Discussion [WSJ]AMD Is in Advanced Talks to Buy Xilinx

[tamz_msc]

WSJ News Exclusive | AMD Is in Advanced Talks to Buy Xilinx

A deal between the rival chip makers could be worth more than $30 billion and mark the latest big tie-up in the rapidly consolidating industry.

www.wsj.com

Advanced Micro Devices Inc. is in advanced talks to buy rival chip maker Xilinx Inc., according to people familiar with the matter, in a deal that could be valued at more than $30 billion and mark the latest big tie-up in the rapidly consolidating semiconductor industry.

The companies are discussing a deal that could come together as soon as next week, the people said. There is no guarantee they will get there, especially given that the talks had stalled before recently restarting, according to some of the people.

AMD’s market value now tops $100 billion after its shares soared 89% this year as the coronavirus pandemic stokes demands for PCs, gaming consoles and other devices that use the company’s chips. Second-quarter revenue rose 26% to $1.93 billion, while net income jumped more than fourfold to $157 million on the back of record notebook and server-processor sales, AMD said.

The surge in AMD shares could embolden the company to make an acquisition using its stock as currency. Xilinx has a market value of about $26 billion, with its shares up about 9% so far this year, just ahead of the S&P 500’s 7% rise.

With a typical takeover premium, a deal would value the company at more than $30 billion.

AMD, based in Santa Clara, Calif., is led by Chief Executive Lisa Su. In addition to desktop and notebook computers, it makes components used in gaming systems such as Xbox and PlayStation that have been in high demand as the pandemic forces people to stay at home. It also has a growing data-center-processor business that increasingly rivals that of Intel Corp., long the dominant player in that segment.

The addition of Xilinx, led by CEO Victor Peng, would put AMD on a more even competitive footing with Intel and give it a bigger position in fast-growing telecommunications and defense markets.

San Jose, Calif.-based Xilinx’s chips are used in wireless communications, data centers and industries such as automotive and aerospace. Its shares have been hurt by trade tensions between the U.S. and China, and namely the Trump administration’s limitations on shipments to China’s Huawei Technologies Co. because of security concerns. Analysts estimated Huawei accounted for roughly 6% to 8% of Xilinx’s revenue.


Xilinx makes microchips called field-programmable gate arrays, or FPGAs. Unlike standard chips, they can be reprogrammed after they are produced. That makes them valuable in rapid prototyping and in fast-emerging technologies where there isn’t enough time to go through a yearslong development process necessary for other chips.

FPGAs are commonly used in new superfast 5G telecommunications infrastructure, although they may be replaced later by standard chips once the technology is more mature. They are also often used in military communications and radar systems.

Intel is the other main player in the FPGA market, having built its business by acquiring Altera Corp. in 2015.

AMD, which specializes in central processing units that serve as computers’ digital brains, has been gaining share on Intel in recent years, releasing new generations of processors that match or beat its larger rival’s on many performance benchmarks. AMD had around a 20% share in personal computer CPUs as of the second quarter, according to Mercury Research, up from around 8% three years ago.

Consolidation has swept through the semiconductor industry as chip makers seek scale and expand their product portfolios to support the increasing number of everyday items that are connected to the internet. Xilinx, for one, has considered a number of potential tie-ups in recent years that didn’t come to fruition.

I don't know - 30 billion for Xilinx when NVIDIA is paying 40 billion for ARM?

 

[Doug S]

Intel bought Altera as much to help them fill their fabs (which would have been a problem if they weren't stuck on 14nm forever) as because its tech was complementary to Xeon. AMD has no fabs, so they can only buy it on its tech being complementary.

I don't know enough about the FPGA market and where it is expected to go in the next 5-10 years to judge it from that perspective, but I'll say the same thing I said about Nvidia's purchase of ARM. Acquisitions larger than $10 billion almost never work out from a financial perspective. For every one you can point to that was clearly profitable, I can point to ten that weren't.

Seems like a risky move for a company that has only recently got back on its feet, but I guess they feel they need to be able to offer the same suite of technologies Intel does - and this allows them to stay one step ahead of the Nvidia/ARM combo (if indeed FPGAs in the datacenter become a must have someday)

 

[sandorski]

Only AMD knows what their Future Vision is and which acquisition helps them achieve that the best. Now is the time for them to make such moves as they are on top of the strongest position they have ever been and their Revenues offer them the ability to make such moves. Anyone who still thinks acquiring ATI was a mistake is clearly stuck in the past and not seeing the big picture. Time will tell how this works out, if it happens. I suspect that AMD already has some ideas on Products to be produced as a result of the acquisition and/or how to implement certain technologies into their existing Products with them.

