Good morning,
In yesterday’s Update I said that “a CAGR calculation from before the pandemic includes all of the pandemic growth”; I should have been more precise and made clear that I was referring to a CAGR calculation that started its measurement from just before the pandemic, and thus included the pandemic.
On to the update:
AMD (and Intel) Earnings
From Bloomberg:
Advanced Micro Devices Inc. gave a strong sales forecast for the current quarter, indicating that the chipmaker continues to make strides in its most lucrative market: data-center processors. AMD predicted second-quarter sales of roughly $6.5 billion, compared with an average analyst estimate of $6.03 billion. That helped send the shares up as much as 8.3% in late trading Tuesday.
The outlook helped allay concerns that the chip market is slowing — and signaled that AMD is making further gains on Intel Corp. The company, which for years lagged far behind Intel in computer processors, is on pace to end 2022 with almost four times as much revenue as in 2019. New products and better execution have helped AMD win over customers who were once skeptical about its capabilities…
Despite AMD’s rapid growth, investors have shunned the stock this year, part of a broader pullback for semiconductor shares. Investors have been particularly wary of chipmakers that made rapid gains over the past three years, fearing that a collapse is near. AMD closed at $91.13 in New York Tuesday, down 37% this year.
I’ll take Bloomberg’s word about investor worries — I’m a company analyst, not a shareholder psychologist — but I will say that as a multi-decade observer of technology I get the sentiment: AMD has always done best in boom periods, particularly when it could fill in demand that Intel either couldn’t cover (because it didn’t have capacity) or didn’t want to cover (because it was cheap); on the flipside, any slowdown in demand meant that AMD was in trouble. Remember, making chips is a fixed cost endeavor, with minimal marginal costs for each additional chip; that meant that the pain Intel felt during slowdowns was lower prices, which left AMD with no market at all.
That was then, this is now, and to that end, I thought that AMD’s comments about slowing PC growth — which hurt Intel’s forecast last week — were particularly striking. CEO Lisa Su said on the earnings call:
Although the PC market is experiencing some softness coming off multiple quarters of near-record unit shipments, our focus remains on the premium, gaming and commercial portions of the market where we see strong growth opportunities and we expect to continue gaining overall client revenue share.
Did you catch that? AMD is actually expecting to grow its client revenue even as the category declines as a whole because it expects to take share in the higher ends of the market. AMD just makes better chips now, and its increased pricing power reflects that; for the first time in a long time (I think the first time ever, but these two companies have been around for a long time, so I can’t be certain), AMD (ex-Xilinx) had better gross margins than Intel:
This gap is expected to grow, particularly once Xilinx is included (which is fair, given that Intel’s results include Altera). I also thought this bit from the Bloomberg article was interesting:
AMD is gaining ground at the largest buyers of computer processors, owners of the giant data centers that are the backbone of the internet. Some 48% of all new processors installed in these data centers were bought from AMD in March, according to Jefferies & Co. analyst Mark Lipacis.
I am looking forward to AMD breaking out these numbers starting next quarter; for now, remember that Intel’s data center business shrunk in 2021 (although it was up this quarter). This isn’t a surprise: given their scale hyperscalers are going to care more about differences in performance, and be willing to overcome any software issues necessary to support the better product (unlike, say, an enterprise or government agency, where Intel still dominates); if hyperscalers are buying AMD, it’s a safe bet that it is better on a total cost of ownership basis (which incorporates price, performance, and energy use). Oh, and these chips are the most expensive, with the highest margins — it’s the server business that is driving the chart above.
TSMC Earnings
A few weeks ago, again from Bloomberg:
Taiwan Semiconductor Manufacturing Co. raised its sales outlook for the year after quarterly earnings jumped 45%, helped by solid demand for chips used in everything from smartphones to cars. Annual revenue in dollar terms will top the previous outlook for as much as 20%-plus growth, the world’s biggest contract manufacturer of chips said Thursday. Sales will rise to $17.6 billion to $18.2 billion in the quarter through June, it said, implying growth of more than 30%. Analysts were estimating $16.9 billion on average, according to data compiled by Bloomberg.
