JHVEPhoto/iStock Editorial via Getty Images
There are many moving parts that together made up the Intel (INTC) Q2 Armageddon, although none of those were too surprising (to some extent, as Intel obviously missed by many billion dollars on revenue) and some were already discussed in my recent Intel downgrade. Ultimately, what stood out the most to me was the discrepancy between the current demand environment (embodied in the horrendous quarterly results) and the incredible investments Intel kept making in contrast, with a record-breaking 5.3k net new employees (mostly engineers) that were added during the quarter.
What this means for investors, though, is that they are still buying Intel for the promised turnaround that is (still) years away. Remember that the post-10nm turnaround already got delayed in 2020 because of the consecutive 7nm delay, a node (renamed to Intel 4) that will only start ramping by the end of the year. This means that 2022 is both financially as well as technologically the bottom for Intel.
Since the deal thesis for the stock hasn’t changed, investors who are still on the turnaround train are getting an ever more attractive price point to acquire shares from the shortsighted market that does not things like regaining process leadership. In other words, Intel remains a stock for the patient.
Since Intel had not pre-announced its downside results, investors were up for quite the shock when they received the downside results as well as guidance, especially given to what lengths Intel had gone to in April to reaffirm its full-year guidance. Instead, in one fell swoop around $10B was cut from the guidance, on top of the $4-5B that Intel is losing this year due to the NAND sale. Intel explained that the demand environment had changed very suddenly during the quarter, which is why Intel couldn't have foreseen the results.
This immediately led some to question how Intel would fund its massive expansion plan going forward. The answer is two-fold. First, less demand means less fabs. This in turn means lower spending, and indeed Intel has immediately taken $4B off the 2022 capex guide. Ultimately, Intel will only build as many fabs as it needs to meet supply.
As a reminder, the main issue that led to the shortages was that neither Intel nor TSMC (TSM) had enough spare fab shells – which take years to build – available which they could fill with capacity (although equipment vendors have had some supply constraints as well). However, fab shells are only a tiny portion of the overall cost of a fab filled with equipment. Hence, Intel will still plan to at least build the shells in order to be able to fill those if the demand eventually does catch up.
Secondly, with the passing of the CHIPS Act, Intel has become incrementally more bullish about receiving tax breaks and grants for building fabs in both the U.S. and E.U. At investor meeting in February, Intel said it was modelling for a 10% offset, but was aiming for 20-30%. Intel said it was now expecting 4x more offsets than it did in February, although Intel also said it isn’t expecting any CHIPS funds in 2022 yet. It wasn’t specified if the 4x number also includes offsets from the E.U., but as a reminder the Germany fab only comes two years after the Ohio fabs. One interpretation I read on Twitter is that only $1B of the $4B capex reduction is due to less equipment, with the only $3B being government funding, although my own interpretation is that the 4x number was about a multi-year timeframe.
As an additional, but important and perhaps very overlooked (given the stock price reaction) comment to close the circle, as Intel detailed at its investor meeting in February (and as I have also been writing about for years): the demand (which is causing the fab buildout) that I talked about in the previous paragraph isn’t coming from the PC market, which is stagnant at best and in decline at worst, and which indeed led to the incredibly weak Q2 results. Instead, it comes from the other, more nascent businesses. Intel expected both DCAI and NEX to achieve mid-teens growth by 2025-2026. In addition, Intel is building out three brand-new businesses that each likely will reach an over $10B scale at some point: Mobileye (robotaxi and AVs as main long-term growth drivers most likely), graphics and foundry.
Overall, the point here, to reiterate for the thousandth time (Intel Stock Is Still Worth Buying (NASDAQ: INTC)), is that most of the revenue decline was due to the PC. Hence, the current results say absolutely nothing about the potential of all the other businesses such as 5G, robotaxi, GPUs and foundry, all of which are compelling businesses which have much room for growth and/or for Intel to take market share from an often small position.
Nevertheless, the bears obviously also noticed the data center underperformance, and indeed Intel’s lack of execution of its data center roadmap has been a big reason for my downgrades and a big theme in my criticisms over the last 15+ months: Ice Lake-SP Tells A Cautionary Tale For Intel Investors (NASDAQ: INTC). The only silver lining here is that once Intel regains process leadership (as Technology Development hasn’t shown any causes for concern for the last two years now), there is theoretically no reason why Intel could not launch a Xeon on this leadership process technology within at most 6-12 of the client CPU, or even before the client CPU. Theoretically. Indeed, Sierra Forest was announced to be the Intel 3 lead product in H1 2024.
With that out of the way, Intel delivered what could be described as a kitchen sink quarter. More specifically, the Q2 results were worse than the underlying market dynamics. The reason for this, as Intel explained, was that Intel was selling less CPUs than what the end-market was buying (in terms of PCs). In other words, companies responded to the decline in the macro environment that occurred during the quarter by massively reducing their inventories, which led to Intel selling less CPUs than the actual TAM.
