hapabapa
Investment Thesis
I have been quite bullish on Intel (NASDAQ:INTC) since the stock reached and even went below the $30 level nearly a year ago – even without having to require any overly ambitious financials besides returning back to pre-downturn numbers: Intel Stock Bottomed And There Is Upside Potential.
In the wake of the recent flawless quarterly Q3 report, the stock has rallied back towards the high $30s, despite the financials still being even below the levels from a year ago. Nevertheless, as the financial trend has finally reversed (i.e. sequential growth) which implies the upturn is getting more and more de-risked, the stock is following suit.
However, despite the outperformance on virtually all metrics, the financials on the face of it (i.e. without looking at the underlying trend) are still quite ugly. This means that as the near-term rally optimism eventually subsides, the stock might at best trade sideways for several quarters. Looking more at the long-term, given the underlying turnaround, which has also getting incrementally more de-risked with each quarter, buying a stock that has ample potential to return to (and even exceed) previous $50+ levels for the current price could nevertheless be a sound investment decision.
Q2 analysis: Flawless
Make no mistake, after the ugly Q2’22 report that marked the start of the downturn last year, a further double digit revenue erosion on the face of it is not something that should be welcomed by investors. Even despite the clear $0.9B revenue beat (compared to guidance), which might signal a somewhat faster recovery than previously expected, the Q3 guidance only sees revenue growing a few hundred million dollars sequentially, which still leaves the company billions of dollars off the near-$20B revenue quarters at the end of the upturn.
All the financial details can be found here, but the main notable item worth calling out is the underperformance of NEX. Investors should note that NEX (formerly the networking part of DCG + IoTG) was touted as one of those above average (mid-teens) growth components in Intel’s investor meeting 2021 plan to accelerate its growth rate. Instead, the business is currently faced with a strong inventory correction as well as reduced 5G infrastructure spending. Note that IoTG alone was already an over $1B per quarter business prior to COVID, so revenue of just $1.4B (down 38%) has significantly brought down the overall financial performance.
So however one slices it, the downturn is taking far longer than expected. Some other examples are how management had initially expected Q2-3 of last year to mark the bottom, or the optimism about the second half of 2023 that was still present at the start of the year.
As mentioned, though, given the context of the current downturn Q2 actually marked a flawless quarter. Besides the sizeable revenue beat, Intel also strongly beat on gross margin (due to the revenue outperformance) as well as EPS. The EPS beat in particular is noteworthy as it marked a quick (and quicker than expected) return to profitability. This shows that Intel’s cost reduction efforts are paying off. Nevertheless, given the still sizeable capex investments during this multi-year turnaround, free cash flow was still quite negative.
In addition, the flawless quarter wasn’t just constrained to beating the downsized revenue expectations and returning to profitability. In the underlying business, the quarter also seemed to go exactly as planned, with no delays in both process technology or product development. This means the plan announced exactly two years ago at Intel Accelerated, in order to return to industry leadership by 2025, remains on track.
Given how Intel’s lack of execution track record had been a major hurdle for many investors, the lack of delays is certainly noteworthy.
Going further back, since the last delay was announced in mid-2020, there have been no further delays in the process technology roadmap for the last three years. Even going back four years, in 2019 the plan was to overtake TSMC (TSM) in (early) 2024 with 5nm (Intel 20A), and that plan still more or less stands (with at most a 6-to-12-month delay), with Intel announcing the first Arrow Lake stepping now being in the fab.
Red flags remain
So far so good, but as indicated there remain two red flags that investors should want to see addressed. The first, as already discussed, is the slow financial recovery, with revenue performance still being at a multi-year low. For the most optimistic investors, though, that may mean Intel has a low bar to clear going forward to deliver revenue growth (although that should be weighed against the stock price).
The second risk is the data center roadmap, which does not seem to be capitalizing as quickly on the improved process technology outlook. While Intel did reveal Clearwater Forest on 18A for 2025 in the E-core swim lane, it is unknown how meaningful the E-core swim lane adoption will be. As such, that likely still leaves the “legacy” P-core line-up as the workhorse for Intel’s DCAI financials.
