Intel Stock: Turnaround Starts In 2025 (NASDAQ:INTC)

Investment Thesis

In my recent coverage of Intel (NASDAQ:INTC) and its Q2 results, I claimed that 2022 would likely mark the bottom, both financially as well as technologically. Perhaps in hindsight also for the stock. As such, with the stock trading for a rock bottom price, it may be fruitful to look at what investors could get in a few years once Intel gets out of its issues and the many investments that have been started by Pat Gelsinger over the last 18 months become more and more visible.

In that regard, I would propose that the future looks quite bright. Whereas Intel is currently faced with declining gross margins, revenue, free cash flow and profit, these trends are likely to reverse over the next few years. As mentioned in the summary, being a patient Intel turnaround investor also gets rewarded with a 4% divided currently.

Analysis

A few years ago, I called Intel a semiconductor conglomerate. Although Intel has since shed several of its businesses, such as NAND and recently Optane memory/storage, there remain many moving parts.

In short, though, ultimately better products will mean less market share loss (or even regaining market share) and the ability to charge higher prices, driving both revenue and gross margins back up.

1. Process technology

The technological bottom for Intel is 2022/2023 since Intel is currently in the final stages preparing for the Intel 4 ramp, which will replace the by now three year old 10nm tech (which itself was delayed by three years). To be fair, Intel 4 may never be a truly high volume node since Intel will use it for just one product, Meteor Lake, and there are actually recent rumors that Meteor Lake is delayed to the second half of 2023 or even later (although Intel just recently said Meteor Lake remains a 2023 product).

In any case, Intel 4 is just the second step of the “5 nodes in 4 years” to regain process leadership by 2025. The most often heard concern here is what confidence investors should have in Intel’s execution.

• First, Intel just confirmed for the fourth quarter in a row that all nodes through 18A remain on schedule.
• Secondly, from Intel 4 going forward Intel is finally using EUV lithography, with Intel in fact being the first to commercially deploy the next-gen “high-NA EUV” tool in 2025. Simply put, more advanced equipment should mean a more robust process development and higher yield (defect density).
• Thirdly, the competition has been having delivery slips of its own as well lately. TSMC (TSM) is late with N3, and its schedule for N2 (three years after N3) should be considered slow as well. In addition, both of these nodes are markedly below the 2x scaling factor that Moore’s Law is known for.
• Fourth, Intel was apparently so confident in its progress that at investor meeting in February, Intel pulled in the last of these nodes, 18A, to H2 2022.

To elaborate a bit further on the third bullet point, investors may observe that (1) it is taking TSMC just as long to go from N5 to N2 as it did for Intel to go from 14nm to 10nm. However, (2) going from N5 to N2 (despite being a jump of two nodes) yields a roughly 25% lower increase in transistor density than going from Intel 14nm to 10nm. In other words, TSMC’s progress in going from N5 to N2 is actually even slower than Intel at 10nm, which as mentioned was a node that was delayed by three years and caused Intel to lose its once firm process leadership.

Meanwhile, from Intel 4 to Intel 20A should end up as Intel’s quickest node transition in its history.

The improved execution in Intel’s process development is quite obvious. More quantitatively, my estimate for 18A is that it will have a transistor density of 800 million transistors per square mm (800MT abbreviated), while my estimate for TSMC is 3x lower at around 270MT. Hence, even if TSMC ends up a bit higher and Intel a bit lower, in terms of transistor density this should be a leading node. In addition, Intel’s claim of regaining process leadership was actually based on performance per watt.

Pat Gelsinger described this as Intel’s “true north star”. With Intel being back in the driver’s seat of Moore’s Law for the first time in eight years (although AMD (AMD) was able to capitalize on only a subset of this period), this will allow Intel to launch clear leadership products, hence this will ultimately act as flywheel that will lift the rest of the company out of its troubles.

Note that since Moore’s Law leadership is basically equivalent to cost per transistor leadership (more transistor per wafer means lower cost per transistor, even if the wafer becomes a bit more expensive), this will mean that Intel’s gross margins should recover nicely back towards the upper 50s or even the low 60s (respectively CFO Dave Zinsner’s official and unofficial goal), benefiting earnings and hence the stock.

2. Foundry

Process leadership not only acts a flywheel for Intel’s traditional business, but also for the new foundry business. Intel Foundry Services will already start with 18A foundry test chips this year, and Intel said recently it is engaged with six of the top 10 fabless companies. Quite obviously, Intel is positioned to build a new multibillion dollar business from scratch.