 

[piokos]

Intel bought Altera as much to help them fill their fabs (which would have been a problem if they weren't stuck on 14nm forever) as because its tech was complementary to Xeon. AMD has no fabs, so they can only buy it on its tech being complementary.

The fab argument is very much correct. It was also among the top reasons to take on SSD, IoT and modem businesses.
Not all of these have to be profitable as separate entities. Some of them are just to cover the unused fab cost (which would otherwise become a loss in CPU business).

But Altera was also a strong technological boost. It helped Intel in the CPU and IoT business, but it was also the first step in their long term plans for autonomous driving domination (which will probably work).

It's hard to point similar synergies for AMD - at least based on their current lineup. Maybe they have plans we don't know about.

 

[piokos]

Now is the time for them to make such moves as they are on top of the strongest position they have ever been and their Revenues offer them the ability to make such moves.

Not really. They're still a relatively small company compared to other IT giants. And with much lower margins than Intel and Nvidia. They will improve for sure, but saying that they're in a great position right now is pretty naive at best.

The more obvious reasoning is that they start to hit a wall when it comes to growing in their core businesses.
If they can't take more market share from Intel and Nvidia, they'll have to enter other businesses - ideally something not very distant from what they know. Because every listed company is expected to grow.

I suspect that AMD already has some ideas on Products to be produced as a result of the acquisition and/or how to implement certain technologies into their existing Products with them.

It would be quite weird if they were thinking about this deal with no idea how to use Xilinx...

 

[sandorski]

Not really. They're still a relatively small company compared to other IT giants. And with much lower margins than Intel and Nvidia. They will improve for sure, but saying that they're in a great position right now is pretty naive at best.

The more obvious reasoning is that they start to hit a wall when it comes to growing in their core businesses.
If they can't take more market share from Intel and Nvidia, they'll have to enter other businesses - ideally something not very distant from what they know. Because every listed company is expected to grow.

It would be quite weird if they were thinking about this deal with no idea how to use Xilinx...

What other point in their history were they in a better position to make such a purchase? Their relative size to others is moot. All that matters is their capability at making a deal and Now is where they find themselves at the height of opportunity. Intel is down and out for another 1-2 years, AMD's income continues to Grow and that should accelerate with Zen3 along with significant Marketshare growth, their competitiveness with NVidia also appears to be hitting a high point as well. In 3-4 years things could change significantly, to the point where such a move becomes impossible. Like with anyone else, AMD needs to make moves when they become possible.

 

[HurleyBird]

I don't believe there are any other FPGA competitors other than Altera, which Intel owns now. Sure, AMD could have a partnership to get FPGA chiplets but buying them ensures that will happen.

List and comparison of FPGA companies

With the top two FPGA companies taking up 89% of the FPGA market, you can be forgiven for thinking there was no one else out there. Xilinx and Altera have done a good job of defending the duopoly but a few companies are gradually winning market share by targeting specific applications and...

www.fpgadeveloper.com

 

[LightningZ71]

I don't see AMD as having any other real choice in the matter. FPGA is an important part of large scale computing going forward. AI and ML are the big areas of growth right now, and, in mass installations those are either happening on glorified video cards, or FPGAs and custom ICs. With Intel owning a large chunk of the market, AMD could either spend many multiples of this money trying to R&D their way into the other half of it, or just by in for product, revenue, and profits now. This is definitely a complementary purchase for them, and enables them to offer more highly integrated solutions to pitch against Intel's offerings on some of the largest projects out there.

Going after ARM was not going to be an option. They would have dumped a massive amount of money into something that is essentially an R&D sink hole with limited return on investment. They also would have faced even more scrutiny for the purchase than Nvidia is facing. Buying Mediatek does them no good. They would become beholden to ARM, which is expected to be an Nvidia holding, with lots of high end competition in the market.

While I believe that the price is going to be somewhat difficult to swallow for them, what they are getting is a profitable, on going concern that has lots of upside and better enables them to have multiple income sources that, while related, are separate enough to be more insulating against market volatility.

 

[AmericanLocomotive]

FPGAs have been touted as the next big thing ever since Intel acquired Altera in 2015, but I've yet to see that potential materialize into products that are massive drivers of growth.

FPGAs are EVERYWHERE. So many electronic devices - everything form consumer stereos to high end scientific imaging equipment utilize FPGAs. FPGAs command very high margins. The highest-end FPGAs command over $50,000 per chip. They don't exist much in the PC space, but they are in just about everything else.

