Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/amd-to-buy-zt-systems-in-4-9-billion-challenge-to-nvidia.20818/
        )

    [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
)

AMD to Buy ZT Systems in $4.9 Billion Challenge to Nvidia

Daniel Nenni

Admin
Staff member
7153c388fe756cd2d267ad1dadcaf851

(Bloomberg) -- Advanced Micro Devices Inc. agreed to buy server maker ZT Systems in a cash and stock transaction valued at $4.9 billion, adding data center technology that will bolster its efforts to challenge Nvidia Corp.

ZT Systems, based in Secaucus, New Jersey, will become part of AMD’s Data Center Solutions Business Group, according to a statement Monday. AMD will retain the business’s design and customer teams and look to sell the manufacturing division. The purchase price includes a contingent payment of $400 million based on certain milestones.

AMD investors applauded the deal, sending the shares up as much as 3.5% in New York on Monday. The transaction consists of 75% cash and 25% AMD common stock.

Closely held ZT has extensive experience making server computers for owners of large data centers — the kind of customers that are pouring billions of dollars into new artificial intelligence capabilities.

“AI is the most transformational technology of the last 50 years and our No. 1 strategic priority,” AMD Chief Executive Officer Lisa Su said in a call with investors on Monday, adding that the ZT takeover will allow the company’s largest customers to more rapidly deploy AMD’s AI infrastructure. ZT’s engineers “understand the challenges of designing and managing high-performance and high-density systems at a massive scale,” she said.

The company has its sights set on Nvidia, the runaway leader in the market for data center gear needed to support AI computing. AMD is adding software and hardware capabilities that will better match the breadth of Nvidia’s offerings and speed the uptake of its chips.

AMD is the second-biggest provider of the graphics processors that have become so vital to developing AI software. And it’s spent more than $1 billion in the last 12 months to expand its reach in that market. In July, the company agreed to buy Silo AI for $665 million to add a maker of artificial intelligence models.

ZT had about $10 billion of sales in the last 12 months, almost all from manufacturing. AMD doesn’t want to compete with customers such as Dell Technologies Inc. and HP Enterprise Co. so it will sell those production operations after closing the transaction in the first half of next year. The chipmaker will retain about 1,000 engineers, Su said.

Santa Clara, California-based AMD is considered Nvidia’s closest rival in AI processors. Its new MI range of accelerator chips will bring in more than $4.5 billion of new revenue this year, the company has projected. That puts it ahead of all other challengers to Nvidia, but still way behind. Analysts expect Nvidia to generate $100 billion in data center revenue this fiscal year.

Part of Nvidia’s expansion has been fueled by offering chips, networking, servers, software and services, all aimed at making AI use more pervasive in the economy. Under Su’s leadership, AMD is showing a commitment to match that reach.

AMD’s gains in recent years have pushed its market value to well over $240 billion, more than twice that of longtime rival Intel Corp.

The company is looking for “a strategic partner” to buy ZT’s US-based data center infrastructure manufacturing business. A deal for that segment is expected to follow the close of the ZT takeover in the first half of 2025.

“We believe ZT’s manufacturing business will be a very attractive asset to multiple players in the ecosystem given the scale of the business and the US and European footprint,” Su said.

 
“AI is the most transformational technology of the last 50 years and our No. 1 strategic priority,” AMD Chief Executive Officer Lisa Su said in a call with investors on Monday...

Seriously? What about the SoC? SOCs put supercomputers in our pocket. AI is, however, the most overhyped technology of the last 50 years. :LOL:
 
Personally, I don't like this kind of acquisition unless there is a strong strategic rational.

If AMD's bottleneck was getting themselves server design wins, it would make sense, but I feel like that's not the case. Maybe 5 years ago it was, but now all the hyperscalers are already supporting AMD. That said, maybe there is still a bottleneck.

I feel like the area that AMD needs to focus on is software, resolving lingering driver issues, and putting something out there that can be competitive to CUDA.
 
Personally, I don't like this kind of acquisition unless there is a strong strategic rational.

If AMD's bottleneck was getting themselves server design wins, it would make sense, but I feel like that's not the case. Maybe 5 years ago it was, but now all the hyperscalers are already supporting AMD. That said, maybe there is still a bottleneck.

