Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/nvidias-rubin-its-next-gen-gpu-faces-possible-delay-due-to-redesign-fubon.23390/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Nvidia's Rubin, its next-gen GPU, faces possible delay due to redesign: Fubon

Daniel Nenni

Admin
Staff member
1755105196705.png


Nvidia's (NASDAQ:NVDA) Rubin, its next-generation graphics processing unit, might face a delay in its production ramp at TSMC (NYSE:TSM) due to a redesign, according to Fubon Financial, a Taiwanese financial services firm.

"We think it is very likely that Rubin will be delayed," said Fubon analyst Sherman Shang, in a research note. "The first version of Rubin was already taped out in late June, but Nvidia is now redesigning the chip to better match AMD's (NASDAQ:AMD) upcoming MI450."

"We think the next tape out schedule will be in late September or October, and based on the tape out schedule, the Rubin volume will be limited in 2026," Shang added.

A tapeout in semiconductor manufacturing refers to the final stage of an integrated circuit design. The design is verified and finalized before being sent to the foundry for fabrication.

The Rubin GPU will be the successor to Nvidia's Blackwell models, which continue to ramp up, according to Moore Morris, an analyst at Nomad Semi. In a post on X, Morris said Blackwell volumes hit 750,000 in the first quarter of 2025, 1.2M during the second quarter, and will reach 1.5M and 1.6M in the third and fourth quarters, respectively.

Morris also said that AMD and Broadcom (AVGO) are currently the fastest-growing customers for CoWoS, or Chip-on-Wafer-on-Substrate, for TSMC. However, Nvidia still dominates capacity at 51.4% in 2025 compared to Broadcom at 16.2% and AMD at 7.7%. However, Broadcom and AMD's capacity allocation is expected to reach 17.4% and 9.2%, respectively, in 2026, while Nvidia dips slightly to 50.1%.

Seeking Alpha reached out to Nvidia regarding the potential delay.

According to multiple reports, Rubin was slated for mass production in late 2025 and available for purchase in early 2026.

 
@tomatoma @Daniel Nenni

Help me out with clarifications.
Is Rubin the first HBM4 product? it certainly is the major HBM4 Product.
It was planned for 2H 2026. Now it may be more like 1H 2027. correct?
Micron this week repeatedly referred to "our customer product timelines will determine HBM4 ramp" .... seems like they know its delayed.

This has implications on 2026 volume and market share IMO.
 
@tomatoma @Daniel Nenni

Help me out with clarifications.
Is Rubin the first HBM4 product? it certainly is the major HBM4 Product.
It was planned for 2H 2026. Now it may be more like 1H 2027. correct?
Micron this week repeatedly referred to "our customer product timelines will determine HBM4 ramp" .... seems like they know its delayed.

This has implications on 2026 volume and market share IMO.
Per wikipedia, The Rubin chips will be manufactured by TSMC using a 3 nm process and will use HBM4 memory. It is scheduled for mass production in late 2025 and will be available for purchase in early 2026.[1][2]

I don't quite understand how NVDA can still command such a high valuation if these rumors are true — its next flagship product has already been matched (or even surpassed) by competitors, so much so that it needs to revise its design just 3–6 months ahead of its scheduled release.

Btw, NVDA market cap is 15x of AMD, and 45x of INTC. I am long both $amd and $intc.
 
Per wikipedia, The Rubin chips will be manufactured by TSMC using a 3 nm process and will use HBM4 memory. It is scheduled for mass production in late 2025 and will be available for purchase in early 2026.[1][2]

I don't quite understand how NVDA can still command such a high valuation if these rumors are true — its next flagship product has already been matched (or even surpassed) by competitors, so much so that it needs to revise its design just 3–6 months ahead of its scheduled release.

Btw, NVDA market cap is 15x of AMD, and 45x of INTC. I am long both $amd and $intc.
Smart, Kevin.

When you see this....

""The first version of Rubin was already taped out in late June, but Nvidia is now redesigning the chip to better match AMD's (NASDAQ:AMD) upcoming MI450."

....IF it's accurate, AMD may quite possibly be ahead of NVDA in technology/performance. Now, we all know they are WAY behind in market share, eco-system, customer base, etc., etc. But this is still a good sign for AMD.
 
Well as someone said Software and Ecosystem matters

SemiAnalysis does a pretty good job looking at software and ecosystem for cloud-scale inference. They did this assessment of AMD back in June. My take is that chip and even cluster-level specs are somewhat meaningless, except for individual GPU memory capacity (since fitting a model plus operating memory into a single GPU can make a huge difference).

AMD Advancing AI: MI350X and MI400 UALoE72, MI500 UAL256
Software Improvement, Marketing RDFs, AMD Fostering Neocloud, MI355 is not Rack Scale, MI400 is UALoE, Not UALink​


 
Back
Top