For most of our existence, our AI & Big Data theme has accounted for 30-40% of our portfolio. The big idea is that it is (almost) inevitable that data will compound exponentially, thereby leading to a compounding of insights, and companies enabling this could be some of the safest high-return investments over the next decade. Ironically, investing in AI is a no-brainer.
When people think AI & Big Data, they normally think of software. But we found that the boring hardware workhorses behind this compounding could be safer, higher-return, lower-risk investments. Intel is the OG of these workhorses, otherwise called Semiconductors. And the company is still a massive force despite their recent troubles. We are not holders of Intel stock. We used to hold the stock for a short while, but we got out in a hurry. There were two main reasons:
- 1.They were (and still are) being outflanked by rivals, but more importantly…
- 2.We thought that the Apple decision of switching to custom ARM-based chips for its latest laptops started a movement that could make Intel a dinosaur.
However, some things have changed since then – in particular, there are 2 possible improvements to the Intel story:
- They have a new CEO – Pat Gelsinger – who’s a bona fide semiconductor guy, having spent the formative years of his career under the tutelage of Andy Grove, the legendary Intel CEO.
- Gelsinger wants to open up Intel – to manufacturing chips for other companies, including fast-growing rivals. This is a doozy. This is like Netflix announcing (last night) that they want to explore the advertising model.
Overall, Gelsinger seems to want to take Intel back to its halcyon days of being “the semiconductor company”. It’s a tough job, especially since Intel has been ceding ground to rivals in this new era of “ubiquitous computing”. This new chapter – that includes making chips for rivals – is an interesting proposition given the current geopolitical happenings. Does Gelsinger make a good case?
Intel is still the big boy in the semiconductor arena. But the stock hasn’t moved anywhere in years because investors are concerned about growth. Compared to its smaller, agile rivals, Intel looks like a sluggish old ship that could take years to turn in the right direction. Check out these 3-year stock returns:
When we had decided to invest in Intel in 2019, we were fully aware of Intel’s revenue growth sluggishness. However, we had found the stock cheap. It was a “value buy”. We don’t do pure “value buys” anymore, although that would have served us well over the last 12 months. We still believe that most “cheap” stocks are cheap for a good reason; it’s very rare to find one that’s cheap purely because of market apathy. But with Intel, there was a believable story attached to the cheap valuation. We believed that the market was underpricing Intel’s “XPU” plans – the ability to make different types of semiconductors talk to each other seamlessly in order to facilitate more intelligent software (aka AI). Gelsinger still believes Intel can do this.
But we exited INTC that same year, not because the XPU story fizzled, but because Apple made the momentous announcement that they will start furnishing their flagship Macbooks with their own custom-made ARM-based CPUs. Now, Apple itself wasn’t a massive part of Intel’s revenue stream. But the danger was (still is) that Apple had shown Microsoft that one needn’t depend on Intel’s x86 for desktop CPUs in this day and age of “ubiquitous” or “heterogenous” computing.
Microsoft has yet to make its move on this front; if it does decide to furnish its Surface “laplets” with custom-made ARM-based chips (made in conjunction with, say, Qualcomm), other Windows PCs manufacturers could be left out in the cold. We don’t know whether Microsoft would do this because (to state the obvious) Windows is built differently compared to Macintosh. But the possible Apple-ing of Microsoft (Bill Gates would hate that phrase) could spell doom for Intel. On a sidenote, why would Microsoft want to Apple-ify? Maybe this…
So, we got out of Intel because under then-CEO Bob Swan, Intel didn’t have a robust response to this threat of insourcing chip design (which in hindsight was very well executed by Apple). Investors got no assurance that:
- 1.Intel’s X86 architecture will still beat out ARM-based architecture – comprehensively – in this brave new world of heterogenous computing.
- 2.Intel will have compelling GPUs and FPGAs to compete with Nvidia and Xilinx respectively, with a software ecosystem to rival that of Nvidia’s CUDA.
But that was all pre-Gelsinger. He’s giving investors some hope.
Gelsinger started his career in Intel when Intel was new and shiny. He learned his chops under the tutelage of Andy Grove. He practically has Intel inside him, in his DNA! That, in our view, is a big upgrade from Bob Swan, who was (is) a finance guy. No offense, Bob. Under Swan (and his predecessor, to be fair), Intel ceded ground to AMD on CPUs, to Nvidia on GPUs, and they allowed Apple to even think they could make their own CPUs for Macbooks. This chart represents everything that Gelsinger wants to change:
Gelsinger’s strategy – as we understood it from their recent Investor Day – can be summarized as follows:
- 1.Get Intel internal manufacturing up to speed to be back on the cutting edge of the “node wars” with TSMC and Samsung.
- 2.Think ahead of the curve – focus on System-on-Chip architectures – that combine Intel’s capabilities in CPUs, GPUs and FPGAs under one unified software ecosystem.