I feel like the area that AMD needs to focus on is software, resolving lingering driver issues, and putting something out there that can be competitive to CUDA.
The issue is that merchant silicon in the datacenter is a shrinking TAM. The future is systems. Be they from the systems companies (like Graviton, IPU, Cobalt, etc) or from merchant chip companies offering fully integrated systems (ie NVIDIA). If AMD wants to compete with NVIDIA they must look at a systems level and optimize from the software down. Just designing a mean chip seems to be insufficient in this day and age.
 
The issue is that merchant silicon in the datacenter is a shrinking TAM. The future is systems. Be they from the systems companies (like Graviton, IPU, Cobalt, etc) or from merchant chip companies offering fully integrated systems (ie NVIDIA). If AMD wants to compete with NVIDIA they must look at a systems level and optimize from the software down. Just designing a mean chip seems to be insufficient in this day and age.
I think I understand the rational, you want to sell a system with an AMD CPU, and a bunch of GPUs for AI, where everything works together optimized by software, and you don't need to worry so much about supporting every single hardware/software combination.

I guess that makes sense.

However the same thinking got Intel in trouble, where they were doing the same thing, designing systems assuming a certain CPU/GPU mix, optimized for that, and got caught off guard when their assumptions about the market ended up not being accurate.
 
I think I understand the rational, you want to sell a system with an AMD CPU, and a bunch of GPUs for AI, where everything works together optimized by software, and you don't need to worry so much about supporting every single hardware/software combination.

I guess that makes sense.

However the same thinking got Intel in trouble, where they were doing the same thing, designing systems assuming a certain CPU/GPU mix, optimized for that, and got caught off guard when their assumptions about the market ended up not being accurate.
Intel isn't a systems company though. Sure you can say they are better at optimizing with partners in say laptop and making new standards. But IMO what intel was trying to do with Altera and Optane is not even remotely similar to what NVIDIA does with their networking, GPU, blades, and servers. Intel never really went above providing silicon solutions and working with HPE, DELL, GNU/Linux, MS, etc to make sure said solutions were well supported. What NVIDIA is trying to do is moving up the value chain and focusing on selling complete servers. I imagine that there are many burnt bridges among OEMs and the like since NVIDIA is trying to steal their business, so I guess we will have to see how this pans out long term. But alas this is the way the winds of change seem to be blowing for many years now. All that is "different" is that NVIDIA is the only merchant chip vendor that seems to be successful in actually moving up the value chain.

Also a final note on intel; just because intel bungles something doesn't make it a bad idea. A good idea with bad execution is still a good idea. Take for example intel's many ill fated adventures into the world of phones. Phones are a good place to be but intel bungled the execution multiple times. Put another way the problem was intel not the mobile business being a bad place to be.
 
The issue is that merchant silicon in the datacenter is a shrinking TAM. The future is systems. Be they from the systems companies (like Graviton, IPU, Cobalt, etc) or from merchant chip companies offering fully integrated systems (ie NVIDIA). If AMD wants to compete with NVIDIA they must look at a systems level and optimize from the software down. Just designing a mean chip seems to be insufficient in this day and age.
Intel isn't a systems company though. Sure you can say they are better at optimizing with partners in say laptop and making new standards. But IMO what intel was trying to do with Altera and Optane is not even remotely similar to what NVIDIA does with their networking, GPU, blades, and servers. Intel never really went above providing silicon solutions and working with HPE, DELL, GNU/Linux, MS, etc to make sure said solutions were well supported. What NVIDIA is trying to do is moving up the value chain and focusing on selling complete servers. I imagine that there are many burnt bridges among OEMs and the like since NVIDIA is trying to steal their business, so I guess we will have to see how this pans out long term. But alas this is the way the winds of change seem to be blowing for many years now. All that is "different" is that NVIDIA is the only merchant chip vendor that seems to be successful in actually moving up the value chain.

Also a final note on intel; just because intel bungles something doesn't make it a bad idea. A good idea with bad execution is still a good idea. Take for example intel's many ill fated adventures into the world of phones. Phones are a good place to be but intel bungled the execution multiple times. Put another way the problem was intel not the mobile business being a bad place to be.
I agree that the merchant chip market is in decline, but I disagree that the future is in systems, it is in fabs. Fabs are the rarest and most difficult skill capability. Modern chip design software tools and IP means a small team with a target volume of mere millions of chips can be viable. Chip design used to be difficult, and it still is if you insist on full-custom chips, like the merchant chip vendors mostly do, but it doesn't have to be that way.