However, despite the initial rumors of Diamond Rapids for 2024 a few years ago, the recent roadmap showed that the Granite Rapids successor isn’t even planned for 2025 currently. This means that after the quick succession of Sapphire to Emerald to Granite, there will again be quite a void until Diamond Rapids to make use of Intel’s leadership process technology. This is obviously a missed opportunity.
Another way to see this is that basically the new management has undone the prior management’s efforts to make Xeon a first-class citizen (adopter) of Intel’s most advanced process technology, as Granite Rapids on Intel 4 (at the time called 7nm) was initially planned for early 2022. This means that if Intel had executed flawlessly on the Granite Rapids roadmap, then there would have been a massive four year void between the transition from 7nm (Intel 4) to 5nm (20-18A) in the data center, which is half as fast as what Moore’s Law suggests.
For a company that has been publicly calling out the competitive pressure in the data center for quite a while now, the way Intel is handling its data center roadmap is sloppy at best. In contrast, in the PC space Intel was touting how it regained market share for another consecutive quarter, clearly (and unsurprisingly) evidence that a competitive roadmap matters.
In addition, although not completely related, a more near-term and discussed topic on the earnings call was the impact of the shift to GPU spending. While GPUs are “merely” accelerators and still need CPUs as host, the GPU to CPU ratio can be as high as 4, and word is going around that Nvidia’s (NVDA) latest H100 is selling for as much as (or even over) $40k, with the latest reports even indicating $70k in China. That is nearly two orders of magnitude below the ASP of Intel’s Xeons, which I estimate to be below $1k (source: during one quarter where Intel reported that it shipped over 8Mu, its DCG revenue was well below $8B).
Intel did indeed acknowledge that big CSPs and enterprises are spending heavily to build out their AI training infrastructure. However, as shouldn’t be a surprise to most investors, Intel has long had an “AI everywhere” strategy, building in AI capabilities in existing products as well as building new dedicated AI accelerators.
Given some hiccups in line with Intel’s general sloppy execution in years past, though, Intel hasn’t been able to capitalize on AI yet besides in the inference space where AI still often tends to run on CPUs, for which Intel has long had a strong leadership position, against AMD (AMD), with its DLBoost capabilities (which Intel further extended with Sapphire Rapids’ new AMX engine).
Nevertheless, with current product lines such as Gaudi2 and GPU Max series (Ponte Vecchio), Intel is in principle by now well able to capitalize on this opportunity. Indeed, Intel said that its AI pipeline (GPU Max/Flex + Gaudi) grew a massive 6x QoQ to over $1B. While this is still tiny compared to Nvidia, since a pipeline by definition pertains to multiple quarters of revenue, clearly Nvidia having difficulties with keeping up with demand is creating opportunities for competitors to gain momentum.
My thesis or expectation in this regard would be that if customers are forced to try out alternatives to Nvidia, they will find that Gaudi and/or GPU Max are very potent as well as very cheap in comparison, which should significantly lower the barrier for further adoption.
One last remark would be that there could be some cautiousness given Intel’s plan to sunset the Gaudi product line in favor of Falcon Shores (with Pat Gelsinger saying Falcon Shores 2 is already planned for 2026), although Pat Gelsinger denied this concern when asked during the call, referring to Intel’s oneAPI software stack.
Further discussion
I have made a concise overview of the Q2 earnings call, which is available here.
To highlight a few items, one quite unexpected announcement was that after announcing the Meteor Lake PRQ for Q3 (whereas I would have actually expected Q2 given historical examples), Dave Zinsner later in the call nevertheless hinted at a Q3 (read: September) launch for the product. With regards to Intel 3:
I would highlight Intel 3 met defect density and performance milestones in Q2, released PDK 1.1 and is on track for overall yield and performance targets. We will launch Sierra Forest in first half ’24 with Granite Rapids following shortly thereafter, our lead vehicles for Intel 3.
