In line with my Intel (NASDAQ:INTC) downgrade earlier in January, I don't see much upside for Intel in the near-term.
In particular, investors likely shouldn't expect a large beat in Q4: given the huge outperformance in Q4'20, Intel will face tough comps. Further pressure on the top (and bottom) line may come from continued China data center weakness in the wake of the mediocre Q3. Lastly, there have been more rumors of even further product delays in the data center (which was the main reason for my downgrade in the first). Although all these rumors come from the same, single source, Intel hasn't done any effort to disprove those.
As such, the promise of Intel's turnaround leading to large shareholder returns remains far from becoming a reality. To be sure, there are many signs that the ship is being steered in the right direction, such as the progress on regaining process leadership and the many new fabs being built, but investors will likely need to be patient for several more years, as Intel may only start harvesting returns on investment in the second half of the decade.
Intel will soon publish its fourth quarter results. I expect two main topics to be on analysts’ and investors’ agenda: signs of improvements of the shortages, and what (if any) truth there is to the Sapphire Rapids delay rumors.
With regards to the shortages, Intel hasn’t provided a ton of clarity this year. Intel said it expected the worst impact in Q3, but in fact Q3 results (as I covered them at the time) were quite mediocre. Although Intel blamed this on Chinese regulatory changes (which impacted gamers and hence the cloud), if there is supposedly much higher demand than supply (which is the definition of a shortage), then wouldn’t Intel have been able to redirect those CPUs to other (impatient) customers?
In addition, another question where the answer isn’t clear is whether Intel is CPU-constrained (because it doesn’t have enough fab capacity) or whether it is constraint due to the “matched set” issue (because there is a shortage of other components that are required to build a complete server or PC), which Intel initially blamed to be the cause for its constraints.
My own guess would have been that given the Apple modem and Mac transitions, it seems highly unlikely that Intel’s fabs are fully constraint, as Intel has been on a ~$70-80B run rate for a while now. In addition, during the previous shortage in 2018-2019, Intel already started some projects (as reported by rumors at the time, although never quite confirmed by Intel) to increase outsourcing things like PCHs (chipsets).
On the other hand, in the last two quarters or so, Intel mentioned that it wasn’t able to fulfill all FPGA orders (demand “significantly” outpaced supply), which is noteworthy since FPGA is only a ~$0.5B per quarter business and one would think Intel would perhaps prioritize producing those 70%+ gross margin FPGAs over much lower margin low-end CPUs. So apparently, despite losing all those Apple sales, Intel’s fabs are still being outpaced due to the high Windows PC demand.
Intel had also said earlier in the year that it started to produce substrates in-house, but Intel hadn’t provided any further information about this during its last call. I came across a report that predicted a shortage in substrates until 2026, but nevertheless other reports actually show that Intel is pretty much the only company that didn’t show double digit revenue growth in 2022. The report I am referring to in fact claimed Intel revenue in 2022 declined, although that goes against Intel’s own Q4 guidance, but in any case Intel should be roughly flat year-on-year. (I don’t have sources for both reports because they appeared and then disappeared from my Twitter feed.)
So moving to the financials, as mentioned for the full year Intel will be about flat, perhaps eking out a slight new record year. Considering the Apple transitions and AMD’s (AMD) momentum over the last year, if anything it should be considered quite remarkable that Intel’s revenue remains about as high as it has ever been. (Some bears on Seeking Alpha likely had predicted Intel revenue would have collapsed by now.) Obviously the record PC demand (“1 million units per day”) has helped considerably in that regard – without the pandemic Intel might have had to report much worse numbers over the last year, most likely.
With regards to Q4 specifically, I would expect records for Mobileye and IoT, and perhaps some further recovery for FPGAs given the aforementioned trends I would also expect a (relatively) strong data center environment – any lingering China issues (which Intel did expect to continue for the next few quarters) notwithstanding. Remember how Intel had guided a year ago that data center revenue would improve through the year (although this flattened in Q3 due to the China slump).