- 3.Open up Intel’s extensive semiconductor manufacturing capabilities to other companies that have “fabless” models.
The last one – about opening up manufacturing – is the big departure from previous Intel regimes.
Gelsinger is right about one thing: The world needs semiconductor fabs in the western hemisphere – in countries that are relatively safe from the scornful gaze of autocrats. Intel is in the best position to offer an alternative. Putin’s invasion of Ukraine has reinforced this view. Other American companies like GlobalFoundries now have a new raison d’etre – like Intel, they were completely outflanked by TSMC and Samsung in the last few years. But the world needs an alternative, fast! And Gelsinger wants to oblige.
Here’s a chart from Gelsinger’s presentation during Intel’s Investor Day 2022:
While we sat through the whole (not that boring) presentation for 3 hours, we noticed that Gelsinger and team spoke of the Foundry business for at least 25% of the time. That’s because this is clearly a massive shift from Intel’s traditional “closed-loop” business model. They used to manufacture logic semiconductors only for themselves. Compared to AMD, that was Intel’s big differentiator: a vertically integrated company a la Apple. How self-centered? But now they want to be the TSMC of the west. This shift in intention has been received well, in general. Qualcomm and Nvidia have signaled their openness to collaborating with Intel on manufacturing. It’s understandable – they want to diversify their risk even if it means collaborating with a rival AMD’s position on this is not clear.
Qualcomm, Nvidia, and AMD (Intel’s main pain in the a**) all have a “fabless” model. They focus only on R&D and designing chips (and the software required to optimize performance of those chips). They leave the nuts-and-bolts manufacturing headaches to TSMC or Samsung. This model took off in the last 5 years, with these companies becoming more creative with their designs specially to accommodate Cloud Datacenter workloads that are becoming more “intelligent” by the day. TSMC has been godsend for them – it takes an enormous level of talent, investment, skill, and creativity to make AI-ready chips at the cutting-edge of Moore’s Law. Intel does both – design and manufacturing. While the biggest market for semiconductors was PCs, this worked well for Intel. But as smartphones became, well, smarter, and the double trouble of Cloud Computing and AI hit mainstream, Intel’s been ceding ground.
TSMC has dominated the semiconductor manufacturing space, taking upon itself the manufacturing headaches for AMD, Qualcomm, Apple, Nvidia and others:
Gelsinger probably has this chart etched in his brain. Being “America’s Fab” could be a massive revenue windfall, given the current direction of geopolitical winds.
Intel: Back inside?
Gelsinger and his team have high hopes. They expect 2023 and 2024 to be “investment years” followed by double-digit revenue growth numbers from 2026 onwards. We believe that this is a plausible scenario.
Let’s assume that Intel’s Management executes like a well-oiled machine. In that case, what should Intel be worth in the market? Let’s work this through based on Management guidance for 2026:
- Revenue of $114 billion (from $79bn currently)
- Gross Margin of 55%
- FCF margin of 20%
We normally attach a 20X multiple on free cash flow of solid, steady, predictably growing companies. That would imply a valuation of $454 billion. That’s a 132% premium over current valuation. That would make Intel stock a steal today! Obviously, that would be true if management is able to convince American fabless semiconductor companies to trust them. That is still a big IF.
Manufacturing cutting-edge semiconductors for others ain’t a walk in the park. Only 2 companies in the world – TSMC and Samsung – have mastered it.
Cloudy, with a chance of ****
There are several reasons why Intel might never make it back to the top. Here are just a few:
- China ends hostility towards Taiwan. This would put a dagger through Gelsinger’s geopolitical case for investing in Intel.
- TSMC shifts most of its manufacturing to “safe” zones like Arizona or Europe – another dagger in Gelsinger’s “geopolitical hedge” pitch.
- “Ubiquitous Computing”, as Gelsinger calls it, means that the CPU will just not be as critical to computing as it used to be – a lot of the workload will be handled by GPUs, FPGAs, SOCs, and/or ASICs. And the chips could be anywhere – on a Cloud server, on a desktop or a smartphone. Intel’s main bread-and-butter, CPUs, are losing some of their utility.
- Intel’s x86 CPU architecture could be deemed as “overkill” if a lot of computational workload shifts to GPUs, FPGAs etc.; Apple certainly thought so w.r.t their laptops.
- The AMD + Xilinx combo may become the premier choice for AI-ready SOCs (system-on-chips). Intel may be too late to the AI party.
- Nvidia could make GPUs programmable enough, that cutting-edge CPUs are no longer needed. Intel is clearly too late to the GPU party.
- Becoming a semiconductor manufacturer in a post-x86, post-CPU world could significantly dilute Intel’s margins and profitability. They may not be as cost-effective as TSMC.
The worst-case scenario would be “all of the above” – unfortunately, this perfect storm is not so far-fetched. A lot of the list above is already in motion. With the way things are going now, in 5 years Intel would be a day late and a dollar short to the Ubiquitous Computing party.
