Intel Stock: World-Class Financial Engineering (NASDAQ:INTC)

Investment Thesis

Despite some distractions like the financial engineering discussed here, overall Intel (INTC) stock remains heavily impacted by two near-term issues: the recessionary and post-COVID decline in PC demand and the lingering process delays while Intel catches up to regain leadership in 2025. Since both of these issues will most likely end up being rather temporary, investors with a longer-term mindset could benefit from the current stock weakness.

Financial engineering: Smart Capital

Intel initially announced its memorandum of understanding with Brookfield in February as part of its Investor Meeting, and as part of its “Smart Capital” strategy, which Intel first talked about late last year when it had delayed its 2021 Investor Meeting to February. Smart capital is an umbrella term for all initiatives Intel is undertaking to reduce its capital burden of building new fabs to catch up with demand from the shortages as well as to accommodate future anticipated demand due to growth.

The first pillar, and the key part to prevent future shortages, was a “shell-first” strategy, where Intel would build empty shells (~20% of the fab cost) before filling them with equipment. The idea is that in terms of lead times, it is actually the other way around, as it costs perhaps two years to build a shell but only one year to fill it with equipment.

Intel tried to spin this as an advantage, but obviously it is actually a burden to have to build empty shells first (about $2B per fab), since these investments obviously do not generate revenue until they are filled with equipment when the fab buildout is complete and the fab starts running. For clarity, an empty fab shell does not depreciate, but obviously does impact free cash flow.

Nevertheless, it is understandable that Intel wants to prevent future shortages, especially since Intel is now becoming a foundry. Given how early it is in the foundry start-up, there is likely a lot of uncertainty about the ramp of this business, so it is better to be safe than have to decline customers because there is no capacity available. If the business ramps quickly, these shells will immediately be filled with capacity.

The second pillar of smart capital was already part of the original definition of “IDM 2.0” when Pat Gelsinger became CEO: increase the use of third party fabs. Intel being able to leverage both its own fabs as well as foundries obviously makes it more agile to deal with things like shortages. Using foundries likely results in lower gross margins, but saves some capex investments like traditional fabless customers.

The third pillar are customer commitments. I would refer to TSMC (TSM), which seems to have benefited very nicely from Nvidia’s (NVDA) prepayments, which given the recent downturn ended up far too large, resulting in excess capacity. Nvidia had no other option than to swallow this capacity itself, resulting in price reductions and write-offs. Intel had previously also used this model for its NAND business.

The fourth pillar, which has been the subject of extensive public exposure and debate, encompasses government incentives aka subsidies aka handouts. Given the benefit of bringing in high wage high-tech jobs, traditionally such incentives have been on the order of 10% of the initial buildout cost. For example Intel got about $1B for its upcoming $10B Israel fab. However, given topics such as tech and supply chain sovereignty and geopolitics, the recent U.S. and E.U. CHIPS Acts are providing up to 30% incentives. Intel has argued that this is merely bringing Intel back to par with subsidies in Asia (including Taiwan).

Not so smart capital

So far, aside perhaps from the use of foundries, the four pillars of smart capital do not really change the underlying economics of the business. However, this is where the fifth pillars comes in, the recent SCIP (semiconductor investment program) announcement with Brookfield. Intel called this a “first of its kind”. While some media have compared the model with other, supposedly similar, co-investments programs of other foundries, Intel is correct that it is indeed a first of its kind program.

The ‘explain like I’m five’ version is that while Intel will still be operating its fabs as usual, the fabs (and their associated costs, revenue and profits) subject to the SCIP program will actually only be owned for 51% by Intel. At the time of the 7nm delay, people talked about having to spin off the fabs and/or the end of IDM. Well, this model is the closest Intel will likely ever get to spinning off its fabs.

The saying goes that there is no free lunch, and this also applies here. As mentioned, both the costs as well as the profits are split 51-49 between Intel and Brookfield. Basically, this means Intel is giving up near-term costs for long-term profits.

Put differently, imagine you have to money to build one fab. However, you see AMD (AMD) and others taking market share, which you want to prevent. Hence, you go to Brookfield, and they will give you the money to build two fabs, but in return they want the profits from one of the two fabs.

This is why I coined this the not so smart capital strategy, since for Intel shareholders, there is no benefit. Yes, Intel will be able to build more fab than it could have without the program (and without taking on debt), but exactly none of the surplus profits will go shareholders, as Brookfield will instead be the shareholder.

Specifically, Intel talked about receiving $15B from Brookfield to build out its two Arizona fabs worth $30B, confirming that this is like the “buy one get one free” sales tactics. Of note, Intel first announced these fabs in early 2021 as part of Pat Gelsinger’s Intel Foundry Services announcement, but it seems the cost (and with that the scope?) of the fabs has increased, since Intel initially talked about a $20B investment.

