Intel Remains In Its Investment Phase
Summary:
- Intel’s CEO departure is surprising as the company is on track with its 5 nodes in 4 years roadmap, crucial for regaining competitiveness.
- The board’s decision to replace Pat Gelsinger seems premature, as financial improvements were expected towards the end of the turnaround period (2024-25).
- Intel’s strategy focuses on technological advancements to improve profitability, with significant investments planned until 2025, aiming for manufacturing breakeven by 2027.
- Reports on 18A yield issues should be dismissed; Intel’s defect density metrics indicate progress comparable to TSMC’s successful nodes.
Investment Thesis
Intel’s (NASDAQ:INTC) board arguably had zero reasons to retire Pat Gelsinger as Intel remains on the course it must be on. Of course, its financials aren’t showing this yet, and Intel also hasn’t been successful in some of its newer businesses which aren’t much reliant on its process technology (robotaxis and AI). Nevertheless, since the company’s entire viability basically depended on its process technology, its improved execution in that regard had been most important.
Since so far there aren’t any further changes, the Intel thesis remains one of improved profitability and perhaps some more minor revenue growth in the years following the 18A launch (which has yet to happen), with Intel management for example having already predicted manufacturing breakeven by 2027.
Simply put, replacing the CEO won’t change any of these things, and it is equally likely he could improve the revenue side either. Hence, unless the board is preparing more drastic changes such as taking the company private of complete splitting the Foundry and Products sides, the whole exercise seems quite futile.
Background
Previous coverage of Intel already pointed out the issue: the technology and competitiveness has been improving quite strongly this year, but the profitability and other financial metrics not yet.
CEO Departure
On December 1, Intel announced the departure of CEO Pat Gelsinger. From what I’ve further read (not from Intel), it appears Pat Gelsinger got the choice between resigning or retiring. This implies it was the board’s decision to find a new CEO, and apparently quite quickly.
This is quite an important distinction, since the reverse would have suggested Pat Gelsinger himself, who came back after a decade hiatus from Intel to lead the turnaround, was throwing in the towel. It is not sure what the exact reasons would be for the board to make this decision. That is to say, the most obvious one is the stock price trend, which in turn is based on the revenue and profitability trends from the last several previous and expected future quarters.
However, basing such a drastic decision on such superficial a metric would be a giant mistake. The plan from that very start was that the turnaround would last until 2024-25, and it became very early apparent that it would be rather weighted towards the tail end of that 5-year range (2021-25). The reason is because the 5 nodes in 4 years roadmap basically ended up having the last 4 of those nodes in 2024-25.
In addition, as has been discussed extensively in most previous coverage during the last several years, it cannot be stressed enough that the return of technological competitiveness is both a necessity and required condition for Intel to have even any chance of improving its financials. Just imagine if AMD (AMD) had fired Lisa Su in 20216 already before its Zen architecture had launched (as it did in 2017). But that is exactly what the board has now done. The board seems to be judging the CEO based on financials rather than technology, which would be the same mistake as it has done in the last 1.5 decades, and indeed might be a cause of concern.
In addition, is not sure what a new CEO could really change, which might ultimately prove to render this whole hassle quite pointless: Intel already quickly implemented a reorganization to reduce its cost structure, obviously out of necessity. As just mentioned, in order to invest more, the financials would have to improve, for which the new CEO would have to wait for results of Pat Gelsinger’s investments. If the new CEO would invest less instead, then that would only increase the risk of Intel’s execution issues, which had led it to this place, to reappear. Improving its execution had been a key part of the strategy of each of the last three Intel CEOs, and Pat Gelsinger has been succeeding in this (technological) part.
Finally, it also not apparent what and how the new CEO could actually change in Intel’s strategy. In particular, Intel since BK had been investing in a myriad of areas, technologies and products in order to expand its TAM and increase its revenue and gross profit. Pat Gelsinger initially continued this pattern by adding the foundry business to Intel’s array of investments (which doesn’t mean Intel never exited any businesses, which it did with modems and NAND). However, since the 2022 downturn, Intel has exited about a dozen businesses – out of necessity – to focus more strongly on its core segments (PC, data center, network, edge, graphics and AI, Mobileye/Altera/IMS, and foundry).
18A Yield Reports
There has been quite a bit discussion/speculation lately about and in the wake of a report that suggested 18A had a yield of less than 10%, which (if it would be accurate) would suggest Intel is having issues with this node (the one supposed to return intel to leadership).
However, this report should never have appeared in the first place, as Intel itself had already provided a defect density number in September, at “less than 0.4 [defects per cm2]”. There are two ways to interpret this metric.
