AMD leaving billions on the table in mobile sales?

I mean, if you put data center first something has to suffer. PC Gamer’s perspective is clearly that of people looking for game-worthy performance in client PCs, and OEMs they talk to who are focused there probably feel like second-class citizens sometimes.

Multiple reports are claiming AMD has comprehensively dropped the ball with its new Ryzen AI chips for laptops (via TechSpot). Abazovic Consultancy Analysis (ACA) reports that AMD has “lost the trust” of laptop makers due to chip shortages, while ComputerBase claims to have been told AMD has left “billions of US dollars on the table” thanks to its missteps in supplying manufacturers.

ACA says that the problem stems from AMD’s decision to shift its focus and manufacturing volumes away from client chips, including laptop APUs like the new Ryzen AI 300 family, and towards more lucrative enterprise CPUs and GPUs. It’s certainly true that AMD’s CEO Lisa Su recently declared that AMD is now a “data center-first” company.

The ACA report contrasts apparently very limited availability of laptops based on the new Ryzen AI APU, codenamed Strix Point, with the success of Qualcomm and its new Snapdragon X chip, noting that Qualcomm bagged seven big OEM laptop makers for the launch of Snapdragon X while AMD could only muster three.

Meanwhile, a report on ComputerBase traces AMD’s problems back a little further and says that AMD’s existing Ryzen AI chips can only be had in “premium” devices. It claims to have been told that notebook makers have “no prospect of quickly obtaining many chips from AMD,” a situation that perpetuates “a persistent problem that has been known for a decade now.”

The net result? One laptop maker reportedly told ComputerBase that AMD has “left billions of US dollars on the table” with its many partners over the years. ComputerBase quotes sales figures showing that AMD has gained laptop market share in the last 18 months or so. But the gains from around 16% to 20 % over that period arguably are not dramatic.

If AMD has indeed missed an opportunity, it’s arguably only going to get harder from here. Intel’s new Lunar Lake laptop CPU looks pretty fantastic on paper and could provide very stiff competition in the premium thin-and-light segment, while Qualcomm has announced a cheaper 8-core variant of Snapdragon X which should enable AI-capable laptops in the $600 to $700 range.

1 Like

To me, this is a very “I want to make more money” as opposed to “I want AMD to make more money” point of view. Several factors go into this.

First of all, how many extra employees would AMD need to satisfy the demand for (say) $3B worth of extra laptop parts? The skills needed to sell data center chips are not an exact match for those needed to sell laptop chips. So there is the expense of finding, hiring, and keeping all those new employees.

Secondly, does TSMC really have the wafers AMD would need to get those chips at laptop margins given data center margins are so much higher. So it makes no sense to AMD to move the higher end wafers into lower margins.

Third and perhaps most importantly, what about Intel. Their GPUs and APUs are not highly desirable in the data center, to the point where Intel had to pay the TSMC to be competitive at all. Intel is I believe the second largest chipmaker, so they need to keep their fab occupied, or close them down. Closing is a bad option. They’d have to lay off employees, pay severance, and lose out on years of amortizing the existing equipment. So Intel’s best wafers that a decade agi were funding the highest margin chips are not making the chips Intel can sell, in the lower margin laptop and desktop sectors. I personally believe that Intel is being too optimistic in catching up to TSMC. I no longer see that happening. That means as Intel’s customer base will continue shrinking, possibly to the point where some of their foundries will no longer be profitable. Their problems all go back to forcing a theoretically superior architecture that so far has not competed well against chiplets. Be that as it may, Intel is pushing to own the laptop sector to keep its foundries open until such time if ever when Intel finally matches AMD in the data center. We’re looking at, what, a full decade of Intel’s margins dropping, high end margin sales dropping, profits and EPS dropping, all the while promising their next product would pass AMD. But neither AMD nor TSMC is sitting on their hands, and every time either one improves, then AMD’s data center products when Intel finally ships they are no longer as good.

I have an AMD gaming laptop that still plays all the games I want. I had to wait six months for it to arrive. I feel for the laptop makers. But I can’t fault AMD for what they’ve done. The profits back me up.