There is one other factor that Gelsinger forgot to mention completely during his geopolitical hedge pitch: That about 27% of Intel’s revenue comes from China! This might defeat the purpose of having Intel as a geopolitical hedge. If China does something as belligerent as Russia, what if there is an embargo on semiconductors? That would knock out 27% of Intel’s business right away!
But let’s give it a fair shake. The China factor is a problem with almost all semiconductor companies.
If China invades Taiwan, the whole semiconductor industry will get pummeled. Let’s zoom out a bit. Here are some dominators in their respective fields, along with their China + Taiwan exposure. In fact, we should do this for our entire portfolio, and for every company we analyze in the future. With China and Taiwan, we get the feeling it’s not a matter of IF, but a matter of WHEN.
This is a pretty dire scenario. Apart from Apple, all these semiconductor companies (and their stocks) would face severe losses if China invades Taiwan. To be fair, Intel looks pretty good in this comparison. In a previous section, we saw a comparison of Intel’s and TSMC’s margins. But to analyze Intel’s investability, let us put it in more intuitive terms:
This chart gives us some context about the “positive optionality” attached to Intel’s Foundry play. What do we need to believe about Intel’s growth over the next 5 years to buy the stock today? The crux of any Intel thesis, in our opinion, should be this: How big does Intel’s new Foundry business need to be to compensate for both the China factor and the continued loss of market share in CPUs?
- Let’s assume Intel’s core business underperforms the broader market (ceding market share). So, core revenue grows at a modest 4% per year (as it has been over the last 3 years).
- Let’s subtract out 27% of Intel’s business (the China portion).
- But to buy Intel stock at current prices, we need to believe that revenue can grow at 10% per year, based on the current stock price. This number will change when we take out the 27% attributable to China. But for now, let’s minimize the number of moving parts.
- Now, let’s say that Intel’s Foundry business grows like TSMC’s – at about 18% per year over the next 5 years. In reality, Intel won’t have a Foundry business until 2025. But let’s keep playing this out.
- What would be the size of the Foundry business in terms of % of revenue to arrive at that blended revenue growth number of about 10%?
- With this sort of analysis, we’ll need to make some basic assumptions. Let’s assume that the unit economics of Intel’s core business and its new Foundry business will be similar – so, similar profit margins. This assumption will likely be reexamined when doing a deep dive. But let’s play for a minute.
Now, in our Watch List, the “revenue GROWTH we need to believe” of 10% corresponds to a “REVENUE we’d need to believe” of $124 billion after 5 years. So, that’s our goal. Now, with the assumptions about segment revenue growth rates we’ve made above, what would the size of the Foundry business need to be today, if Intel suddenly loses all its China business?
It turns out, after some elementary math, that the Foundry business would need a starting point of about $25 billion. To put it in context, TSMC’s revenue is about $58 billion; only about 10% of that comes from China. So, Intel would need to snag nearly 50% of TSMC’s business (if, say, heaven forbid, TSMC is abruptly nationalized by China) in this new world.
Would you believe the Intel can do this – snag 50% of TSMC’s business? Obviously, Samsung would keep the rest. Imagine this crazy world in which semiconductors – the unsung heroes of all the gadgets we love – are suddenly in flux because China invaded Taiwan. The arc of global technological progress slows. We won’t get nicer things for years. And that’s assuming there is no direct confrontation between nuclear powers. Putin’s belligerence is making us think the unthinkable.
Intel looks like it could be a credible hedge to the China Factor in semis. While TSMC’s revenue stream is not heavily exposed to China, its existence is. We’d like to sleep well at night, so we’ll probably offload our TSMC bet over the next few months altogether. Sure, TSMC’s recent results were fantastic, and we expect that to continue until China does something nasty. So, the good news is that we’ve probably got time to offload our TSMC position at our own pace, and to think about replacing it with Intel. But…
The main reason is that we want to reduce our China-Taiwan exposure. Our other compelling semiconductor bets (AMD, Qualcomm, Nvidia, Apple) are dependent on TSMC (and Samsung) AND are exposed to China directly. If they’re smart, they’d be thinking of re-allocating to Intel (just as we are). But they’d need to believe that Gelsinger & Team can build a semiconductor fab on par with TSMC. This is a huge, complicated project.
Our focus in this new world of belligerent autocrats would be to reduce our China-Taiwan exposure as much as possible. To that end, we’ll keep Intel in the mix of some of our other possible buys. We have not done a deep dive on most of these possible buys, so this is not a recommendation to buy. But we’ll be analyzing these companies over the next few weeks (spoiler: none of these have much China/Taiwan exposure):
- Salesforce (waiting for a 5-10% price drop)
- Zoom Communications
One thing is for sure: our semiconductor exposure will decrease over the next few months. Ultimately, we don’t want to lose even one good night’s sleep by playing with dragon fire.
Let’s keep investing in progress, while sleeping well.
Many Happy Returns.