Editor’s Note: This race to the next generation of artificial intelligence is on…
The next leg in the AI boom will look a little different from what we’ve seen so far. It isn’t just the industry heavy hitters like NVIDIA that will benefit.
The smaller companies – the ones creating the foundation for this next level of artificial intelligence to flourish – are the ones that will truly thrive.
Jeff did a deep dive into the future of artificial intelligence in his AI Emergency Meeting last week – including a small portfolio he’s put together of companies helping make the next generation of AI happen…
You can still access the replay of Jeff’s meeting, but hurry… there’s an AI industry event scheduled for August 28 that Jeff believes is going to shake up the market. We want you to be ready when it hits. Just go here to watch the replay.
Six o’clock on a Monday morning is not a normal time to announce a major acquisition…
But it’s one heck of a way to kick off the week.
Semiconductor powerhouse Advanced Micro Devices (AMD) announced a $5 billion acquisition of ZT Systems, a company I’m confident very few have ever heard of.
Unless you’re from the industry, I wouldn’t expect anyone to know what the company does. Go to its website and you’ll find descriptions like, “We create cloud-enabling server solutions that help the world’s leading companies turn their ideas into realities.”
If that confuses you, don’t worry, you’re not alone. It’s just marketing fluff that can mean just about anything, or nothing at all.
Making the acquisition even harder to understand is that ZT Systems has always been a private company and has taken no investment from venture capital. That means there is very little public information regarding the business AMD just acquired.
It’s worth $5 billion, but actually, I would argue it’s worth even more than that. It’s a big deal and absolutely worth understanding.
ZT Systems used to be a boring business.
It designs, manufactures, assembles, integrates, and tests servers and server racks used in data centers.
Source: ZT Systems
Boring right? It was, but not any longer.
Not since the demand for artificial intelligence kicked off a once-in-a-generation growth and productivity megacycle. AI turned ZT Systems into a business that now has more than $10 billion in annual sales.
That’s a big business. And its annual sales represent about 40% of AMD’s forecasted sales of $25.6 billion for this year. That makes it a big acquisition even for AMD.
AMD is currently sitting on $5.3 billion in cash and will generate about $3.5 billion in free cash flow this year. It’s not clear yet how AMD structured the deal. Will it pay cash, or will it raise additional debt to help finance the acquisition?
In reality, it doesn’t matter. AMD expects the deal to be finalized in the first half of next year. It’s on track to generate an additional $6.7 billion in free cash flow in 2025. It will have no trouble raising additional capital if it wants to use debt, and it can pay cash for the deal as well.
Better yet, AMD is planning to sell off the manufacturing business of ZT Systems to another player in the industry. If I had to bet, I’d say that AMD will have a buyer lined up by the time the acquisition is completed next year.
I’d also bet that the sale of ZT Systems’ manufacturing business will allow AMD to capture somewhere between $1–2 billion, reducing its overall cost of acquisition of ZT Systems.
The strategy is for AMD to sell off the low-margin manufacturing business of ZT Systems and retain the higher-margin, higher-value data center systems, server, and server rack design, testing, validation, integration, supply chain management, and related services and support.
Some are already mistakenly claiming that AMD is “buying additional revenues” by acquiring ZT Systems.
But that’s not what the deal is about.
The acquisition is about scale. Not revenue scale. Scale as it relates to having the systems design expertise to help those who wish to build data centers to scale their businesses.
The industry has been moving so fast that those who are buying and building data centers will take what they can get, even if it is a generic design. AI is improving so quickly that many companies are purchasing servers with GPUs in any way that they can.
Yes, there is such a thing as a one-size-fits-all server rack. We can oversimplify and think of it like an off-the-rack suit. It fits OK and gets the job done, but it doesn’t look great.
This issue is that these data center facilities are massive.
In the image above, I’m standing in front of just one of Amazon Web Services’ data centers in Ashburn, Virginia. The scale is so large that, even at that distance, the entire facility isn’t in the background. To the left of where I’m standing, it continues on.
In total, there are 131 data center facilities like this in the “Data Center Alley” of Ashburn. Amazon Web Services owns 15 of them. All of them are impressive in scale. I know this because I drove around all of them.
My point is that because the scale is so large, even a relatively small server design optimization – like 5% for particular computational workloads – can save tens of millions of dollars in operational expenses.
And those optimizations naturally improve performance as well.
This is important because all workloads aren’t the same. A data center designed to train large language models (LLMs) will be different than one designed to run machine learning applications.
That’s why AMD’s acquisition is strategic and important. It will multiply AMD’s ability to work with its customers to customize data center and server solutions… resulting in higher performance and improved operational expenses.
And AMD has something that NVIDIA doesn’t have.
A publicly traded company comparable to ZT Systems is Super Micro (SMCI). Super Micro is down by about 50% from its March all-time high. It’s now a $37 billion company.
There’s nothing wrong with Super Micro. It’s still growing rapidly. It’s just that its EBITDA margins are just around 10% for its current fiscal year.
Super Micro is trading around 1.3 times the forecasted 2025 fiscal year (ending June 30, 2025) revenues of $28 billion.
What does this tell us?
AMD was able to negotiate a great price for ZT Systems. If we apply the same multiple of 1.3 against roughly $10 billion in revenues, we’re at a $13 billion valuation. AMD picked up ZT Systems for $5 billion. And it will sell the low-margin manufacturing business.
Nice deal.
The media are positioning the acquisition as a way to compete with NVIDIA. I don’t really see it that way.
The data center market size is about $250 billion this year and is growing to more than $500 billion by 2030 – doubling in just six years.
What’s really happening is that the size of the “pie” is growing so quickly, it’s hard to keep up. There’s so much business for the AMD and NVIDIA duopoly to chase. AMD will continue to grow as it acquires larger chunks of that expanding business opportunity.
The ZT Systems acquisition was about accelerating its ability to scale to even more data center opportunities.
AMD needed more design and systems expertise for servers, racks, power systems, cooling systems, and wiring so that it could continue to scale…
…from a quarter-trillion-dollar business to a half-trillion-dollar business in the years to come.
The Bleeding Edge is the only free newsletter that delivers daily insights and information from the high-tech world as well as topics and trends relevant to investments.
The Bleeding Edge is the only free newsletter that delivers daily insights and information from the high-tech world as well as topics and trends relevant to investments.