First, comparing this to similar data provided by TSMC, taking into account the node is still nearly a year from ramping in volume, suggests Intel at worst might be bit behind where TSMC was at that point in time for its N7 and N5 nodes, but those are two of its most successful nodes in its history, so it shouldn’t be a very large issue. In addition, TSMC’s chart suggests that the defect density drops to around 0.1 eventually, so a lot would still depend on the pace of further improvement.
Secondly, for a given chip, one can readily calculate its theoretical yield using a 0.4 defect density. Herein lies the issue, and it is the reason why defect density is a more insightful metric than yield: the yield depends on the size of the chip. The bigger the chip, the bigger the impact of one single defect on a wafer. It is, for example, like one of the reasons why Nvidia’s Blackwell process node is 4NP instead of N3 (which is an improvement based on the 4-year old N5 node): because Blackwell is a very large chip, it benefits from using this technology node since newer nodes (as just discussed) have a higher defect density.
In general, the first chips manufactured on new process nodes are much smaller (for example Apple’s (AAPL) iPhone chips) because those will have a higher yield. Therefore, the reports that said Intel has 10% yield on 18A might be considered misleading since Panther Lake and most likely Clearwater Forest as well will be much smaller chip(lets). For example, using the 0.4 defect density and a 10mm by 10mm chip (100mm2) results in a 68% yield. If the defect density would be reduced to 0.2 (which for TSMC took just a few quarters after achieving the 0.4 value), then the yield would improve further to 82%, should be quite viable (economical) to manufacture.
Earlier, in September, there had been a more concrete report by Reuters, which claimed that Broadcom (AVGO) had encountered issues with the 18A process, calling it a “setback”. However, the only specific point mentioned is, again, yield, which as just discussed seems to progressing on schedule. The report also cites comments by Intel and Broadcom:
A Broadcom spokesperson said the company is “evaluating the product and service offerings of Intel Foundry and have not concluded that evaluation.”
This echoes earlier comments by Intel on the Q2 earnings call, where management said the 1.0 PDK (process development kit) release had caused increased design activity. Intel has also previously hinted it didn’t necessarily expect very large external 18A revenue until beyond 2026, which makes sense given it takes several years to develop a chip. By that time, quite surely, the yield will (and should) have improved, solving this one single issue that both reports have pointed out.
In summary, defect density is a more general metric than yield, since yield depends on the size of the chip (among other factors), with larger chips generally having a lower yield (which is one of the reasons the industry has been moving towards chiplets). The defect density is one the most critical metrics as it might reduce the economics of a node or result in delays, but a comparison with TSMC’s provided data for N10/7/5 suggests that while, indeed, it needs to be reduced much further, at this point in development (pre-production), there does not seem to be a fundamental issue that suggests there will or might be a delay to (for example) Intel’s own 2025 18A-products.
Path forward
As also discussed previously, contrary to AMD whose strategy with Zen was quite forward (to take back market share), for Intel given that it still has quite a high market share (especially compared to AMD back then) and revenue, it isn’t all that obvious that it could easily return to strong growth in the wake of launching its 18A node and products.
Rather, the thesis would be some minor growth at best combined with a strong improvement in profitability. So not necessarily due to increasing leverage from strong revenue growth, but simply due to the improved economics/profitability profile of its more competitive nodes going forward.
Investor Takeaway
The CEO departure was quite surprising given that Intel has been successfully executing its 5 nodes in 4 years roadmap, with the product portfolio following (and increasing in competitiveness) as well. Hence, Intel was on track to its strategic imperative as a tech company, which is arguably the only thing that mattered for Intel given how far it had fallen behind TSMC (TSM).
It’s only after regaining competitiveness (which still isn’t fully the case with 18A still several quarters out) that the discussion can shift towards financial targets. Of course, as a public company Intel has already been required to do so far earlier (set profitability targets), but overall Intel had been quite clear that it would be mostly investing until 2025, with at best about breakeven free cash flow.
Analyst’s Disclosure: I/we have a beneficial long position in the shares of INTC either through stock ownership, options, or other derivatives. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Seeking Alpha’s Disclosure: Past performance is no guarantee of future results. No recommendation or advice is being given as to whether any investment is suitable for a particular investor. Any views or opinions expressed above may not reflect those of Seeking Alpha as a whole. Seeking Alpha is not a licensed securities dealer, broker or US investment adviser or investment bank. Our analysts are third party authors that include both professional investors and individual investors who may not be licensed or certified by any institute or regulatory body.