I'm not sure what the exact synergy is with AMD's lineup though. I think maybe a partnership or collaboration would be a better idea than a straight acquisition, as AMD would be paying this off for literally decades.

Also I'm really against giant semiconductor mergers. There has been way too much consolidation in this industry.

 

[SarahKerrigan]

FPGAs are EVERYWHERE. So many electronic devices - everything form consumer stereos to high end scientific imaging equipment utilize FPGAs. FPGAs command very high margins. The highest-end FPGAs command over $50,000 per chip. They don't exist much in the PC space, but they are in just about everything else.

I'm not sure what the exact synergy is with AMD's lineup though. I think maybe a partnership or collaboration would be a better idea than a straight acquisition, as AMD would be paying this off for literally decades.

Also I'm really against giant semiconductor mergers. There has been way too much consolidation in this industry.

Basically agreed. The overlap here right now looks like "they're both semiconductor companies," which is pretty limited - I feel like Renesas or TI would have made more sense as a merger partner for Xilinx. On the other hand, if we see massive FPGA growth as datacenter accelerators (whether smart NICs or as something else) that could change the equation a bit.

 

[Zstream]

Nah, I think you're missing the point. The main reason why Intel is picked over many other offerings is due to the FPGA being available. The high scale environments, such as Cisco, F5.. "insert whatever hardware company" all primarily use Intel due to the FPGA chips. This would increase their server population by many times. However, to your point, 30B is way, way too much IMO.

 

[zinfamous]

From -$30 billion (debt) just under 3 years ago, to considering $30 billion purchase today.

Lisa Su is a freaking rock god of tech, lol.

 

[Atari2600]

Ahhh, this smells to me like Lisa's ATi.

IMO, the good people in every company are constrained by both the bad ones and the systems/processes imposed on them by the bad ones.
The better thing to do is get the good people, not the company. Let them at a clean sheet design with processes & systems they draw up themselves.

Far cheaper and you end up with an easier integration as well as better product.


How long did it take AMD to get something fruitful out of ATi? Can anyone name a significant technology merger that worked well before rivals rendered the supposed benefits negligible?

 

[scannall]

Ahhh, this smells to me like Lisa's ATi.

IMO, the good people in every company are constrained by both the bad ones and the systems/processes imposed on them by the bad ones.
The better thing to do is get the good people, not the company. Let them at a clean sheet design with processes & systems they draw up themselves.

Far cheaper and you end up with an easier integration as well as better product.


How long did it take AMD to get something fruitful out of ATi? Can anyone name a significant technology merger that worked well before rivals rendered the supposed benefits negligible?

Most of Apple's acquisitions have been great. PA Semiconductor, Next and Intrinsity being vital ones.

 

[nicalandia]

Please, we're not on AMD.com. You spearhead the attack on Intel on this forum. Even in this thread, you couldn't help but take a dig at Intel. I have had many arguments with @tamz_msc on this forum where he's either attacking Intel products or defending AMD products. He's not all of a sudden a pro Intel member because he's sharing opinion that's not seen to be worshipping AMD like you know who.

You are on thin ice buddy, you won't last long here.

 

[HurleyBird]

Most of Apple's acquisitions have been great. PA Semiconductor, Next and Intrinsity being vital ones.

Absolutely. Small-ish, relatively cheap companies with key tech and/or world-class talent. That's not what AMD is doing, but it is what they should be doing. For much less than $30B AMD could buy all the promising CPU startups, and then throw in Imagination and one of the smaller FPGA outfits for good measure.

 

[scannall]

Absolutely. Small-ish, relatively cheap companies with key tech and/or world-class talent. That's not what AMD is doing, but it is what they should be doing. For much less than $30B AMD could buy all the promising CPU startups, and then throw in Imagination for good measure.

I agree with that all the way. Just use Apple as your Acquisition 101 textbook.

 

[Doug S]

Most of Apple's acquisitions have been great. PA Semiconductor, Next and Intrinsity being vital ones.

Apple paid $800 million total for all three. Even considering inflation that's less than 5% of what AMD is rumored to be forking out here.

The difference is that Apple was willing to wait for these acquisitions to bear fruit. Yes, Steve Jobs came aboard right away (and that ended up being worth almost any price since Apple might not exist if he hadn't) but it took a few years to turn NeXTStep into OS X (and later iOS etc.) and likewise it took a few years before PA Semi enabled them to deliver the fully custom cores in A6.

