gap, as can be seen below, is about 2 years.

- H100: Q3 FY23 (or the quarter ending October 2022)

- B200: Q3 FY25 (or the quarter ending October 2024)

- Rubin: Q3 FY27? (or the quarter ending October 2026)

Note that this does not count the mid-life refreshes.

- H200: Q2 FY25 (or the quarter ending July 2024)

- B300: Est Q3 FY26 (or the quarter ending October 2025)

On the other hand, AMD’s generation gap has been reducing.

- MI300: Q4 CY23 (or the quarter ending December 2023)

- Mi350: Q2 CY25 (or the quarter ending June 2025)

- MI4XX: Q2 CY26 (or the quarter ending June 2026)

That is a gap of 1.5 years between the current and previous generations but only 1 year between the current and next generations!

AMD also had two mid-life refreshes planned.

- MI325: Q3 CY2024 (or the quarter ending September 2024)

- MI375: Cancelled – likely because of small gap between MI350 and MI400

In other words, AMD roadmap pace has accelerated while Nvidia’s has stagnated despite great hype about “one year cadence”. There is a company that is delivering on a one-year cadence and that is not Nvidia.