Intel announced its Q3 earnings results with very strong numbers and an acknowledgment that the company is still facing constraints on its 14nm business. Intel smashed guidance in Q3 2019, beating Q3 revenue expectations by $1.2B, with $19.2B in revenue. According to CEO Bob Swan, Intel has made significant capacity investments over the past year and increased its overall wafer production on 14nm by 25 percent. This, however, has not been sufficient to meet increased demand for 14nm chips.
One important thing to remember is that increasing capacity by 25 percent doesn’t directly translate to 25 percent more chips. Building 25 percent more wafers would only translate to a 25 percent increase in chip throughput if all CPUs
are the same size. At the same time that Intel has increased manufacturing, it’s also increased core counts. Intel’s desktop and laptop chips now field more cores than they did several years ago, while it’s also launched the Cascade Lake-AP family, which offers far more cores. We don’t know what the exact relationship is between Intel’s 25 percent wafer production increase and its total number of CPU shipments.
Here’s Bob Swan:
We expect our second-half PC client supply will be up double-digits compared to the first-half. And we expect to further increase our PC client supply by mid-to-high single-digits in 2020. But that growth hasn’t been sufficient.
We’re letting our customers down. And they’re expecting more from us. PC demand has exceeded our expectations and surpassed third-party forecasts. We now think the market is stronger than we forecasted back in Q2, which has made building inventory buffers difficult. We are working hard to regain supply demand balance. But we expect to continue to be challenged in the fourth quarter.
Elsewhere, Swan noted that Intel intends to increase capacity in 2020 by 25 percent over 2019, just as 2019 represented a 25 percent increase over 2018.
10nm and Beyond
There have been rumors that Intel’s entire desktop line of products on 10nm were canceled. Desktop products were not discussed in any particular on the call and no analysts asked about any specific issues related to 10nm product cancellation. This doesn’t mean 10nm wasn’t a topic of conversation. Much of the discussion on the call was related to capacity management and the difficulty of balancing 14nm versus 10nm production. According to Swan, Intel is now building 10nm in Oregon and Israel and will soon begin shipping volume production out of its fab in Arizona. Ice Lake and FPGA parts are shipping. In 2020, Swan said, “[W]e’ll continue to expand our 10-nanometer portfolio with exciting new products including an AI Inference Accelerator, 5G base station SoC, Xeon CPUs for server storage and network and a discrete GPU. This quarter we’ve achieved power on exit for our first discrete GPU DG1 an important milestone.”
None of this is a specific reference to any desktop parts, but Intel might not have a 10nm desktop chip. Way back in 2017, the 10nm graphs the company showed for per-transistor performance implied that 10nm+ or 10nm++ might be the replacement for 14nm++ on desktops. Intel has specifically said that 10nm hardware would come to desktop, but it hasn’t said which node. No details here, unfortunately, but the process is clearly going to be used for a number of other products at Intel.
This slide from 2017 shows the expected relationship between 10nm and 14nm++. Intel has not yet provided updated guidance.
Once it shifts to 7nm, Intel intends to slow the cadence push a bit. 5nm won’t be introduced until 2.5 years after 7nm, which puts it ~2024 (a mid-2021 launch for 7nm pushes 5nm into very late 2023 or early 2024). As far as competitive pressure from AMD, however, Intel isn’t currently concerned. When asked if it had seen much competitive impact from AMD outside of data centers, Intel implied the only pressure has been at the low end of the market, where its own supply issues are having an impact. In DCG, Swan said: “Look it’s a more competitive world. And in that world we just raised our full year outlook by $1.5 billion and increased our operating margins. So yeah, I think, competitively nothing’s really changed in the last three months, six months, nine months relative to what we expected. And our — the only thing that’s really changed is our performance.”
This is not necessarily demonstrative of much. AMD’s share of the server and data center market is still quite small and 7nm server hardware has only been in market for a few months. It’s entirely possible for AMD to be gaining market share without having a particularly large impact on Intel’s earnings at the moment. Intel’s desktop volume fell 11 percent, while laptop volume fell 10 percent, but ASPs in both categories rose by 3 percent and 4 percent respectively.
AMD announces its own Q3 earnings this week, so we’ll have a better sense how the company is performing once that data is available.
Now Read:
- Intel Acknowledges It Was ‘Too Aggressive’ With Its 10nm Plans
- Intel Is Finally Shipping Ice Lake in Volume
- Intel: CPU Shortage Will Extend Into Q3 2019
No comments:
Post a Comment