Nvidia does not build and sell servers. Nvidia has a certification program for products from server OEMs, but they don't sell servers or systems. I don't think it is accurate that Nvidia is trying to steal server OEM business. AMD, on the other hand, is now on a path I do not understand with ZT Systems.

Long ago, Intel used to be a systems vendor, for supercomputing. Intel actually sold systems, and they were pretty good in their day. I knew that team.

 
I agree AMD is not Intel. I guess you have to trust that Lisa Su has some kind of strategic rational driving this, and I trust Lisa Su a lot more than I trust Pat G.
 
I agree AMD is not Intel. I guess you have to trust that Lisa Su has some kind of strategic rational driving this, and I trust Lisa Su a lot more than I trust Pat G.
I agree about LS and PG, but inexplicable doesn't come in various colors. :)
 
I agree that the merchant chip market is in decline, but I disagree that the future is in systems, it is in fabs. Fabs are the rarest and most difficult skill capability. Modern chip design software tools and IP means a small team with a target volume of mere millions of chips can be viable. Chip design used to be difficult, and it still is if you insist on full-custom chips, like the merchant chip vendors mostly do, but it doesn't have to be that way.
I don't disagree. I was speaking from the perspective of a fabless company where owning and running a fab is completely out of the question.
Nvidia does not build and sell servers. Nvidia has a certification program for products from server OEMs, but they don't sell servers or systems. I don't think it is accurate that Nvidia is trying to steal server OEM business. AMD, on the other hand, is now on a path I do not understand with ZT Systems.
Thanks for correcting my understanding of this.
Long ago, Intel used to be a systems vendor, for supercomputing. Intel actually sold systems, and they were pretty good in their day. I knew that team.
Interesting.
I agree AMD is not Intel. I guess you have to trust that Lisa Su has some kind of strategic rational driving this, and I trust Lisa Su a lot more than I trust Pat G.
What she has done for AMD was truly great, but I do think she is overvalued. Lisa's predecessors did a lot of work planting the seeds that she got to harvest, and in my opinion those folks don't get enough credit for AMD's turnaround. As for the future, I feel like AMD needs a shake up. Lisa was a great catch-up/comeback/underdog CEO. But I think AMD needs a CEO who knows how to win and is visionary. I've said this before, but I think AMD spends too much time following and not enough time looking where the puck will be. I also think that they somewhat squandered the time they had with a process lead. Even if you think all intel nodes are doomed to fail, intel is still collaborating with TSMC to get access to their best early. Heck isn't intel getting N3 products out faster than even MTK or QCOM. Meanwhile AMD is focused on doing as much as possible with as little as possible rather than swinging for the fences with big bets, big pre-pays, and early process node engagement.
 
What she has done for AMD was truly great, but I do think she is overvalued. Lisa's predecessors did a lot of work planting the seeds that she got to harvest, and in my opinion those folks don't get enough credit for AMD's turnaround. As for the future, I feel like AMD needs a shake up. Lisa was a great catch-up/comeback/underdog CEO. But I think AMD needs a CEO who knows how to win and is visionary. I've said this before, but I think AMD spends too much time following and not enough time looking where the puck will be. I also think that they somewhat squandered the time they had with a process lead. Even if you think all intel nodes are doomed to fail, intel is still collaborating with TSMC to get access to their best early. Heck isn't intel getting N3 products out faster than even MTK or QCOM. Meanwhile AMD is focused on doing as much as possible with as little as possible rather than swinging for the fences with big bets, big pre-pays, and early process node engagement.
This is a very unique and interesting perspective. I need to think about my reaction for a while.
 
I don't disagree. I was speaking from the perspective of a fabless company where owning and running a fab is completely out of the question.

Thanks for correcting my understanding of this.

Interesting.