Nevertheless, despite the accelerating Ice Lake-SP ramp, Intel expected “just” 1 million sales in Q4, which compares to a total volume that should be at least 6-8 million units. Therefore, Ice Lake-SP would still only make up on the order of just ~20% of DGC, even though Sapphire Rapids is already supposed to launch in Q1.
As mentioned, whether Sapphire Rapids will make this timeline is doubtful. Most realistically, assuming at least 12 months between these two CPUs, then the launch may come during Q2. Given how long it is taking Ice Lake-SP to represent a significant fraction of total DGC units, Intel likely won’t be too concerned about “launching” Sapphire Rapids, since initial volume will remain quite limited for a few quarters. (Of course, “limited’ is relative since compared to AMD’s size – although AMD has grown quite a bit over the last few years – even a slow ramp would still quite quickly dwarf AMD's data center business.)
Overall, though, while I won’t bother with providing any numerical estimates for Q4, directionally I would remind investors that Intel had a blockbuster Q4'20 due to the ramping PC demand in Q4 last year (especially at the low-end). Therefore, as the overall roughly flat revenue for the year indicates, investors shouldn’t expect too many or significant surprises on the financial side.
Of course, as in any quarter, many investors and analysts may be quick to dismiss any beats or misses as they will look forward to guidance for the next quarter and year. For Intel, this is especially relevant given the new CFO appointment and the looming Investor Meeting, although management may therefor opt to refer to this event during the call when certain questions are asked.
Undoubtedly, there will be the usual scrutiny regarding gross margin and capex. Although gross margin in 2022 will be low compared to Intel’s standard, last I saw TSMC has guided for just 53% either, despite all the rumored price hikes.
Regarding capex, both Intel and TSMC have been one-upping each other over the last year, although TSMC’s spending has clearly outpaced Intel’s – even when accounting for the fact that only 70-80% of TSMC’s capex is actually for leading edge capacity. In fact, despite Pat Gelsinger’s talk of “torrid pace” it seems the manufacturing group didn’t quite get the message since both the E.U. and U.S. (mega)fabs (at least their announcements) have been delayed, since Intel had expected to announce its next locations by the end of 2021.
Pat Gelsinger had said that IDM 2.0 was Intel’s strategy and wasn’t influenced by any politics, but clearly Intel is waiting for greenlight for juicy subsidies. In December, Pat Gelsinger said (transcript available on Seeking Alpha) that total subsidies could fund up to 40% or so of a fab, with prepayments further reducing the capital burden. With regards to the latter, though, at this point it seems Qualcomm may be the only foundry customer significant enough in size to be able to pay any meaningful prepayments. Still, such a prepayment would be a huge sign of commitment at this stage for Intel Foundry Services (if it materializes).
With regards to revenue, excluding the sold NAND business, non-GAAP revenue should again be roughly flat for the year. Maintaining flat revenue in 2022 should be a bit easier given that most of the Apple headwinds happened in 2021. While PC will be facing quite tough comps, there could be some upside from an improved overall supply (chain). While also the data center will continue to face tough competition, there are no signs that demand will subside, and the comps in the first half should be relatively easy to beat. However, as discussed above the 10nm ramp takes a few quarters to become material in size, so it remains to be seen if Intel can ramp Ice Lake-SP and Sapphire Rapids faster than that customers get tired of having to keep buying old, outdated 14nm parts – and therefore move to competitor CPUs.
Lastly, after about three years, Intel will be looking to start its transition beyond 10nm as it starts to ramp Intel 4 (Meteor Lake) in the second half of the year. Overall, despite the delay, this is still a big improvement compared to the (much longer) 10nm transition, and obviously Pat Gelsinger intends to follow this up even more quickly with the following nodes.
