Nvidia CEO Reiterates Solid Partnership with TSMC

[TheLostSwede]

One key takeaway from the ongoing GTC is that Nvidia's AI empire has taken shape with strong partnerships from TSMC and other Taiwanese makers, such as those major server ODMs.

According to the news report from the technology-focused media DIGITIMES Asia, during his keynote at GTC on March 18, Huang underscored his company's partnerships with TSMC, as well as the supply chain in Taiwan. Speaking to the press later, Huang said Nvidia will have a very strong demand for CoWoS, the advanced packaging services TSMC offers.




He said the company is embarking on a new journey and will work closely with TSMC. He said he believes TSMC's capacity will continue growing.

A major reason for the shortages of Nvidia's AI chips is insufficient CoWoS support from TSMC. To serve Nvidia and other AI chip customers, TSMC is accelerating the expansion of CoWoS capacity and has just confirmed that it will build two advanced packaging fabs in Chiayi, southern Taiwan.

Nvidia's AI foundry service
Nvidia is taking inspiration from TSMC's business model. Huang admitted that the way Nvidia is working for its partners and the AI industry is like the way TSMC is making chips for Nvidia.

Nvidia is learning from TSMC's business model where the foundry only handles the manufacturing. Nvidia is working in the same as an AI foundry, according to Huang.

In November 2023, Nvidia introduced an AI foundry service to customize generative AI applications for enterprises and startups deploying on Microsoft Azure. Sources from Taiwan's server supply chain noted that Nvidia has been the biggest winner in the AI boom with its prowess from the CUDA and GPU platforms.

Taiwanese firms have played a key role in helping build Nvidia's AI empire. At the same time, the Taiwanese supply chain also has been a major beneficiary of Nvidia's rise to dominance.

Nvidia is expected to introduce upgrades to its H100 and GH100 in the second quarter of 2024 - the H200 and GH200. It has also unveiled the B100/B200 and GB200 on its new Blackwell platform at GTC.

The launch of new chip products will benefit not only their contract maker TSMC and equipment suppliers but also the server supply chain.

TSMC is sole maker of Nvidia AI GPUs
The sources said TSMC has been the sole maker of Nvidia's AI GPUs. Nvidia had Samsung make its RTX 30 series at the 8 nm node for the gaming market.

However, yield and performance issues at Samsung later prompted Nvidia to let TSMC make all of its next-generation RTX 40 series at the 4 nm node, the sources said. The upcoming RTX 50 series will still be made by TSMC using 3/4 nm processes, sources added.

The Blackwell GPUs will also be made by TSMC. Its next-generation AI GPUs to be introduced in 2025 are expected to embrace 3 nm processes, the sources said.

As switching chipmaking partners will incur very high costs, Nvidia will continue using TSMC's 2 nm, A14, and A10 services in the coming years, barring unexpected developments. Nvidia's server partners in Taiwan, including ASRockRack, Asustek, Foxconn, Gigabyte, Inventec, Pegatron, Quanta Cloud Technology (GCT), Wistron, and Wiwynn, will supply servers running on the Blackwell platform.

View at TechPowerUp Main Site | Source

 

[Daven]

I don't think this is a press release. It reads like a news article. Did Nvidia or TSMC publish this as a press release? And the source link just goes to Digitimes Asia main page.

 

[TheLostSwede]

I don't think this is a press release. It reads like a news article. Did Nvidia or TSMC publish this as a press release? And the source link just goes to Digitimes Asia main page.

Digitimes did.

 

[Daven]

Digitimes did.

Ok so this isn’t a press release. I thought it would be strange for Nvidia or TSMC to submit an official statement worded like the article above.

You might want to consider removing the Press Release tag.

Edit: BTW, here is the direct link to the article (paywalled). Your source link above just goes to the main homepage.

Nvidia CEO reiterates solid partnership with TSMC

One key takeaway from the ongoing GTC is that Nvidia's AI empire has taken shape with strong partnerships from TSMC and other Taiwanese makers, such as those major server ODMs.

www.digitimes.com

 

[Guwapo77]

TSMC would be fools not to have a solid and robust relationship with one of the world's richest companies.

 

[TheLostSwede]

Ok so this isn’t a press release. I thought it would be strange for Nvidia or TSMC to submit an official statement worded like the article above.

You might want to consider removing the Press Release tag.

Edit: BTW, here is the direct link to the article (paywalled). Your source link above just goes to the main homepage.

Nvidia CEO reiterates solid partnership with TSMC

One key takeaway from the ongoing GTC is that Nvidia's AI empire has taken shape with strong partnerships from TSMC and other Taiwanese makers, such as those major server ODMs.

www.digitimes.com

It was sent out as a press release.