One more thing that IMO shows had badly Intel has declined. When Zen first shipped, I believed Intel was 3-5 years away from catching AMD. I did not intend to sell my AMD until Intel was two years at most away from catching AMD. But my negative target fixated wife just saw another half a million dollars in retirement savings, and I sold around $57 knowing I was leaving money on the table but if I am not making my wife happy, it’s ice floe time for me. Intel’s new products are not going to catch AMD. All AMD needs to make their current products much better is to overpay TSMC for their best wafers, and AMD’s high end chips will instantly be much better (and cost a lot more). There are some price points where that can make sense for AMD. But AMD has been steadily claiming market share in the high margin data center while also not paying premium prices for TSMC’s best wafers. The chiplet approach is just significantly higher in yields, because a defect in one large CPU takes out a much higher percentage of the yield than a defect destroying a chiplet. I also believe that chiplets, being smaller, allow better air flow access to disperse heat.

I’m mostly in “safe” dividend stocks these days, to keep my wife happy. But AMD sure looked inviting at $132 recently. The only reason I didn’t was the stock to sell was too close to it’s ex-dividend date, and that’s guaranteed money.

BTW, in personal new, my new kidney is working far better than I had any reason to expect. My anticipated hemoglobin was 11 or so in a normal range of 13.5 - 18.5. Part of why I stopped posting (but not reading) here is I could no longer trust myself to write cogently. But about six weeks ago, my hemoglobin (stable in the low 11s) started rising again. Last test about two weeks ago was 13.9. I don’t think it’s been that high this century. With hemoglobin comes brain (and muscle) power, and I’m playing complex games and have insomnia again. I’m also not staying lazy and letting my dividends accumulate but slowly getting back into research and real investing. Next year my wife will need to start withdrawing from her retirement accounts; I’m three years away. Now that I have a new kidney, I could easily last more than a decade, and I would have had four hotel trips this year but the second one got cancelled because of the transplant surgery. More travel is in our future. AMD and TMF were a big part of why our retirement is so comfortable all of our savings is likely to go unused save by our heirs. So thank all of you for giving me things to consider and teaching me to learn from others about investing.

5 Likes

Intel has moved to a chiplet/“tile”-centric approach to chip design as of Meteor Lake. So that advantage is gone.
https://www.intel.com/content/www/us/en/support/articles/000097683/graphics.html

In this generation they will depend on TSMC for Arrow Lake and Lunar Lake, so that means they basically can’t generate meaningful profit there.

I can’t imagine Intel “catching up” to TSMC in a way that is a threat to TSMC. Intel’s big objective I think is to displace Samsung and become the #2 player to TSMC. In process I don’t see them actually hitting parity with TSMC any time soon either, or ever.

I was greatly gratified to see TSMC Arizona is now producing the chip used in the iPhone 15 with yields said to be equivalent to those achieved in Taiwan (though apparently their production will go to the next Iphone SE). It may have been harder than anticipated but they got there. And TSMC in Japan is up and running as well AFAIK – they went from groundbreaking to production in 2 years by working 24/7 on construction, showing what you can do if you’re committed.

Greatly pleased to hear of your recovery. Enjoy the new vigor. Definitely curious to hear any new investing ideas or any news of your travels.

WRT your first post, I still think that Intel’s chiplets are a generation or three behind AMD’s because AMD has far more experience with it. I am willing to accept that with time Intel will catch up to and even pass AMD, but I can’t put a time frame on that shorter than five years right now. Intel has a LONG history of comparing their next product to AMD’s current product, and by the time their next product arrives, AMD’s next product will have also arrived.

The two trips I’ve taken post transplant were to Tacoma, for the founding the Adventure Game Hall of Fame. Some things went very well (friends, the convention) and some went poorly; GPS without voice does not work at all well with so many obscured street signs in the area. I was “on”, with lots of jokes hitting, and I got an unexpected standing ovation when I introduced myself at the voice actors panel (random seating assignment made me first) with this short sentence: “I am Cedric.”

The second trip to Chicago between the DNC and Labor Day went much better because of what I learned at the first trip. But both trips my hemoglobin was much lower. She who must be obeyed is insisting I go visit friends in California while she is in Sweden, so I’m considering it. But if I drive off, our cat won’t feed herself.

Next year I’m tentatively scheduled to visit the battlefields in France of D-Day and beyond with other wargaming friends. But it may conflict with a trip to see Sweden and our grandkids in situ of the first time.