3. GPU

Investors may remember that GPUs are actually one of the prime beneficiaries of Moore’s Law given their inherent focus on parallelism, and hence Intel’s nascent GPU business, which has been plagued by issues, should benefit from process leadership as well, both in the data center and perhaps for gaming as well.

Note that due to the process delays, this business is a notable example of Intel increasing its use of foundries (Arc on N6, Ponte Vecchio on N5) so it remains to be seen what happens once Intel regains process leadership. Still, GPUs are one of the most obvious products for Intel to use its regained process leadership to its advantage for.

Intel has set itself a quite ambitious goal to grow this into a near-$10 billion business by 2026, although admittedly this should be taken by a grain of salt until there’s more evidence (since this implies a revenue ramp even steeper than AMD’s acclaimed Epyc business).

4. Mobileye

Aside from (or due to) the IPO, Intel has been relatively quiet about Mobileye, even though Mobileye reaccelerated to >40% growth in Q2. But more importantly, Mobileye is preparing for its own disruption, with the launch of its initial robotaxi business by the end of the year and the initial autonomous vehicles for consumers by 2024-2025.

My assumption remains that once this technology proves itself on the road, there will be an incredible demand for this technology as expensive human drivers are getting replaced by technology. For example, in robotaxis there is no reason why Mobileye couldn’t become as big as the likes of Uber (UBER), which means this is another multibillion dollar business in the making.

5. Data center

For completeness I will mention the data center, where Intel is obviously bleeding the most market share currently, as Alder Lake on the PC side is already a competitive product. (On the PC side, the only bastion that I see in my Twitter feed is that AMD has higher performance per watt, but clearly this is an esoteric metric that falls well outside what the average consumer would do research about, if the average consumer even does research at all. So for Intel’s real customers, which are OEMs, Alder Lake delivering leadership performance combined with Intel’s still higher brand value should be sufficient.)

Although Intel is making noise about Granite Rapids and Sierra Forest on Intel 3 in 2024, by definition of process leadership, if Intel can launch the successors of these products in 2025 on 18A, then those would be its real undisputed leadership data center CPUs.

Beyond 2025

In some ways, the turnaround by 2025 to regain process leadership is only just the beginning. I have previously likened it to AMD’s Zen launch in 2017, of which AMD and its investors continue to reap the benefits from to this day half a decade later.

Although Intel’s turnaround is unlikely to be quite as successful (although retaking process leadership would be a major feat, technologically, which many investors actually thought would never even be possible), this does imply Intel should be invested in with a time horizon to 2030 or beyond.

Risks

The main risk is that Intel already had a bunch of products in development before Pat Gelsinger joined. For example, the heavily delayed Sapphire Rapids Xeon went through no less than 12 “steppings” (revisions of the manufacturing blueprint), fixing 500 bugs in the process. In addition, Meteor Lake on de PC side seems to be delayed as well, although this seems to be due to issues with the GPU, which is made by TSMC (previously announced as using N3, but the node may even have changed).

Due to this, and the fact that it takes several quarters before any new product is ramped, as indicated in the previous paragraph 2025 should be the year where the initial seeds (which are currently being planted) become visible, while full harvest may only happen several years later.

On the other hand, while 18A will likely launch in H1 2025 if it goes in production in H2 2024 according to the roadmap, before that Intel already has Intel 3 and 20A, which although not discussed in detail should already be more competitive nodes. From that view, the more risk-averse investor may decide to wait until 2024 to see the first results, although this may perhaps come at the opportunity cost of having to get in at a higher cost basis.

Secondly, the risk is in the execution. While delays in specific product groups are always possible, the biggest execution risk is the process technology, since at best this would take away the advantage of having process leadership (in case of parity), or at worst could lead to further delays across the company. While Intel is clearly increasing its use of foundries (as evidenced by the GPU business), outsourcing everything will never be possible, and obviously Intel instead has entered the foundry business itself.

Investor Takeaway

As noted, investors are currently caught up in the downward spiral of Intel’s declining revenue, profits, (nonexistent) free cash flow, and gross margins (declining to nearly 35% in Q2 on the back of exiting the Optane business, or around 45% non-GAAP), all while capex and R&D are ballooning.

However, in simple terms, Intel actually has a very clear path to get out of this spiral. It is called process leadership. It will vastly improve Intel’s cost structure, while simultaneously allowing for Intel to create much more advanced products, which should be rewarded in the market by an improved market share position and higher average selling prices.

As argued 2022 and 2023 will most likely be the bottom technologically and financially, so investors who until now perhaps weren’t convinced may try some bottom fishing over the next year or so, before signs of Intel’s progress across many areas (including process, foundry, GPUs, Mobileye and more) becomes too obvious to be ignored by the stock market.