 

[Vya Domus]

Well duh, all of these companies are nothing without TSMC.

 

[AnotherReader]

TSMC would be fools not to have a solid and robust relationship with one of the world's richest companies.

Don't be fooled by market capitalization; TSMC is the more important of those two companies. In any case, unlike Nvidia, there haven't been many complaints about TSMC from customers. The sole customer to badmouth TSMC publicly is Nvidia.

 

[Wirko]

Well duh, all of these companies are nothing without TSMC.

It goes both ways. How would TSMC fund their R&D without, say, their top two or three customers?

 

[evernessince]

It goes both ways. How would TSMC fund their R&D without, say, their top two or three customers?

TSMC was completely fine when Nvidia went to Samsung. Nvidia was forced to come crawling back after they realized the 4000 series was going to be 600w+ on Samsung's process.

Nvidia's words here mean nothing, they've already backstabbed TSMC and will do it again in a heartbeat if the business makes sense.

 

[phints]

This seems like damage control after the articles from a few days ago saying Blackwell will be on a similar lithography to Lovelace (TSMC 4N or 4P?). Pressumably limiting performance per watt gains. Time will tell but I'm sure it'll still be a solid RTX 5000 generation, just maybe not the large jump from 3000 to 4000.

 

[bonehead123]

they've already backstabbed TSMC and will do it again in a heartbeat if the business makes sense

And so is the way of corporate greed-mongering....

People don't call them "Ngreediya" for nothin

 

[evernessince]

This seems like damage control after the articles from a few days ago saying Blackwell will be on a similar lithography to Lovelace (TSMC 4N or 4P?). Pressumably limiting performance per watt gains. Time will tell but I'm sure it'll still be a solid RTX 5000 generation, just maybe not the large jump from 3000 to 4000.

Makes sense, with how big Nvidia's die sizes are I don't see them working well on TSMC 3nm with it's yields only targeting 80% by the end of 2024. I can only hope that AMD is able to jump on 3nm with it's smaller chiplets, that would given them an edge in terms of efficiency and transistors per square mm.

 

[Guwapo77]

Don't be fooled by market capitalization; TSMC is the more important of those two companies. In any case, unlike Nvidia, there haven't been many complaints about TSMC from customers. The sole customer to badmouth TSMC publicly is Nvidia.

Don't be fooled into thinking TSMC doesn't want Nvidia's money, they want/need each other. There is no company on TSMC's level at this time; however, Nvidia could go elsewhere like they did before on Samsung's inferior node and Nvidia will still come out on top.

 

[AnotherReader]

Don't be fooled into thinking TSMC doesn't want Nvidia's money, they want/need each other. There is no company on TSMC's level at this time; however, Nvidia could go elsewhere like they did before on Samsung's inferior node and Nvidia will still come out on top.

Both of them need each other, but of the two, TSMC is more indispensable. Nvidia went with Samsung and only came out on top because AMD didn't build as big a chip as GA102. Navi 21, sans the 128 MB of LLC, would have been closer to AD104 in die size. One can imagine that a bigger Navi 21 would have taken the crown from Nvidia and Nvidia knows that. This is why they tucked their tail between their legs and came back to TSMC. Don't forget that even at that time, GA100, Ampere's datacenter focused GPU, was built by TSMC.

 

[evernessince]

Don't be fooled into thinking TSMC doesn't want Nvidia's money, they want/need each other. There is no company on TSMC's level at this time; however, Nvidia could go elsewhere like they did before on Samsung's inferior node and Nvidia will still come out on top.

Except it's not just AMD vs Nvidia anymore. Nvidia is competing against every company that wants to make AI chips which includes all the big tech companies and a whole host of startups. In addition, just in the GPU space AMD isn't that far behind them and Intel has joined the fray. Nvidia's threats of pulling AI chip allocation have held companies off from making public statements in regards to AI hardware but make no mistake many of them absolutely have plans in motion.

I don't see how Nvidia leaving TSMC hurts TSMC, litterally every AI hardware manufacturer is chomping at the bit for CoWoS packaging. AMD at the very least would suck up as much as it could for it's MI300/MI300X and future accelerators.

 

[Guwapo77]

Both of them need each other, but of the two, TSMC is more indispensable. Nvidia went with Samsung and only came out on top because AMD didn't build as big a chip as GA102. Navi 21, sans the 128 MB of LLC, would have been closer to AD104 in die size. One can imagine that a bigger Navi 21 would have taken the crown from Nvidia and Nvidia knows that. This is why they tucked their tail between their legs and came back to TSMC. Don't forget that even at that time, GA100, Ampere's datacenter focused GPU, was built by TSMC.