Investors may also have seen a few 18A-related announcements, such as two new RAMP-C members and the Ericsson (ERIC) partnership for a 5G chip on 18A. A new announcement on the call pertained to what seemed encouraging progress regarding two potential big (“whale”) customers for the node.
Intel further noted progress on its capital offset program, with the first U.S. Chips Act application having been sent, and three more applications in progress. Intel also said the offsets are now trending towards the high end of the 20-30% range, which was already up from the 10-30% range from the 2021 investor meeting (with Intel at the time conservatively using assumptions at the low end of the range for its models).
With regards to the cost reduction efforts, besides being on track to the $3B and $8B+ targets for 2023 and 2025, Intel also said that its nine exited businesses are saving $1.7B. However, it seems this was only referring to the cost of running those business lines; it should be expected that those businesses were also bringing in revenue and some perhaps even profits. It would have been less disingenuous if Intel has instead reported the net P&L effect of these disinvestments.
With the decision to stop direct investment in our client NUC business earlier this month, we have now exited 9 lines of business since Pat rejoined the Company with a combined annual savings of more than $1.7 billion.
Guidance
As mentioned, the guidance (which calls for a relatively minor sequential recovery) is lower than I would have expected, and likely lower than many others as well would have expected given the initial optimism for H2 at the beginning of the year.
While Intel indicated that the PC inventory digestion is finally complete with CPU shipments now approaching PC sell-through rate, the same cannot be said for the data center, which in addition seems to have a weak macro outlook in general (except obviously for the GPU part of the data center TAM). Intel expects a sequential decrease in DCAI revenue in Q3 followed by a recovery in Q4, so perhaps combined with seasonality Q4 could be a relatively strong finish to the year (at least compared to Q1).
Nevertheless, there should be some gross margin tailwind in Q3 from marginally higher revenue, lower PRQ costs and slightly lower underloadings. Intel also expects sequential Mobileye (MBLY) and IFS growth. The IFS growth in particular is noteworthy given that this business, which in its early stages seems quite lumpy, already doubled QoQ in Q2 (and quadrupled YoY). Intel called out the IMS Nanofabrication business as well as packaging (real silicon revenue is likely still some quarters out).
Investor Takeaway
Intel had a flawless quarter with beats on revenue, gross margin, EPS and guidance. Intel also seems to have had a flawless quarter in terms of roadmap execution, further reinforcing and de-risking the underlying turnaround effort. However, investors obviously shouldn’t be oblivious to the fact that this quarter’s outperformance comes after basically four straight quarters of underperformance (whether due to misses or lowering guidance in advance) during this downturn.
Clearly, in the early days (given for example management’s initial optimism of a quick recovery) few would have expected revenue to be this far down at this point. Put differently, revenue was still down double digits YoY, and despite the return to profitability free cash flow breakeven is only expected by the end of the year.
While a stock pop might be warranted after beating estimates, the stock is already up quite significantly from its lowest point. So given that the recovery in financial performance is only happening gradually, this raises the question of how much further upside there could realistically be in the near-term.
In that regard, the first part of my thesis from a few quarters ago (that Intel’s stock had declined too much since a meaningful part of the financial performance was due to the inventory correction, which resulted in CPU sales well below sell-through rate, exacerbating the downturn) seems to have played out nicely. In contrast, some daring predictions by others such as that the stock would go to $20, $15 or so have not.
This still leaves a second part of the thesis, which is a full recovery in financials. However, this might take quite a bit longer given (1) the low likelihood of full recovery to a COVID-era PC TAM, (2) the data center market share erosion in the last few years, and (3) the many exited businesses. Nevertheless, there is also still a third part underway that should start to play a role going into 2024 and beyond, which is the actual technological turnaround itself.
So while the stock has recovered some of its losses, overall I would say that in the mid- to long-term outlook there are still more tailwinds than headwinds (most of which by now are already in the rearview mirror), which provides and still leaves the stock with a favorable investment opportunity.
Editor’s Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.
