Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/rapidus-faces-three-major-challenges-in-2nm-chip-production.21687/
        )

    [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
)

Rapidus faces three major challenges in 2nm chip production

Daniel Nenni

Admin
Staff member
1_b.jpg

Rapidus chairman Tetsuro Higashi. Credit: SEMI Japan

According to Nikkei, Rapidus chairman Tetsuro Higashi stressed in an interview that achieving the goal of producing 2nm chips involves overcoming three significant challenges- the technical feasibility of mass production, market and customer positioning, and funding issues, particularly regarding its heavy reliance on resources from the Japanese government.

Technical feasibility of mass production​

As Japanese manufacturers have only reached mass production on processes down to 40nm, Rapidus's direct pursuit of 2nm technology is typically seen as a reckless endeavor.

Higashi believes that since semiconductor equipment manufacturers have long been involved in advanced semiconductor technologies, and Rapidus is collaborating with institutions like IBM to integrate technology and experience, they have confidence in achieving mass production of 2nm chips.

Higashi had also considered entering the mid-tier semiconductor manufacturing market but concluded that existing companies with completed depreciation would create unfavorable conditions for new entrants. In contrast, the advanced market offers more equitable competitive conditions.

Market and customer positioning​

From the outset, Rapidus has positioned itself as an alternative supplier to TSMC, aiming to differentiate itself by creating a market for specialized semiconductors, unlike larger firms that excel in producing large-scale and standardized products.

During Semicon Japan 2024, Higashi discussed how standard chips waste excessive energy, arguing that semiconductor products should shift from standardized offerings to specialized chips tailored for specific applications. Through technological innovation, future semiconductor products could reduce energy consumption to one-fifth of current levels.

These specialized areas include robotics, autonomous driving, and telemedicine. Higashi anticipates that the generative AI market will pivot towards optimized dedicated chips for various algorithms, leading to dramatic changes in the market environment.

Funding issues​

Due to private investors perceiving high risks regarding Rapidus's ability to actually produce advanced chips, the company currently relies heavily on financial support from the Japanese government.

Rapidus estimates that reaching the stage of mass production for 2nm chips will require JPY5 trillion (approx. US$32 billion). Currently, its primary source of funding comes from government subsidies over three consecutive fiscal years, totaling JPY920 billion.

The Japanese prime minister and other officials have repeatedly stressed that the government plans to allocate over JPY10 trillion to support the development of the semiconductor and AI industries by the fiscal year 2030 to stimulate private investment, with support for Rapidus being a focal point of this policy.

However, if Rapidus continues to rely solely on government support, it may effectively become a state-owned enterprise. Concerns persist within the Japanese industry and media about state-owned semiconductor companies potentially becoming costly ventures with limited effectiveness.

Higashi stressed that while Rapidus currently depends mainly on government support, the ultimate goal is to achieve independence.

In the coming years, Rapidus aims to increase the proportion of private funding, seeking to finance factory construction through bank loans or capital increases from private shareholders, targeting approximately half of the equipment investment costs to come from private sources.

 
I certainly hope Rapidus is a viable foundry alternative at 2nm but I have concerns. Anytime IBM is involved I have concerns but that is my bias. Synopsys and Rapidus are working together on IP which is a good sign:

 
By "standard chips" I assume Rapidus is referring to GPUs and CPUs, and the energy savers are specialized ASICs with state machine logic. That strategy played out for bitcoin mining, a small number of networking protocol chips (I'm thinking of the Nvidia (Mellanox) InfiniBand/Ethernet offerings), but many so-called custom ASICs include embedded processor cores doing a lot of the work. I'm not convinced the Rapidus vision is practical. State machine ASICs are one hell of a design and testing/debug challenge.
 
1_b.jpg

Rapidus chairman Tetsuro Higashi. Credit: SEMI Japan

According to Nikkei, Rapidus chairman Tetsuro Higashi stressed in an interview that achieving the goal of producing 2nm chips involves overcoming three significant challenges- the technical feasibility of mass production, market and customer positioning, and funding issues, particularly regarding its heavy reliance on resources from the Japanese government.

Technical feasibility of mass production​

As Japanese manufacturers have only reached mass production on processes down to 40nm, Rapidus's direct pursuit of 2nm technology is typically seen as a reckless endeavor.

Higashi believes that since semiconductor equipment manufacturers have long been involved in advanced semiconductor technologies, and Rapidus is collaborating with institutions like IBM to integrate technology and experience, they have confidence in achieving mass production of 2nm chips.

Higashi had also considered entering the mid-tier semiconductor manufacturing market but concluded that existing companies with completed depreciation would create unfavorable conditions for new entrants. In contrast, the advanced market offers more equitable competitive conditions.