I don't know the FPGA landscape, but I wonder if there is a smaller player AMD could buy for 5% of the price of Xilinx that could do just as well for them with a few years of growth and investment if AMD was willing to be patient.

 

[scannall]

Apple paid $800 million total for all three. Even considering inflation that's less than 5% of what AMD is rumored to be forking out here.

The difference is that Apple was willing to wait for these acquisitions to bear fruit. Yes, Steve Jobs came aboard right away (and that ended up being worth almost any price since Apple might not exist if he hadn't) but it took a few years to turn NeXTStep into OS X (and later iOS etc.) and likewise it took a few years before PA Semi enabled them to deliver the fully custom cores in A6.

I don't know the FPGA landscape, but I wonder if there is a smaller player AMD could buy for 5% of the price of Xilinx that could do just as well for them with a few years of growth and investment if AMD was willing to be patient.

That's a valid question, and should be a serious consideration.

 

[itsmydamnation]

Ahhh, this smells to me like Lisa's ATi.

IMO, the good people in every company are constrained by both the bad ones and the systems/processes imposed on them by the bad ones.
The better thing to do is get the good people, not the company. Let them at a clean sheet design with processes & systems they draw up themselves.

Far cheaper and you end up with an easier integration as well as better product.


How long did it take AMD to get something fruitful out of ATi? Can anyone name a significant technology merger that worked well before rivals rendered the supposed benefits negligible?

tandberg

There i just did it.

people here who think ARM is a better buy for AMD then Xilinx have rocks in there head. ARM is complete duplication, Xilinx is completely complementary.

 

[HurleyBird]

The difference is that Apple was willing to wait for these acquisitions to bear fruit. Yes, Steve Jobs came aboard right away (and that ended up being worth almost any price since Apple might not exist if he hadn't) but it took a few years to turn NeXTStep into OS X (and later iOS etc.) and likewise it took a few years before PA Semi enabled them to deliver the fully custom cores in A6.

Imagine if, after buying up Nuvia and Ampere, AMD started licencing out ARM cores better than NVIDIARM itself is able to come up with, effectively wiping out most of the value of that $40B acquisition. That's the kind of lateral thinking you want to utilize.

 

[Hitman928]

Imagine if, after buying up Nuvia and Ampere, AMD started licencing out better ARM cores better than NVIDIARM itself is able to come up with, effectively wiping out most of the value of that $40B acquisition. That's the kind of lateral thinking you want to utilize.

Why would AMD need to buy Nuvia and Ampere to do that though?

 

[SarahKerrigan]

Imagine if, after buying up Nuvia and Ampere, AMD started licencing out ARM cores better than NVIDIARM itself is able to come up with, effectively wiping out most of the value of that $40B acquisition. That's the kind of lateral thinking you want to utilize.

Are you sure that architectural licensees can license their cores out? Aside from one case a long time ago (Faraday) I'm not aware of it ever happening.

 

[chrisjames61]

Well there's a chance that AMD might fail to integrate just like Intel. AMD's last big acquisition nearly made them bankrupt.

That was lifetime ago in the tech industry. Plus in the long run the acquisition has worked out and the ceiling is high if AMD can become more competitive with Big Navi.

 

[Zucker2k]

You are on thin ice buddy, you won't last long here.

I won't be provoked, so am just going to add this to the pile of hot air you've been blowing the last couple of days, noob.

 

[piokos]

What other point in their history were they in a better position to make such a purchase?

When they had solid ~40% market share (15-20 years ago) among most (all?) segments.

AMD's situation today isn't that great. But you have to look beyond CPU reviews to notice that...

Also: what would "to make such a purchase" mean? It looks like you though taking over other companies is the goal. It's not. The goals are: profitability and growth.
Trying luck in a new business (either on their own or by a merger/takeover) means they think it's a better choice than expanding what they have right now.

Intel bought Altera (and few other companies) when they had 80% market share. That's how they wanted to grow.
The fact that AMD considers this so early (having around 20% of CPU and GPU consumer market and a lot less of datacenter market) means they're starting to hit a wall in the core business.

In 3-4 years things could change significantly, to the point where such a move becomes impossible. Like with anyone else, AMD needs to make moves when they become possible.

That's very optimistic and, once again, showing a point of view formed by CPU benchmarks/reviews, not actual financials.

Yeah, 3-4 years from now AMD may be in much better shape. Or much worse. There are quite a few variables.

Today they have good position and competitive products.
What if this deal doesn't pay off? What if they start losing to Intel and Nvidia once again?

The pessimistic scenario (but still not worst case) is repeating what happened 10 years ago. I'm sure you don't want that.