Intel fabs

Intel announced early this year that it had opened its $3B D1X Mod3 expansion in Oregon. Intel is further currently busy with equipping its $7B Ireland EUV expansion, which will produce Intel 4 next year.

Furthermore, Intel has several fab projects early in the buildout phase. As already mentioned, Intel is building a new $10B fab in Israel, and the groundbreaking of the two Arizona fabs ($30B) happened last year ahead of schedule. (These are the fabs subject to the initial SCIP program.) There are already four fabs on the same Arizona campus, with the latest $9B Fab 42 starting production in 2020. Note that Fab 42 is an (accidental) example of the “shell-first” strategy discussed above: it was first built about a decade ago (in 2012) as a 14nm fab, but was never filled with equipment due to the PC decline. (It was later announced as a 7nm fab by BK in 2017, but came earlier into service to deal with the shortages.)

In addition, Intel aims to build two brand new megafabs over the next decade, one in Ohio (“silicon heartland”) and one in Germany (“silicon junction”). At full buildout, these would likely encompass 6-8 fabs each, worth up to $100B at each site ($200B total). The initially pledged investment is about $20B for each. For reference, If Intel would get the full 30% offset for both fabs, then the total capital offset from government incentives would be $60B. Contrary to the SCIP program, this would be a true “buy two fabs, get one for free” handout.

Obviously, though, the scope of the project will be subject to the demand Intel can generate for its products. Pat Gelsinger had previously said he wanted to go bigger and faster, but in light of the current flip from shortages to oversupply/recession, it remains to be seen if either the subsidies or the demand will limit the pace of the buildout. Some investors may also note that TSMC had most recently actually increased its full-year outlook (although the high-end of the capex guide was reduced from $44B to $40B, but this may have been due to equipment shortages), showing that not every company and every sector is impacted the same.

Besides fabs, Intel is also investing in assembly and test (packaging) facilities, with a $7B investment in Malaysia and a $3.5B one in Rio Rancho.

Note that Intel does have some leverage in building new fabs, as even in case of underwhelming increase in demand, Intel could bring back some foundry capacity in-house, as this may yield improved economics once Intel regains process leadership and high process yields.

Capital allocation

In the wake of the Q2 downturn (the miss and lower, the reverse of beat and raise), some analysts have started to question the sustainability of the $6B annual dividend payout. Intel had already stopped the stock buybacks, but the big reduction in revenue and hence reduction in gross profit puts further pressure on the free cash flow, which due to the current “investment phase” Intel had already guided would be breakeven at best. Hence, Intel has already made some adjustments in order to keep the FCF guidance unchanged, although these adjustments were less than expected since Intel is now expecting more capital offsets in 2022 than previously forecast.

Still, the new SCIP further raises the questions if Intel should not have cut the dividend, or to simply take on more debt, since as discussed the only thing this program do is to reduce upfront capex investments but “in return” will impact future profits from those fabs, as 49% of the profits will flow to Brookfield instead. Of note, Intel’s Investor Meeting model assumed a rather sharp increase in FCF from (below) breakeven currently to 20% by 2026.

Financial engineering, again

This is not the first time Intel is playing the financial engineering game, for better or worse. Intel has previously already quit the NAND market, and in the wake of the gigantic Q2 and guidance misses, Intel immediately decided to also quit the (admittedly money-losing/unprofitable) 3D XPoint/Optane business.

In addition, late last year Intel further sought to IPO a minatory stake of Mobileye, reportedly by raising some $5B at a $50B valuation, although it is not sure if the plan will still continue given the market downturn.

Investor Takeaway

Pat Gelsinger had grand plans to restore Intel to its former glory when he became CEO. The three-headed dragon the consisted of his strategy encompassed following parts: restore process leadership (see Intel Stock: Turnaround Starts In 2025 (NASDAQ:INTC)), use this to restore product leadership (see Intel Stock: No Pain, No Gain (NASDAQ:INTC)), and build more fabs to accommodate the increased demand this will result in (both from entering new markets as well as retaking market share in existing businesses).

The latest SCIP announcement serves as the latest move in the Smart Capital strategy to reduce the cash flow burden of building new fabs. However, investors who saw the announcement likely were left with disappointment, as the program essentially says that Intel will build two fabs for the price of one… and indeed only one of the fabs will contribute to Intel’s profits, while the other fab will benefit Brookfield instead. Hence, there is no free lunch.

Overall, investors were looking for Intel to return to its engineering roots when Pat Gelsinger returned to Intel, instead of the financial engineering path that Bob Swan had threaded with the many buybacks. While Intel has indeed strongly stepped up its capex and R&D investments, perhaps due to the CFOs’ influence Intel is still experimenting with new ways to financially engineer its way back to greatness. While the E.U. and U.S. CHIPS Act handouts will save Intel billions in investments, SCIP does not seem to benefit shareholders long-term since Intel had already forecast that FCF would improve substantially over the next few years anyway.