The company also predicted wider earnings margins, signaling sustained demand for mobile phones, smart televisions and other gadgets from makers such as Apple Inc. and Samsung Electronics Co. even as consumers exit pandemic-era work-from-home arrangements. Meanwhile a chip shortage is yet to ease — the wait times for semiconductor delivery grew again in March due to China’s Covid lockdowns and a Japan earthquake that hit production, according to research by Susquehanna Financial Group…
Gross margin, or what’s left of sales after production costs are deducted, will expand to 56% to 58% this quarter from 55.6% in the first quarter, TSMC predicted. That’s the widest in at least a decade.
TSMC is another company which has historically trailed Intel in gross margins, which makes sense: in a modular value chain both the designer and manufacturer of a chip need to collect their margin; Intel, as an integrated supplier, should capture both, and for many years its margins reflected that reality: Intel was in the low 60% range for most of the last decade, while TSMC was in the high 40% range (AMD was in the 30% range, although hit a nadir of 23% in 2016). Today, though, both TSMC and AMD have higher margins than Intel does, despite the fact that TSMC is making AMD’s chips!
That language, though, isn’t quite right: while AMD’s designs have dramatically improved, a major reason for AMD’s pricing power is because TSMC is manufacturing their chips; that gives TSMC pricing power of its own, which it is increasingly leveraging. Tim Culpan wrote in Bloomberg:
“Gross margin increased mainly due to cost improvement and value-selling efforts,” Chief Financial Officer Wendell Huang told investors Thursday…that “value-selling” he refers to is TSMC’s ability to convince customers that it’s worth spending more money for their world-beating manufacturing technology. And clients like Apple Inc., Nvidia Corp., Advanced Micro Devices Inc. and Intel Corp. are buying it. In fact, TSMC was able to raise rates by the most in over three years. The Taiwanese company doesn’t release prices, but we can approximate it from the data it does provide. For the March quarter, it brought in $4,650 for every 12-inch wafer it churned out for customers. That’s 10% higher than the prior quarter.
In the past, such stronger pricing could be explained by technology mix. A higher ratio of more-advanced products raises that average, and vice versa. This time, however, its best offering — the 5 nanometer node — experienced an overall decline in sales from the prior quarter, accounting for 20% of revenue from 23%. There’s no need to worry about the drop — Apple is the largest buyer of these chips for its iPhones, and the March quarter is a lull in the annual smartphone cycle. We did, however, see a 20% rise in sales for the slightly older, but still very advanced, 7nm product. TSMC posted record sales of chips used in high performance computing, which includes advanced graphics and artificial intelligence, that used this manufacturing node. Clients like Nvidia and AMD are major customers and are doing a roaring trade thanks to continued demand for machine learning and cloud computing…
Investors, and clients, should expect that pricing strength to remain, with Chief Executive Officer C.C. Wei on Thursday noting the company won’t cut prices, even in a downturn.
This is why Wei had no problem waving away investor concerns on the earnings call about TSMC making chips for Intel.
Let me answer the question first on the competition. As a leading pure-play foundry, TSMC have never been short on competition in our 35-year history. And we know how to compete, all right? And you asked about how to protect the TSMC’s IP or technology detail. In fact, we have a well-established process and design enablement system to ensure a productive engagement with all our customers where we can protect our own IP as well as customers’ IP. We do not anticipate any issues at all. And for the future [taking business back]…we have already taken this into our capacity planning consideration.
Remember, TSMC benefits from Intel as a customer in multiple ways: they not only gain more revenue, which they can plow back into the next node, but also get that much more competition for leading edge nodes; no wonder Wei is confident TSMC won’t have to cut prices anytime soon. The real bet, though, is that Intel never catches up, and has to stay with TSMC to avoid AMD continuing to eat its lunch.