Market and customer positioning​

From the outset, Rapidus has positioned itself as an alternative supplier to TSMC, aiming to differentiate itself by creating a market for specialized semiconductors, unlike larger firms that excel in producing large-scale and standardized products.

During Semicon Japan 2024, Higashi discussed how standard chips waste excessive energy, arguing that semiconductor products should shift from standardized offerings to specialized chips tailored for specific applications. Through technological innovation, future semiconductor products could reduce energy consumption to one-fifth of current levels.

These specialized areas include robotics, autonomous driving, and telemedicine. Higashi anticipates that the generative AI market will pivot towards optimized dedicated chips for various algorithms, leading to dramatic changes in the market environment.

Funding issues​

Due to private investors perceiving high risks regarding Rapidus's ability to actually produce advanced chips, the company currently relies heavily on financial support from the Japanese government.

Rapidus estimates that reaching the stage of mass production for 2nm chips will require JPY5 trillion (approx. US$32 billion). Currently, its primary source of funding comes from government subsidies over three consecutive fiscal years, totaling JPY920 billion.

The Japanese prime minister and other officials have repeatedly stressed that the government plans to allocate over JPY10 trillion to support the development of the semiconductor and AI industries by the fiscal year 2030 to stimulate private investment, with support for Rapidus being a focal point of this policy.

However, if Rapidus continues to rely solely on government support, it may effectively become a state-owned enterprise. Concerns persist within the Japanese industry and media about state-owned semiconductor companies potentially becoming costly ventures with limited effectiveness.

Higashi stressed that while Rapidus currently depends mainly on government support, the ultimate goal is to achieve independence.

In the coming years, Rapidus aims to increase the proportion of private funding, seeking to finance factory construction through bank loans or capital increases from private shareholders, targeting approximately half of the equipment investment costs to come from private sources.

The biggest issues IMO are not having experience as a foundry or company and getting a technology from a company without a Fab. The "fast cycle time" is another comment showing an idealistic plan.

Lets see the output from the pilot line in spring 2025 as promised I would expect revised plans and strategy to be announced by the end of 2025

The Powerpoint foils for a new foundry are easy. Its the pesky Fabs and billions in financing that make it hard LOL
 
The biggest issues IMO are not having experience as a foundry or company and getting a technology from a company without a Fab. The "fast cycle time" is another comment showing an idealistic plan.

Lets see the output from the pilot line in spring 2025 as promised I would expect revised plans and strategy to be announced by the end of 2025

The Powerpoint foils for a new foundry are easy. Its the pesky Fabs and billions in financing that make it hard LOL

Rapidus' major technology development partners are IBM, IMEC, Tokyo Electron, and Japan's Leading-edged Semiconductor Technology Center (LSTC). They all have "labs" but no commercial "fab".
 
Rapidus' major technology development partners are IBM, IMEC, Tokyo Electron, and Japan's Leading-edged Semiconductor Technology Center (LSTC). They all have "labs" but no commercial "fab".
A list of people who don't do manufacturing.... not good ....

unless just perhaps, Rapidus won't do HVM either.... hmm
 
A list of people who don't do manufacturing.... not good ....

unless just perhaps, Rapidus won't do HVM either.... hmm

There are some speculations about Japanese government's Plan B for Rapidus. A wildest one is to ask TSMC to take over the Rapidus' fab because either way Japanese government is giving money out. Although I think it's very expensive to reconfigure Rapidus' fab design to TSMC's.
 
Rapidus' major technology development partners are IBM, IMEC, Tokyo Electron, and Japan's Leading-edged Semiconductor Technology Center (LSTC). They all have "labs" but no commercial "fab".

Similar to other reshoring fab projects, hopefully the Japan market will fill Rapidus fabs and get them profitable. I still think working with TSMC would be a better option for leading edge fabs. Japan might be hedging their bets with a second TSMC FinFET fab which could easily be N3 if Japan is serious about HPC/AI applications and Rapidus fails to yield.

 
Similar to other reshoring fab projects, hopefully the Japan market will fill Rapidus fabs and get them profitable. I still think working with TSMC would be a better option for leading edge fabs. Japan might be hedging their bets with a second TSMC FinFET fab which could easily be N3 if Japan is serious about HPC/AI applications and Rapidus fails to yield.


According to this article, $32 billion is needed to get Rapidus' 2nm production up and running. Without multiple large-volume customers, Rapidus' financial viability remains questionable.
 
$32B sounds about right to me. And less than it will cost to bail out Intel.

I wonder at the defense strategic planning of governments. When it costs nothing, they don't care of about semiconductor competitiveness. When it costs $10s of billions, suddenly they care. I think it has to do with how governments work--once the figure is large enough it's useful, political pork.
 
Back
Top