Intel's Foundry Ambitions and Updating the WFE Playbook
Just a quick update on WFE 🔥🔥🔥
Intel had a pretty awesome first month with Pat and is beginning to execute a bold vision to get back to manufacturing parity by the launch of new investment in their foundry and foundry services and a renewed commitment to their fabs and Intel CPUs.
I don’t want to replicate work where it isn’t needed, and I found that this wonderful (and free!) post by Ben Thompson really summarized everything well. I won’t offer much commentary on Intel besides that the boldness is what is needed, and that despite my excitement it is going to be a long road of execution (Pat said Grovian execution for a reason) before we can seriously believe the turnaround at Intel.
Skepticism is warranted, but we should still be cheering on “Intel unleashed”. The world needs a healthy semiconductor ecosystem, and America needs a domestic fab.
But I’m not really here to talk to you about Intel, but rather Semicap (semiconductor capital equipment). This is an update post to my previous WFE Playbook post. What I want to focus on is what Intel’s capex increases mean for WFE, and what their bold $19-20 billion in capex means for companies downstream. Below is their updated guidance, which brought down full-year FCF expectations and revenue guidance, but importantly raised their capex outlook.
That capex outlook compares to ~$14 billion last year (note that on a 2-year basis this is less impressive), which represents an impressive ~37% YoY increase in capex. In conjunction with the impressive ~50% increase in capex at TSMC, it looks all but certain that total fab spending will increase by ~24-25%. I believe this is the floor, and 30% WFE growth is possible. Below is my WFE Model comparing bottom-up capex guidance to top-down expectations.
I previously believed in ~20% YoY WFE in 2021 and now feel confident that 25% is more reasonable with an upside bias to possibly 30%. What’s more, is that CY2022 is likely to not be a contraction in WFE in my opinion. Given that Intel’s fab expansion will take many years of investment to pan out, and there seems like an imminent EU fab, I believe multi-year growth is now probable. Let’s discuss what this means in the near term.
Near Term Implications for Semicap
Even with consensus estimates recently rising, it looks like again that from a purely top-down consideration that semicap revenue estimates are too low to the tune of 1-4%. This assumes no market share gains and leans conservative. Most companies, especially the front-end manufacturers, tend to take a little bit of share during growth periods.
If the past is any expectation, it seems possible that most companies can add 100-300% bps over WFE growth. Additionally, if you followed my WFE playbook, I assume that there is flatshare maintenance and just assume segment growth, and from that perspective, it seems like revenue is still too low. I think to the tune of 2-4% in terms of revenue and 5%+ for EBIT if WFE is 25%, and 5-7%+ in revenue if WFE is closer to 30% (probable at this point). Let’s now turn it to the slightly longer view.
This is where I am starting to get really excited. First and foremost I think it’s almost impossible to believe the three simultaneously. This tweet is absolutely money.
You cannot believe that the supply shortage continues into 2022 and believe that most of the companies are over-earning. Meanwhile, the Intel announcement all but assures that CY 2022 will likely not be a WFE contraction year. The Arizona fab will likely not even be ready for tool additions until CY 2022. And if there continues to be a cyclical recovery, it’s likely that semis will benefit given their historical relationship to NGDP (it’s very high).
Next, I want to put a bit of attention to where the fabs are going, and that is not Taiwan. Taiwan is the absolute lowest-cost provider of leading-edge chips. There are huge local network effects with talent and providers that make deploying tools to Taiwan have a higher utilization rate with a large infrastructure to start production. Anywhere outside of Taiwan is dilutive to utilization, which is a sneaky way of increasing semicap intensity. Greenfield fabs in the US and EU will probably have worse tools per wafer than in the US. That’s great for semicap.
Lastly, I want to point to the clarity of demand. When TSMC originally expanded capex by 50%+ it was assumed that Intel outsourcing was going to be a large part of that. We now know that this is obviously not the case and there clearly seems to be a broad-based demand that is ex-Intel. I am going to trust the process here and say this is not vaporware but the fruition of many things I have been writing about. Namely heterogeneous compute and the huge incremental AI demand wave, and the nascent increase in semi content in Automotive. Broadcom and many other providers are saying they have the longest visibility they have ever had in terms of demand and future booking orders, with an immense focus on preventing double ordering. I believe that the demand wave is real.
I think if you put it all together it seems that semiconductor manufacturing is becoming more critical than ever, benefits not only from the increasing digitalization of our lives but AI and multiple other growth drivers, all while being awesome businesses with high cash flow margins and returns on capital. I understand the fears of cyclicality and “peak” WFE, but I think that given supply shortages and strong broad-based demand, this isn’t it.
Almost any given semicap seems like a great risk-reward right now. They are cyclical yet secular businesses with extremely deep moats (not just ASML folks!) and trade at market-ish multiples despite growing faster than average and likely should sustain faster growth in the future. Recent volatility has given investors pause for tech stocks, but it’s rare that a sub-industry goes down / stays flat on good news. I think that’s an opportunity.
This is more typical of a paid post but given the pause, I wanted to have a quick blurb on Intel’s capex. I don’t really have any spicy takes on Intel. Also to answer the people who have asked me to open up paid, I will not be doing that. It’s a bit of a headache for billing and I want to make sure my subs are not billed for time not on the clock. Ciao!