What is going on at Intel?

We do not spend that much time on Intel, but it has been a hot topic lately and so we want to put down some of our thoughts on the company. We touched on some this on Ben Thompsons’s podcast a couple weeks back (paywalled link, but really, you should be subscribed to him by now).

Put simply, Intel is embarking on a major effort to save themselves. The company once defined the pinnacle of semiconductor manufacturing excellence, the standard by which everyone else measured themselves. But it has fallen off Moore’s Law and now needs to catch up in a hurry. (Here is some background material on this and here and here. Maybe we do spend a lot of time on Intel.) Intel has set itself two tasks – catch up and regain its manufacturing lead and to open its fab to third party customers – Intel Foundry Services (IFS). Intel has set itself a 2024 goal for both of those.

The ostensible reason for our appearance on Stratechery was Intel’s earnings call held a few hours before we recorded. Those results were a mix. In places where Intel’s manufacturing deficit is less important the numbers looked good. In places where that deficit is important the results were not great – the fifth straight quarter of declining sales to cloud service providers. for example. To be fair, they are on a long journey which has just begun. In that light, the results at the very least demonstrated that nothing else is going wrong. The ship is sinking, but slowly and the crew is moving to plug the leak.

All of this is of course playing out against the backdrop of a global chip shortage. Intel’s chip shortages are largely self-induced (see above re:manufacturing capabilities), but this has not stopped them from leading the charge for the US government to subsidize their expansion plans. We will leave it to the policy people to tell us if all those subsidies will do anything to lower chip costs, but Intel is using the funding to partially foot the bill for a massive expansion with new manufacturing capacity coming on stream in Arizona and Ohio in coming years. Of course, if they cannot catch up technologically those plants will look fairly empty, but that is a problem for a different day.

A question we get frequently is what kind of early indicators will we see to know if Intel is achieving its targets? Given that Intel’s problems are largely tied up with its manufacturing process it will be tricky for outsiders to gain any signal until fairly late in the game. So the company itself will be the main source of information. Beyond the typical corporate presentations, we will likely see some technical information seep out through company presentations at big industry conferences. Of course, Intel is long accustomed to this, and we are unlikely to see leaks of any data Intel does not want us to see. That being said, the company will need some early outside help in this process – from suppliers and sub-suppliers. Technical conferences are one way they can seek this support. If a year from now a senior Intel engineer presents at HotChips or Semicon with no new information that would qualify as a warning flag. On the other hand, if they demonstrate some new nano-structure or innovative process that would be sign of encouragement.

Another potential indicator will be comments from other chip companies. IFS will need external customers. Intel has announced a few of these already, but dig a bit and it becomes clear that none of these will really be paying customers for several years. For those customers to actually move to IFS, they will need to start work on that far earlier than production. So at some point in 2023 if some large chip company drops hints about taking on a new foundry partner it will be fairly clear who they are talking about.

The biggest challenge Intel faces, and the hardest to judge from the outside, will be changing it’s culture. Transitioning from a vertically integrated technology leader to a scrappy street fighter grasping for every straw at the edge of the abyss is a big cultural shift. The company needs to make a lot of changes. We have seen some of this already, with a large number of significant personnel changes among the senior ranks. So far, most of these have been encouraging, but the real test will be how much the new mentality trickles down to the rank and file. For this, look to swings in employment or any outflow from the company – someone can generate Alpha by mining LinkedIn. Even here, the signal will not be entirely clear. There is a big war for semiconductor talent raging right now. With junior level employees being poached with massive paychcecks and stock comp. This is probably not sustainable for the industry for much longer, but it has already seen big talent drains out of established companies to “new entrants” like Google and Meta.

To sum up, Intel is not in a great position, but so far at least it is making the right noises.

3 responses to “What is going on at Intel?

  1. The best leading indicator that Intel is starting to right its manufacturing woes will be a significant jump in shipments from ASML. Thus far, North America has accounted for less than 5% of their leading-edge EUV production. Meanwhile, over 50% of EUV machines are going to TSMC, racking them up millions of man-hours of operating experience. Gelsinger’s claim that they’ll be back in the thick of things with high NA EUV machines is a total joke. As far as I can tell, Intel will be the leading manufacturer of n-1 to n-3 products while outsourcing their most critical leading-edge products to TSMC. This is a losing proposition and AMD will continue to destroy them in their highest margin data center segment.

  2. I can’t get past Otellini and Krzanich paying out $58B in dividends 2005-19 while the technology engine that made Intel leaders drove into a ditch. The investment in Inpria appears to have paid off, but it’s still years from being production-ready – and Intel didn’t move to control it, instead agreeing to let JSR acquire.

    I love all the sudden interest in “saving the US semiconductor industry” from Congress, but after funding the FinFET work through DARPA the US government went back to relying on venture capital for its national technology roadmap. That got us apps…

Leave a Reply