Obviously, though, this can’t continue forever… Unless one would believe that Intel was lying and the actual reason for selling less CPUs than the TAM is because AMD (AMD) is now selling those units. However, that makes absolutely no sense as Intel did explicitly call out the competitive environment in the data center while simultaneously suggesting the PC market share is flat (in-line with the PC market). So instead, Intel said it expected the inventory issue to correct itself over time, to again ship in-line with the actual TAM, which indicates that Intel does not expect significant market share changes in the PC market. This also means that once inventories are depleted, OEMs will have to start buying CPUs again, which should fuel revenue.
In summary, the PC is clearly going through a hangover from the COVID-induced boom due to work and learn from home, which has become even more exacerbated due to the current macro environment. Nevertheless, the 2022 PC TAM is likely still higher than pre-COVID.
In addition, though, the data center was also worse than expected, which further deflated the quarterly results. Intel also does not expect this to change for the next few years. AMD continues to take share, and in the near-term the macro slowdown likely does have some contribution and there are still matched set component issues. Intel also admitted that its uncompetitive product line reduces its ability to increase prices due to inflation.
Behind all this noise, though, this ultimately simply ties back to the whole turnaround theme: Intel only expects to regain process leadership in 2025, so it would indeed be quite surprising if Intel magically wouldn’t have any product issues anymore, as it has had for the last so many years in the data center, in 2022 already.
In that regard, Intel really only has been ramping its investments for the last 18 months. For comparison, designing a CPU or process node takes on the order of 4-6 years or even longer (in any case not shorter). Again, this is nothing new as Pat Gelsinger literally said in one of his first calls that the 2022-2023 roadmap was what it was. Pat Gelsinger again reiterated that the products that are coming out over the next few years (such as Sapphire Rapids, Granite Rapids and Sierra Forest) were all still designed under the old methodology. (Note: some people may remember that Jim Keller, one of the most well-known people in this industry and who played a major role, arguably much more than Lisu Su, in AMD’s turnaround, had a two-year stint at Intel. He already likely pushed some changes as well, but apparently, he didn’t stay long enough for those to materialize.)
In any case, this is a point I did mention in my recent Intel downgrade article: it is simply unfair to (fully) blame Pat Gelsinger for the execution on the products that were already in development for literally years before he rejoined Intel. Sapphire Rapids development started in 2016 under BK. In addition, although I get that investors ultimately judge a company based on its financials, ultimately things like PC and data center demand (which lead to the top line quarterly performance) cannot be directly controlled by Intel.
Hence, I would say that blaming Gelsinger for the Q2 earnings report is too simplistic, especially since the technology side (which was the main issue) obviously still isn’t up to snuff. Put differently, as I had predicted, Wall Street will judge Intel based on its near-term financials, ignoring the nuances of the technology progress. This is what ultimately creates the investment opportunity that investors may take advantage of, as the market obviously is not discounting things like regaining process leadership.
More specifically, one of the comments I saw about the Intel Q2 earnings was that Pat Gelsinger was asking for a >$100M package before delivering the goods. To address this, this is actually false. Although I don’t know if the details had been announced, Pat Gelsinger said last year the package was tied to the stock performance, and he gives (at least) 50% to charity anyway. He said something along the lines of that if he didn’t deliver, he would get nothing.
As a further example, Pat Gelsinger mentioned that its employee engagement survey had delivered results that were best-in-industry, reversing the brain drain that investors had been concerned about. Hence, at this point I would regard comments that take issue with Intel’s culture to be FUD. Intel also had another record hiring quarter in Q2 by adding 5.3k net new employees, which might be its strongest hiring quarter ever, even surpassing the Altera absorption quarter in Q1 2016. Intel added nearly 15k employees in just the last year, and nearly one AMD or NVIDIA (NVDA) worth of employees since Pat Gelsinger became CEO. Although Intel said it would slow hiring (and the CFO even hinted to restructuring charges starting in Q3, which would be weird if Intel would start laying off people after first hiring them), at least in Q2 Intel certainly wasn’t deterred by the near-term with also R&D spending reaching another all-time high, coming in at nearly 29% of revenue due to the revenue underperformance. This is a kind of R&D intensity usually only reserved for growth companies instead of ~$70B behemoths and illustrates how Intel has the potential to again become the most innovative company on Earth.
As already mentioned, Intel sold less CPUs than the actual TAM due to inventory depletion, which exacerbated the PC revenue decline. The 2022 PC TAM will likely still be stronger than before COVID-19.