While the data center ramp remains mediocre, Intel for the first time in many years has a solid CPU on the PC side, spanning mobile to desktop. Intel has been touting many of its improvements, and also said it is its fastest rollout ever across the product stack (mobile, IoT, vPro and desktop all launching within months). As Pat Gelsinger said:
"Alder Lake. All of a sudden...Boom! We are back in the game," exclaims the impish tech CEO. "AMD in the rearview mirror in clients [consumer market]," he adds, "and never again will they be in the windshield; we are just leading the market."
Looking forward, while AMD will be moving to 5nm by the end of the year (which does give AMD a chance to pull ahead again, contrary to what Gelsinger said), AMD's advantage at the 5nm node won't be as large as it was at 7nm in 2019: (1) on the laptop side Intel will respond with Meteor Lake on Intel 4 within one or two quarters, and (2) on the desktop side Intel is expected to leapfrog AMD with TSMC 3nm-produced CPUs in 2023.
The main reason for my recent downgrade of Intel was its inability to fully capitalize on the surging data center demand due to its non-leadership product portfolio.
It seems this is set to continue for the next few years, at least according to rumors. SemiAccurate claims that Granite Rapids would launch in the first half of 2024. This is a glaring issue given that Intel had already delayed Granite Rapids by a full 12 months to the first half of 2023 (in the wake of the 7nm delay). Hence, this CPU may now be delayed by two years.
CEO transition or note, clearly such a delay would represent "business as usual" for Intel, which is the opposite of what Intel investors are expecting since the CEO transition.
The reason this rumor may, unfortunately, be rather credible is because Intel still has the Emerald Rapids refresh planned after Sapphire Rapids, the latter of which itself is already delayed by up to a year when it launches somewhere in the next few months.
With regards to Emerald Rapids, investors should remember that Intel has an utterly nonexistent track record when it comes to launching multiple data center CPUs within a short period of time. The most infamous example is how Intel planned to launch both Ice Lake and Cooper Lake in mid-2020, but ended up delaying Ice Lake by nearly a full year to Q2'21. Due to this, investors started to question Intel about how Intel (instead) would deal with launching Ice Lake and Sapphire Rapids both in 2021, but obviously Intel's "solution" was to delay Sapphire Rapids to 2022.
SemiAccurate did claim that Intel's flagship ("unquestioned leadership") Diamond Rapids would still launch in 2024, but obviously this represents the same issue as above of Intel having to launch two CPUs in one year. Nevertheless, what has changed over the last year is that Pat Gelsinger has been ramping up R&D spending, so perhaps that may finally start to yield results in the 2024-2025 timeframe (as Gelsinger originally promised).
As most investors may have noticed, Intel announced its new U.S. fab location (its first in 40 years) for a greenfield fab with much fanfare. As per Intel's usual practice in the last few years, each fab has a price tag of $10B, and Intel will initially build two of them, capable of 18A and below. However, given the potential for Intel Foundry Services to generate a lot of demand (as already evidenced in the last year by the shortages), Intel clearly wants to build out the location to 8 fabs or even more ($100B), to become what Gelsinger argues would be (one of) the largest semiconductor mega-fab(s) in the world.
The neat thing, however, is that given the current political climate, Intel may not have to pay most of the cost themselves. Ohio announced that for every 6 cents it would spend, Intel would invest one dollar. Based on the initial two fabs, that implies a ~$1.2B subsidy. In addition, the federal CHIPS act could add another $3B in funding per fab, which means subsidies might pay up to ~40% of the cost out o (~$7B for the two fabs).
That seems like a really solid deal Intel may be getting here, so again it is not sure what all the fuss from investors about Intel's $26B+ capex plan is about. These fabs, whether they will be used for Intel or IFS demand, will undoubtedly print money and pay themselves back in a very short time.
In total, Intel has committed to on the order of $80B in greenfield new fab space (excluding on the order of ~$10B in additional packaging facilities) over the next five years:
Doing the math, building out all these fabs over the next five years would contribute $20B to annual capex, which is obviously excluding capex required to run the existing fabs. Hence, investors should likely expect ~$30B capex as the new baseline, but may perhaps get close to TSMC's ~$40B around 2024 when all those projects are in progress.