You're basically saying what I said... Yes, TSMC wants that money. Nvidia has choices as they are not using TSMC's most advanced node and AMD still doesn't have anything to compete.

 

[AnotherReader]

You're basically saying what I said... Yes, TSMC wants that money. Nvidia has choices as they are not using TSMC's most advanced node and AMD still doesn't have anything to compete.

End the ban on Huawei and other Chinese companies and TSMC wouldn't even need Nvidia. AMD is using a node slightly inferior to what Nvidia is using and as usual, their top end GPU die is smaller than Nvidia's top end. Given how well RDNA 3 scales with increasing CU count, a 120 CU RDNA 3 SKU would have been uncomfortably close to the 4090. However, it would have still lost in ray traced games.

 

[Guwapo77]

End the ban on Huawei and other Chinese companies and TSMC wouldn't even need Nvidia. AMD is using a node slightly inferior to what Nvidia is using and as usual, their top end GPU die is smaller than Nvidia's top end. Given how well RDNA 3 scales with increasing CU count, a 120 CU RDNA 3 SKU would have been uncomfortably close to Nvidia. However, it would have still lost in ray traced games.

Let's not expand the topic at hand sir. I'm not downplaying TSMC, but Nvidia money is endless at this point. AMD is a lot of coulda, woulda, shoulda... I've supported AMD going way before Nvidia was making motherboards for AMD but I'm not blind to the facts here... AMD is still playing catch up. I thought they had a chance in the AI space, but with the launch of the latest chip and AMD isn't supplying MI300X with any remarkable impact... Nvidia is straight teabaggin' the competition.

 

[AnotherReader]

Let's not expand the topic at hand sir. I'm not downplaying TSMC, but Nvidia money is endless at this point. AMD is a lot of coulda, woulda, shoulda... I've supported AMD going way before Nvidia was making motherboards for AMD but I'm not blind to the facts here... AMD is still playing catch up. I thought they had a chance in the AI space, but with the launch of the latest chip and AMD isn't supplying MI300X with any remarkable impact... Nvidia is straight teabaggin' the competition.

Without a doubt, Nvidia is the leader in AI. However, the biggest limiters for these AI chips are the interposer or COWOS as TSMC calls it and the HBM. Nvidia is very rich, but they are still not in the ballpark of the true big players: Apple, Google, and Microsoft. Of these, Google is the second biggest AI hardware company in the world. It is an easy fact to overlook as their products are only accessible via Google's own cloud.

 

[Guwapo77]

Without a doubt, Nvidia is the leader in AI. However, the biggest limiters for these AI chips are the interposer or COWOS as TSMC calls it and the HBM. Nvidia is very rich, but they are still not in the ballpark of the true big players: Apple, Google, and Microsoft. Of these, Google is the second biggest AI hardware company in the world. It is an easy fact to overlook as their products are only accessible via Google's own cloud.

I will have to challenge you on that one.

Current Value

Nvidia - 2.36 Trillion
Apple - 2.66 Trillion
Google (Alphabet inc) - 1.88 Trillion
Microsoft -3.19 Trillion
Amazon - 1.86 Trillion

I believe Nvidia has a seat at the table. Their value has exploded since the launch AI. I believe they just might become the most valuable company in the world...might.

 

[AnotherReader]

I will have to challenge you on that one.

Current Value

Nvidia - 2.36 Trillion
Apple - 2.66 Trillion
Google (Alphabet inc) - 1.88 Trillion
Microsoft -3.19 Trillion
Amazon - 1.86 Trillion

I believe Nvidia has a seat at the table. Their value has exploded since the launch AI. I believe they just might become the most valuable company in the world...might.

I think we're using different metrics. I'm talking about revenues. Market capitalization is unreliable and is a hopeful indicator of future performance. By market capitalization, I expect Nvidia has a good chance of becoming the most valuable company in the world this year.

 

[Guwapo77]

I think we're using different metrics. I'm talking about revenues. Market capitalization is unreliable and is a hopeful indicator of future performance. By market capitalization, I expect Nvidia has a good chance of becoming the most valuable company in the world this year.

Looks like Nvidia is second on TSMC's list which isn't too bad. @ 11% of TSMCs revenue. I can't see them overtaking Apple, but they are sitting in there nicely.

Analyst estimates Nvidia is now TSMC's second largest customer accounting for 11% of revenue in 2023

Nvidia paid TSMC $7.73 billion in 2023 for its services

www.tomshardware.com

 

[stimpy88]

I'm sure TSMC would like nGreedia even more if they let them make more, instead of nGreedia artificially manipulating their markets by simply not ordering enough to create demand and help them raise prices.

 

[Vayra86]

TSMC to Huang: 'The more you buy, the more you save!' It just works!