Morris Chang on U.S. Semiconductor Production
TSMC founder Morris Chang (who is now retired) appeared on a Brookings Institute podcast and had this to say about the U.S.’s efforts to bring chip production back onshore:
TSMC has decided to invest in a $12 billion foundry that is under construction outside of Phoenix. You’ve made headlines in the past talking about the challenges the United States may have developing a semiconductor supply chain of its own that is competitive with those in Taiwan and in Asia. What do you see as some of America’s top limitations in its ability to ramp up domestic semiconductor production right now?
There’s a lack of manufacturing talent to begin with. I don’t really think it’s a bad thing for the United States actually, but it’s a bad thing for trying to do semiconductor manufacturing in the U.S. We have actually had a manufacturing plant in Oregon for 25 years, and 25 years, that’s a long time. We sent all kinds of people, we changed the managers, changed the engineers, we used both American local engineers and we also send engineers from Taiwan to Oregon to try to improve the performance. Improvement in its performance has happened, however the cost difference between Taiwan manufacturing and Oregon manufacturing has remained about the same. The same product, the Oregon cost is about 50% more than the Taiwan cost. Well of course for us the Oregon product is still profitable — although not nearly as profitable as the Taiwan product — so still we have maintained it.
We started in 1997. and initially it was chaos. It was just a series of ugly surprises, because when we first went in we really expected the cost to be comparable to Taiwan, and that was extremely naive. But after a few years of trying to make it work, we have to settle down and we have to accept it, and since it was still profitable of course we still accepted it, but we didn’t expand it. That was Oregon, and we still have about a thousand workers in that factory, and that factory costs us about 50 percent more than Taiwan.
Now Arizona, that would be a bigger scale venture and bigger scale manufacturing than the one in Oregon. Much more advanced technology, etc., and of course we did it at the urging of the U.S. government, and we felt that we should do it. Basically I was already retired, so at that time the decision was made, the decision was made by the current chairman. But anyway, we think that the recent effort of the U.S. to increase onshore manufacturing of semiconductors, right now you are talking about spending only tens of billions of dollars of money of subsidy, well it’s not going to be enough. I think it will be a very expensive exercise in futility. The U.S. will increase onshore manufacturing of semiconductors somewhat, but all that will be very high unit cost, it’ll be non-competitive in the world market, when you compete with factories like TSMC.
Right now I think the U.S. has a very good position in semiconductor technology design. The U.S. has got most of the design capability in the world, the best design capability in the world. Taiwan has very little — TSMC has none, but there are a few companies in Taiwan that also do design, but they are not nearly as advanced as some of the U.S. companies. Now of course there are people who point out that maybe Taiwan is not safe. Now that’s of course another topic. I’m assuming that there will not be any war. Frankly, if there is a war in Taiwan Strait then IF think the United States will have more than chips to worry about. Now if there’s no war, then I think the effort to increase onshore manufacturing of semiconductors, is a wasteful and expensive exercise in futility. If there is war then my goodness we all have a lot more than just chips to worry about.
Chang actually understates the issue: not only will any U.S.-based foundry not be competitive on a unit cost basis, but the ones not built by TSMC won’t have pricing power either; if anything they will have to price their products lower to compete, magnifying the cost issue. Chang is right: if you ignore the war issue then there really is no logical reason to compete with TSMC at this point.
That, though, is the rub: if a war comes it will already be too late, which is exactly why Intel CEO Pat Gelsinger keeps invoking this exact risk in his campaign for subsidies; then again, perhaps the reality of TSMC dependence is exactly what will prevent a war in the first place.
This Update will be available as a podcast later today. To receive it in your podcast player, visit Stratechery.
The Stratechery Update is intended for a single recipient, but occasional forwarding is totally fine! If you would like to order multiple subscriptions for your team with a group discount (minimum 5), please contact me directly.
Thanks for being a subscriber, and have a great day!