There aren't many deep value turnaround opportunities in the technology world these days. However, Intel (INTC +13.79%) is an interesting case.
After falling behind Taiwan Semiconductor Manufacturing (TSM 0.84%) in process technology, Intel embarked on an ambitious and expensive turnaround plan in 2021 to retake the technology lead and produce chips for other companies. The transition is supposed to bear fruit this year.
But things look ugly right now, as lower revenues and profits from less-good products combined with heavy investment in the turnaround has caused profitability to sink. Intel's stock fell 60% in 2024 and now trades below book value.
While Intel actually beat expectations on last week's fourth quarter earnings call, it guided softly because of a potential pull-forward of sales before potential tariffs. Yet while the fourth-quarter profits only slightly "beat," they were actually much better than advertised.
NASDAQ: INTC
Key Data Points
Intel's fourth quarter looked strong, but...
In the fourth quarter, Intel generated revenue of $14.26 billion, well ahead of the $13.83 billion estimated by analysts. And while adjusted (non-GAAP) earnings per share of $0.13 came in ahead of estimates as well, that bottom line only beat by $0.01. Given the magnitude of the revenue beat, one might have thought more would flow to the bottom line.
Management also guided for a sequential decline in the first quarter, to $12.2 billion in revenue at the midpoint and adjusted earnings per share of breakeven. The first quarter usually sees a seasonal decline, but this one looked a bit worse because of the pull-forward of purchases ahead of tariffs.
Underwhelming Q4 margins seem to have been concentrated in Intel's data center (DCAI) segment. Even though that segment grew sequentially, margins actually fell into the single digits:
Image source: Intel Q4 2024 presentation.
A 6.9% operating margin would be strange and worrisome, not only because the segment's revenue grew, but also because Q4 saw the launch of Intel's new Xeon 6 server products, Granite Rapids and Sierra Forest, which first hit the market late in Q3. Given that these chips were coming out of Intel 3, Intel's most advanced fab in mass production today, one would have expected much better margins.
One-time charges affected the results
Fortunately, upon reading Intel's annual report, the situation is less dire -- at least regarding current server products.
In the report, Intel mentioned the company took a $922 million inventory reserve charge in 2024 for its Gaudi 3 accelerator line. These were the AI accelerators Intel hoped would challenge Nvidia (NVDA +1.73%) in data center AI applications. Earlier this year, former CEO Pat Gelsinger had let on that Intel would sell up to $500 million of these accelerators in the back half of this year. However, now it looks as though Intel has written off those products to zero.
On top of that, interim co-CEO Michelle Holthaus also said on the Q4 call that the upcoming Falcon Shores AI chip, planned for this year, would only be an "internal test chip" and not brought to market. Intel will essentially skip Falcon and move to Jaguar Shores, the next AI chip on the roadmap, faster.
The downside is that Intel's accelerators don't seem to be selling well and may have some architectural flaws. Hopefully, the Jaguar Shores product, likely to be released in 2026, will have fixed those flaws and become a competitive challenger.
But the bright side is that current CPUs are doing much better from a margin perspective. Of that $922 million charge, it appears $313 million was taken in the third quarter. That leaves a $609 million charge taken in the fourth quarter.
If one adds that $609 million back to the data center segment, then fourth quarter data center operating income would have increased from $233 million to $842 million. That would have increased Intel's data center CPU operating margins from the reported 6.9% to 24.9%.
That's actually quite a competitive margin. Rival Advanced Micro Devices (AMD +11.44%) data center operating margins were 29.3% in the third quarter 2024; however, AMD also groups its new MI300 AI GPUs, which probably have a higher margin, into that segment, and AMD has also produced CPUs on TSMC's leading node, whereas Intel is still in transition to a leading-edge process.
Another one-time charge hit results
In addition to the Gaudi write-off, Intel also took a $755 million accrual charge related to its Ireland Intel 3 fab. In mid-2024, private equity firm Apollo Global (APO +4.20%) invested $11 billion for a 49% stake in Intel's Ireland fab 34. However, Intel has scaled back the buildout, probably because of last summer's decision to move its Arrow Lake desktop CPU product to TSMC. Arrow Lake was supposed to be built on Intel's 20A process in Ireland, but Intel decided to accelerate development of its more advanced 18A node to move faster and save money. Given that the Ireland fab won't build as much product, Intel is giving some of that $11 billion back, it appears.
Counteracting that charge in Q4, Intel did receive a $560 million benefit as well, because of a successful challenge of a 2009 European Commission legal ruling.
Combining the $609 million Gaudi write-off and $755 million Apollo charge, then subtracting the $560 million EC payment, and the net effect of these one-time charges would be $804 million, or roughly $700 million after-tax. That equates to over $0.16 per share.
So absent these non-recurring charges, Intel's Q4 adjusted EPS would have been $0.29, more than double the $0.13 reported.
Intel's big year
After former CEO Pat Gelsinger retired in December, things remain uncertain for Intel, as it's still without a permanent CEO, and its data center accelerator business appears pushed out even further.
However, other elements around current products seemed to be better, with strong PC sales and sequential improvement in data center, absent those charges.
2025 will see the all-important introduction of Intel's 18A node, which is supposed to surpass TSMC in process technology. And even though Intel's interim co-CEOs seem to be forecasting conservatively in other respects, both had good things to say about 18A on the conference call with analysts.
Co-CEO David Zinsner said, "We're excited by the launch of Panther Lake this year and the internal ramp of Intel 18A in the second half that will support increased volumes and improved profitability in 2026. From the perspective of external customers, Intel 18A is a very competitive offering that gives each of them a reason to engage with us." Holthaus later added: "What I can tell you is 18A is going well. They earn my business."
So, the bad news is that Intel's AI accelerator roadmap appears to be pushed out. But the good news is that profitability was a bit better than it appears, and the all-important 18A node is on track. If the positives continue outweighing the negatives, Intel will be a turnaround story worth watching in 2025.