Remember that these are all leading edge fabs, so Intel could actually be spending more than TSMC (TSM), which only spends 70-80% of its capex on leading edge capacity.
Key semiconductor manufacturing competitor TSMC recently amazed analysts and investors alike by further increasing its capex budget to (up to) $44B in 2022. This represents a ~3x increase in ~3 years.
However, while some contributors therefor concluded that the war is over and TSMC has won, I would instead argue that it seems a little premature to declare any victories yet, considering that Pat Gelsinger has only been steering the massive Intel battleship for just one year. The only bullet Intel has really fired so far is by starting the Intel Foundry Services business, becoming a direct competitor to TSMC and therefore threatening the latter’s monopoly at the leading edge.
However, Intel’s battle plan in this war is a little different than TSMC. While TSMC remains firm on defending its Taiwan home soil and planning to build as many fabs as it can to safeguard its position, Intel is strategically expanding its territory with the help from allied U.S. and E.U. subsidies.
But I would argue that the main battle is not with regards to who can build the most fabs, but who will actually have the best fabs. For example, consider that WWII was won due to superior technology from one side, not because this side had more troops on the ground. I would argue the same principle will apply here. Hence, Intel’s battle plan consists of regaining process technology leadership in 2025 with the three-headed dragon that consists of high-NA EUV lithography from ASML (ASML), RibbonFET and PowerVia – Intel’s equivalent of the atomic bomb.
Therefore, I would advise investors to take profits from their TSMC investments in a timely manner, in order to prevent their portfolio from becoming radioactive from the fallout of underutilized TSMC fabs once the dust really settles.
However, as explained above, it is still early in the battle given Intel's mediocre outlook in the near-term (while TSMC is firmly growing in the double digits and could overtake Intel in revenue in the next few years). In fact, the leaders are still in the negotiating and scouting phase, quite literally: Intel CEO Pat Gelsinger arrives in Taiwan - Focus Taiwan.
Although TSMC seems to be quite significantly outspending Intel, I would remind investors that Intel had already been investing quite heavily to improve its capacity since 2017-2018 (which doubled from 2017-2020), so TSMC simply had to catch up a lot more than Intel. For example, Intel has two expansions coming online in 2022-2023, in Oregon and Ireland, which both started well before COVID-19. In addition, as discussed, given the many fab plans Intel has, Intel's capex may actually overtake TSMC's leading edge spending over the next few years.
With regards to Intel’s financials, net-net I expect continued relatively stable revenue for Intel – which of course compares to the hypergrowth that pretty much all other semiconductor peers and competitors have seen, which explains the lower P/E multiple Intel gets in the market.
Further, investors will likely want more details about how the shortage is progressing as well as the usual capex and gross margin items. Gregory Bryant, Intel’s resigning EVP of the PC group, had already confirmed around CES that Intel’s process technology development still continues to be on track. Any further details and updates may be delayed until the Investor Meeting in February.
(With regards to the GB executive shuffle, while unfortunate, I would remind investors that overall Intel still seems to have a clear net inflow of talent, both at the executive and engineering level, that has significantly dwarfed any outflow. In addition, I would also remind investors that the most important engineering happens not in CCG but in DEG/PEG, Design/Platform Engineering Group, and therefore PCG overall may be seen more as a customer-facing organization, which explains why GB has been replaced by the GM of sales, who at least in that role had a strong track record.)
Overall, at the start of the year I “downgraded” Intel based on the absence of significant upside catalysts in the near- to mid-term. For example, two of the three main items Intel discussed at CES, discrete GPUs and robotaxis, remain $0 revenue businesses at present. Combined with Intel’s (re-)entry into HPC and AI with Ponte Vecchio, this does mean Intel will launch several important projects in 2022, but it may take a few more years until those start to move the ~$75B needle.
This article was written by
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.