What she has done for AMD was truly great, but I do think she is overvalued. Lisa's predecessors did a lot of work planting the seeds that she got to harvest, and in my opinion those folks don't get enough credit for AMD's turnaround. As for the future, I feel like AMD needs a shake up. Lisa was a great catch-up/comeback/underdog CEO. But I think AMD needs a CEO who knows how to win and is visionary. I've said this before, but I think AMD spends too much time following and not enough time looking where the puck will be. I also think that they somewhat squandered the time they had with a process lead. Even if you think all intel nodes are doomed to fail, intel is still collaborating with TSMC to get access to their best early. Heck isn't intel getting N3 products out faster than even MTK or QCOM. Meanwhile AMD is focused on doing as much as possible with as little as possible rather than swinging for the fences with big bets, big pre-pays, and early process node engagement.
I agree with your assessment with Lisa Su, while she is a great CEO, a lot of the credit should go to Rory Read was at the helm of AMD who lay a lot of the foundational groundwork for Lisa. He got AMD out of the hole that Dirk Meyer and Hector Ruiz created. There is a saying regarding AMD, AMD never missed an opportunity to miss an opportunity.
 
"Intel isn't a systems company though.", nghanayem

Intel runs Tiber Developer Cloud. I think it knows how to design and manage large systems.

 
Intel has way more experience than AMD in design and validation something AMD Lacks .They have started gaining momentum in DC for past few years it is a strategic acquisition to expand in DC which is the major revenue source for AMD
 
I don't get it. AMD is up 3.4% on the NASDAQ while I'm typing this, and NVDA is down about 1.8%, apparently on the ZT Systems news. I must be missing something big.
 
Nvidia does not build and sell servers. Nvidia has a certification program for products from server OEMs, but they don't sell servers or systems. I don't think it is accurate that Nvidia is trying to steal server OEM business. AMD, on the other hand, is now on a path I do not understand with ZT Systems.
Take a look at this year’s GTC. NVIDIA has definitely jumped into the server and even full data center business, while also continuing to sell SoCs / board to OEMs and CSPs. Look up DGX SuperPOD.


I think they discovered they have to get into the system design business to deal with the power, communication and complexity issues at the system level. They can‘t just rely on the secondhand feedback from OEM designs and CSP partners that are building their own chips and systems, as well, based on their own specific workloads. AMD’s move reinforces the necessity of moving into system design to guide the next generations of AI chips and associated communications chips, just like AMD‘s purchase of Silo.ai reflects the need for GenAI chip companies to offer direct software solutions for building GenAI applications (not just models).
 
Last edited:
Take a look at this year’s GTC. NVIDIA has definitely jumped into the server and even full data center business, while also continuing to sell SoCs / board to OEMs and CSPs. Look up DGX SuperPOD.


I think they discovered they have to get into the system design business to deal with the power, communication and complexity issues at the system level. They can‘t just rely on the secondhand feedback from OEM designs and CSP partners that are building their own chips and systems, as well, based on their own specific workloads. AMD’s move reinforces the necessity of moving into system design to guide the next generations of AI and associated communications chips, just like AMD‘s purchase of Silo.ai reflects the need for GenAI chip companies to offer direct software solutions for building GenAI applications (not just models).
Thank you. I missed this somehow, since March of 2023 no less.


Now AMD's acquisition of ZT makes more sense.

Apologies to @nghanayem for posting an incorrect response.
 
I don't disagree. I was speaking from the perspective of a fabless company where owning and running a fab is completely out of the question.

Thanks for correcting my understanding of this.

Interesting.

What she has done for AMD was truly great, but I do think she is overvalued. Lisa's predecessors did a lot of work planting the seeds that she got to harvest, and in my opinion those folks don't get enough credit for AMD's turnaround. As for the future, I feel like AMD needs a shake up. Lisa was a great catch-up/comeback/underdog CEO. But I think AMD needs a CEO who knows how to win and is visionary. I've said this before, but I think AMD spends too much time following and not enough time looking where the puck will be. I also think that they somewhat squandered the time they had with a process lead. Even if you think all intel nodes are doomed to fail, intel is still collaborating with TSMC to get access to their best early. Heck isn't intel getting N3 products out faster than even MTK or QCOM. Meanwhile AMD is focused on doing as much as possible with as little as possible rather than swinging for the fences with big bets, big pre-pays, and early process node engagement.
Choose the right manufacturing partner regardless of history, Choose architectures that are both high performance and efficient to design, set and achieve realistic goals. Execute.
AMD has been successful with these IMO. All companies are successful with these.

Lets see how this ZT thing pans out. I dont think it is a game changer and AI isnt the greatest change in 50 years IMHO
 
Back
Top