In addition, and this is the second point, investors should also remember that Intel is losing over $4B in inorganic revenue due to the NAND sale. Adding this segment for the sake of argument, then the actual high end of the guidance call of over $72B in revenue, which is approximately what analysts in late 2020 had expected Intel to earn in revenue in 2021; Intel ultimately ended up beating analyst estimates by about $4B.
I have touched on most important high-level points in the analysis above, so this section is for some other topics that were discussed in the earnings call.
First, besides the macro stuff, Intel (Pat Gelsinger) did call out its own execution issues with Sapphire Rapids. As argued, this is simply a sign that the BK, Bob Swan, and Navin Shenoy (whom Pat Gelsinger more or less fired a year ago) underinvestment ghosts are still lingering.
Secondly, Intel admitted that the Alchemist delays are due to quality issues with the drivers. I would state this is "simply" a software issue, which should eventually be fixed. Obviously, it is extremely disappointing though, especially given how long-anticipated the GPU launch has been and given how often Intel has been bragging about how many software engineers it has (around ~20k). Intel had set an extremely aggressive target for $10B AXG revenue by 2026, so clearly Intel needs to step up.
Thirdly, Intel announced the exit of the Optane business. The issue with Optane is that it wasn't quite as high performance (low latency) as DRAM, which did not make it a straight(forward) replacement, even despite Intel's software efforts to make Optane a new addition to the memory hierarchy.
In the bullish case, 3D XPoint could have become a business as large as any other Intel business by taking away huge chunks of market share from the large DRAM market since Optane had a much more promising scaling (Moore's Law) roadmap which would have only further increased the cost per GB advantage that Optane already had. Put differently, if you need to be 1TB of RAM, it is much cheaper to do so using Optane, at no noticeable performance penalty in many workloads.
Many tech followers will remember how AMD embarrassed Intel for three full years with its 7nm (64-core) CPUs against Intel's 14nm (28-core) CPUs in the data center. However, given the currently available information and expectations, it seems very likely that Intel will be able to give AMD some of its medicine by 2026.
Given the recent TSMC N2 disclosure, a node which TSMC said would go into production in Q4 2025, TSMC's most leading-edge process technology is expected to have a density of less than 300MTr/mm2 (300 million transistors per square millimeter). To be specific, TSMC said N2 would improve the density of a design consisting of 50% logic, 30% SRAM and 20% analog by "greater than 1.1x".
By contrast, Bob Swan and Murthy had both said that Intel was targeting 2x shrinks. Hence, the 14A node that is scheduled for 2026 should be expected to have 2x the density of 20A, which Intel said would have 2x the density of Intel 3, which has a 2x higher density than 10nm (which has a density of 100MTr/mm2).
Altogether, this means the benchmark for 14A is 800MTr/mm2. This is nearly 3x higher than TSMC's N2. Given how incredibly large the margin of error hence is, if the node launches on schedule, it is virtually guaranteed to strengthen Intel's process leadership, in turn undoing most of the issues Intel has had for the last decade.
Regaining tech leadership alone doesn’t make for a compelling investment case if Intel did not have strong opportunities to monetize this IP. As Q2 showed, despite having tried for many years to diversify its business, Intel’s revenue remains heavily dependent on the stagnant PC market. Intel has hinted that Alder Lake and Raptor Lake will serve to maintain flat market share in the near-term at least, and Intel also said it was shipping less CPUs than PCs being sold (inventory reduction). So while further downside in revenue seems unlikely, neither foundry, graphics nor Mobileye robotaxis are big enough yet to move the needle forward in the near-term. Meanwhile, the only business that does have the scale and growth profile, the data center, is still in the midst of its execution failures induced by prior management.
Pat Gelsinger described the turnaround as a three-step model. First comes regaining process leadership (expected in 2025), then comes launching the products on these nodes (improvement expected in 2024 on Intel 3, but likely not fully until 2025 either), and last comes entering growth markets (with both Mobileye, foundry and graphics currently being around $1B businesses, neither of those will reach a substantial scale before 2025 either).
Nevertheless, given that timeline, the current variability in the PC market will most likely be long gone by then. So while the current decline of the PC TAM in the wake of the post-COVID-19 hangover and the more recent Ukraine and recession risks is a setback, it should ultimately not impact the long-term upwards trajectory if the three aforementioned businesses deliver on their respective theses. (In that regard, Mobileye delivered a strong >40% growth quarter and IFS continued to gain momentum by adding $1B to the pipeline in one quarter, while graphics remains wait-and-see.)
Is it really that bad? sure looks like it right now but I have a 5-year+ horizon so why would I want to dump shares at a 52 week low now if I have no urgency to do so?
These comments were explicitly debunked by Pat Gelsinger during the earnings call:
Intel is trying to chance too late but they are not changing their primary problem: culture. Intel Non-GAAP EPS of $0.29 misses by $0.41, revenue of $15.3B misses by $2.63B
I can’t help but think the culture needs a reboot. I left after 27 years back in 2017. The firm had definitely slipped.
