Intel has been grappling with a CPU shortage for months, due to production constraints associated with its difficult 10nm ramp as well as increased demand for its 14nm data center processors. The company announced during its last conference call that it would spend an additional $ 1B to increase 14nm production.
That shortage could ease in the first quarter of 2019, with component orders expected to tick up before the Lunar New Year in early February, according to DigiTimes. The site writes:
The sources said that Intel’s CPU supply shortfalls seen since the beginning of the third quarter 2018 have disrupted the shipment schedules of Taiwan PC and notebook ODMs. The US-China trade conflicts have further frustrated the supply chain players, including chipmakers, preventing them from enjoying the peak season effect in the third quarter.
Intel’s interim CEO Robert Swan spoke at the 22nd annual Credit Suisse TMT Conference on November 27 and addressed some of these concerns. He reiterated that Intel was on-track to ramp 10nm for holidays 2019, before stating that FPGAs would also meet that guidance, with servers “a relatively fast follow on 10 nm, early in 2020. So, we’ve compressed the timeframe to go from PC to server. So, an increasing portion of our business, as we get into 2020, will be transitioned over to 10 nm.”
This implies that rumors of a limited or mobile-only 10nm ramp could be incorrect. Intel could still plan to roll over to 7nm more quickly than its typical cadence, but the company has released no details on that. Swan did say that Intel’s data center business is on track to grow 20 percent for full year 2018, and that this represents twice the full-year growth that the company expected back in 2017. ASPs have grown faster than unit shipments — the latter were up in the high single digits, compared with double-digit growth in ASPs.
Regarding component shortages, Swan stated:
[W]ith the push of 10 nm going out a little bit, the supply that we’re counting coming from 10 nm was not actually going to happen. So, we got ourselves in a position where we’re supply constrained and we’ve done a variety of things to get to the position where we can unequivocally say to our customers that we will not be constrained on your growth. So, we spent a billion and a half more capital this year to ramp more 14 nm capacity. We’ve redeployed some of our 10 nm equipment that we had back to 14 nm. So, we’re getting more output and we’re tackling the constraint problem to make it go away as quickly as we possibly can.
This generally confirms our original theory about where Intel’s 14nm shortage came from, particularly once you include the increased demand for high-end, wafer-hungry server CPUs. The one potential fly in this ointment is the ongoing trade battle between the US and China. This week, President Trump threatened to slap tariffs on every good imported from China, including an estimated $ 267B worth of goods currently exempt from those tariffs. While this wouldn’t impact the price of Intel CPUs directly, it could very much impact the price of the servers those systems are sold in.
How this would impact the overall market is uncertain. Beyond stating that people “could stand” a 10 percent iPhone tariff, President Trump has not indicated if he would seek a 10 or 25 percent tariff on additional Chinese goods if he is unable to reach a resolution with the Chinese. As we’ve previously discussed, the impact of tariffs always depends somewhat on which goods are taxed and when the tax is assessed. Up until now, laptops and desktops built by the major OEMs have been exempt from tariffs, while components like GPUs haven’t been. Any change to this policy could impact system prices in the not-too-distant future. While the higher margins on servers probably give companies like Dell and HP some leeway to absorb costs, it’s entirely possible that businesses would simply choose to defray their upgrade cycles for a time, assuming that the US and China will work things out sooner as opposed to later.