Intel's CPU supply shortage may continue throughout the year, or AMD may take more market share

Latest update time：2020-01-20

Reads：

According to foreign media reports, the well-known technology media DigiTimes recently released a latest report, saying that it expects Intel's CPU supply shortage problem to run throughout 2020, so many of its partners may switch to AMD's similar products.

This may not be particularly surprising. Intel has previously admitted that it is caught between a rock and a hard place, with CEO Bob Swan giving a very frank explanation of their current situation. However, it does mean that AMD will eat more market share from Intel, as OEMs and AIBs will have to switch to AMD parts to maintain sales because Intel's foundries are running at maximum capacity and still can't meet demand. Every chip order that Intel can't fill means new market share for AMD.

Even Intel's chips being shipped at a premium won't help. There's no point in taking steps to curb the premium now when supply is outstripping demand. OEMs/AIBs will have to pass these costs on to consumers who might prefer Intel products anyway. If there's one thing we can be sure of, it's that 2020 will be a make-or-break year for Intel, and things may not start to improve for the company until late 2021.

One thing Intel is working on right now is that, in theory, the progress of the 7nm process should not be affected by the delay of 10nm, as it is a separate process based on EUV lithography, and while it may not help Intel regain its lead, it should create a level playing field in terms of process technology. According to a report by market research firm Gartner, Intel recently regained the title of "the world's number one foundry company." This is an impressive achievement that shows how huge their foundry is now.

What we know so far is that the first 7nm product will be released in the fourth quarter of 2021, and Intel CEO Bob Swan explained what went wrong with the 10nm process. In addition, Swan also candidly answered the most pointed question, that is, how Intel got into a situation where its CPU market share was taken away by AMD and it could not meet the demand, which was in stark contrast to its long-standing philosophy of prioritizing mistakes, playing it safe, and always having idle manufacturing capacity.

Swan explained: "There are three reasons why we are in this predicament: First, we got into this predicament much faster than we expected, and the demand for CPUs and servers in 2018 also increased much faster than we expected. Remember, we entered 2018 expecting a 10% growth rate, but we actually increased by 21%, so the good news is that the demand for our products was much higher than we expected as we transformed into a data-centric company. "

He continued: “Second, we had 100% market share for smartphone modems, and we decided to build it in our fabs, so we took on more demand pressure. Third, to add insult to injury, we launched our 10nm process, and when that happened, we needed to add more and more performance to the previous generation, which meant more core counts and larger die sizes. Given those three factors - the ramp was much faster than we thought, producing modems in-house, and the delays to 10nm, we didn’t have flexible capacity. ”

While much of this is old news, this is the first time Intel has given a solid reason for why it couldn't keep up with demand, namely that it decided to bring smartphone modem production in-house, which in turn meant they couldn't focus on the CPU side of things. It's also a fairly plausible explanation for why Intel couldn't even keep up with demand at 14nm anymore and had to resort to scaling 22nm products.

When specifically asked to explain what went wrong, Swan responded candidly, acknowledging that Intel was overconfident in its ability to beat industry standards and suffered the consequences. Here is Bob's explanation, which he calls "scar tissue."

"The scar tissue actually started at the beginning of Moore's Law," Swan said. " And Moore's Law had worked for a long time, and we found that even though the physics were getting more and more challenging, we decided to set a higher bar for ourselves in terms of performance. So the 22nm to 14nm transition was not 2x transistor density, but 2.4x, which led to a lot of bumps along the way, but it worked, and that kind of worked gave us confidence. So why don't we increase the density to 2.7x when we move from 14nm to 10nm? The problem arises when you try to get higher and higher density performance. "

Swan went on to say: “Secondly, we are not going to try to do a 2.4x or 2.7x density increase as we did with 7nm. As you know, when we think about 5nm, we will reduce the density to 2x, which is consistent with historical trends. I also think that another positive aspect of the challenges of 10nm is that in the process, we learned how to make 14nm better. We have 14+, 14++, and even though we have been working on the same node for four years, the performance of the chip is now increasing as we reach 10nm process. ”

As Intel pursues 2x transistor density at 7nm and moves to EUV, the company appears ready to launch its first 7nm products (equivalent to TSMC’s 5nm) in the fourth quarter of 2021. Swan further stated that he expects to achieve 5nm (equivalent to TSMC’s 3nm) by 2024.

Source: Tencent Technology

Recommended Reading

Shelling Python? The person in charge of the domestic programming language "Mulan" apologized for exaggerating

Can SMIC take over Huawei's high-end chip foundry orders?

The battle for 5G chips?

Entering the 14nm world, what does SMIC's mass production mean?

AMD becomes TSMC's largest 7nm chip customer, replacing its old rival