This comment was addressed above (he only gets the package if he delivers):
This guy wanted 178M pay package without earning it.
If this article's as well as Intel's thesis that the results will improve in Q4 does not materialize, then following comment (which was actually posted before the Q2 earnings) may be correct:
Before the year end Intel will have to make huge tough choices…cut its dividend, slow the pace of investment on fabs, let many people go, reduce marketing spend…pick your preferred medicine.
This comment was also posted before Q2 earnings, so at least one investor got what he asked for:
Shhh... I want a kitchen sink type moment for INTC so I can try to sneak into it at epic lows.
When Pat Gelsinger became CEO, literally everyone was saying it was going to take years to turn the ship. What’s the solution, bring back Bob Swan to do stock buybacks again while ignoring Intel’s technological position? Investors need to remember that Pat Gelsinger is not responsible for the mess Intel got itself into over the last decade. He’s doing his part to reinvest in Intel by literally hiring tens of thousands of new engineers, install new executive management, build new fabs and enter new growth markets. He’s doing what Intel should have done a decade or even two decades ago. Hence, he should not quit.
Yes, Jim Keller obviously knew all about COVID-19, inflation, Ukraine war and recession.
Well it appears we now know the real reason why Jim Keller left Intel.
The irony is that people are blaming Intel for being behind in technology, hence losing market share to AMD. While simultaneously asking for Bob Swan to return, whose solution would be to buy back more stock and reorg the company by laying for 30k employees, which will destroy competitiveness in the long-term (if it wasn’t destroyed already after a decade of mismanagement). Pick your best poison, but for the long-term the only sensible thing is to invest in best-in-class products, which costs money (which will hurt the stock near-term).
fire Pat Gelsinger..call the previous beancounter..
Following comment I came across made a nice comparison:
Huge numbers for AMD are still one third of a terrible quarter for intel.
Based on around $6B revenue for AMD in Q1, Intel still did a multiple of AMD's revenue even in the "disaster quarter" that was Q2. Note that AMD took over Intel in market cap due to the sell-off, so the discrepancy in valuation remains quite staggering.
Author comment: not a single mention of Sapphire Rapids. Nevertheless, if Intel can ramp Blockscale in one year, then given the Granite Rapids tape-in in Q2 2022, there should be enough time for the supposed H1 2024 launch.
Directionally my analysis of Intel in my opinion has been correct for the last year or more. For example, in April last year I detailed as a red flag how the Ice Lake Xeon delay could not be attributed to the 10nm delays, which given the Sapphire Rapids saga indeed turned out to be the canary in the data center coal mine. More recently, I actually downgraded Intel before its Q2 report. In stark contrast, many analysts only downgraded Intel afterwards. They didn’t have the guts to do so beforehand, yet being forward-looking is what being an analyst is actually about as anyone can “predict” the past.
So when it comes to the future, technologically I would remind people that Intel for the fourth consecutive quarter reiterated that Intel 4, 3, 20A and 18A are all on or ahead of schedule. In addition, Intel has hired nearly a full AMD or Nvidia worth of employees in the last 18 months.
So as a thought experiment, imagine that Intel assembled a team from scratch in 2021 to work on a future Xeon CPU. Given that such a task takes on the order of 5 years, this CPU will not launch until 2026 at the earliest on a process node Intel hasn’t even announced yet. After launch, it will still take another year to ramp into volume (to surpass the previous CPUs in unit shipments), and only then will this CPU be able to retake market share from AMD. This means that the teams Intel hired in 2021 might not have a financial impact until 2027. Although of course some (or many) of the hires could also be to strengthen existing teams and projects, this example should give a more realistic picture than what the comment sections of Intel articles would have one believe.
In summary, on one hand PC revenue declined more than the actual TAM, however this was not due to market share losses but due to a one-time inventory correction. On the other hand, data center continued to lose market share as Intel only expects to regain process leadership in 2025. So overall, even though the results were quite a surprise (Intel Needs To Surprise (NASDAQ: INTC)), poor quarters like Q2 could have been expected, even though admittedly not to the extent of the actual underperformance Intel reported.
Still, without any new process node delays for the last two years now, and with a very competitive and even cutting-edge roadmap in the wake of TSMC’s embarrassing N2 disclosure (which seems to bring a ~1.2x shrink), I remain optimistic about Intel’s future. As long as Intel regains process leadership, there won’t be any fundamental roadblocks that cannot be otherwise solved, such as design failures (Sapphire Rapids CPU) or software issues (Alchemist GPU). Watch from the sidelines or buy the dip and meanwhile harvest the dividend (which, yes, Intel said it remains committed to).
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.