Description

Texas Instruments (aka "TI") is the largest analog semiconductor company in the world.

What does an anlog semiconductor company do?

The semiconductor industry is broadly divided into analog and digital semiconductor products. Analog semiconductors condition and regulate “real world” functions such as temperature, speed, sound and electrical current. Digital semiconductors process binary information, such as that used by computers.

What does TI do?

The company reports in three segments - analog, embedded and other. Analog is what I just described above. Embedded is similar but slightly different. Embedded devices are the "brains" of many types of electronic equipment. They are designed to handle specific tasks and are optimized for some combination of performance, power and cost. They also tend to have a software component (mostly written by the customer). Other at TI is comprised of three businesses - DLP, Custom ASICs and calculators. Other did $1.2b of revenue in 2019, made up of $700-800m of DLP, ~ $400m of calculators and ~ $130m of custom ASICs. DLP is short for Digital Light Processing and these are chips that go into projectors. TI has ~ 100% market share in the cinema space and ~ 50% market share in the educational/commercial space. This is expected to be a slow but secular decliner and is invested-in accordingly. Calculators is exactly what it sounds like. The death of this business has been predicted for many decades now and it too is in slow but secular decline (low single digit volume declines in a typical year). This is also expected to be a slow but secular decliner and is invested-in accordingly. Custom ASICs is not a focus for the company and exited the year at a run-rate of $60m of revenue (less than half a percent of total sales) and likely continues to decline. Other has been a significant drag on total revenue for the last five years (-11% five year CAGR) but that probably moderates quite a bit going forward as the Custom ASIC business was the primary driver of that and is now almost gone and is also less of a drag on consolidated results as Other was ~ 8% of TTM revenue as of Q1.

TI is extremely diversified and sells their products to over 100k customers. TI has - by their estimates - 19% of the $54b analog semiconductor industry and this number has been slowly but steadily climbing over time. This breadth and diversity in their portfolio (and the associated lack of meaningful product cycle gyrations) is seen by some market participants as boring but if you're investing over a longer-time horizon it can give investors comfort that there is no one product or cycle that you're overly reliant on for the thesis that I lay out below to be correct.

TI splits out their revenue as 36% industrial, 21% automotive, 23% personal electronics (phones, PC, tablets, TVs, gaming, etc), 11% communications equipment (wireless/wired infrastructure), 6% enterprise (data centre + enterprise computing) and 3% other (which includes the eponymous calculators). It's worth flagging that this mix has been shifting towards industrial / automotive (the combined share was 44% in 2014 vs 57% in 2019) and away from personal electronics / communications equipment (the combined share was 46% in 2014 vs 34% in 2019). The higher-growth areas are (unsurprisingly) outgrowing the lower-growth areas and the business is slowly mix-shifting over time to a higher-growth overall portfolio. 

Why is this a good business?

TI benefits from a number of classical competitive advantages and favourable aspects to its business model.

1) Economies of Scale

TI is the largest player in the analog market. This benefits them in terms of both manufacturing and in fixed cost leverage.

TI's size makes them one of the few players capable of economically justifying manufacturing their products in-house (and cutting out the foundry partner's gross margin) and effectively the only player capable of filling a new 300mm fab quickly enough to justify the upfront cost (more on this later).

TI also uses their scale to better leverage fixed costs than competitors are able to. If you look at TTM Fixed Costs (R&D + G&A) as a % of Sales, for the last ten quarters, TI has averaged ~ 21%. If you look at the same metric for its two closest peers - ADI & Maxim - those numbers are ~ 31% and ~ 32% respectively. Framed differently TI generated (on a TTM basis) ~ 78% more revenue than ADI & Maxim combined but only spends ~ 27% more on R&D + G&A and as a result generates ~ 95% more EBIT.

2) Switching Costs

The products that TI manufactures and sells tend to have long lifespans. This is a function of the fact that the devices that they are going into tend to have long lifespans and the fact that they rarely get switched out. The bulk of the effort to design, test and integrate an analog chip into a device is done early on in a device's lifecycle and, once complete, doesn't need to be replicated. The upside to a customer from replacing a TI chip down the line is minimal given their very low contribution to the typical bill of materials while having to replicate all of that upfront activity would require significant time, internal focus and spending. The result of this dynamic is that once an anlog chip wins a socket, it tends to hold onto that socket for as long as the device is being produced.

The proof is in the pudding. Analog Devices has a pretty comparable product portfolio to TI and has publicly stated that more than 40% of their FY19 revenue was generated from products released a decade or more ago.

3) Small Percentage of Customer Costs

This switching-cost dynamic is made more acute by the very low absolute prices of the stuff that TI sells. The average TI chip goes for $0.35c. This benefits TI as it makes their products less of a focus for aggressive price negotiation and it meaningfully reduces the economic incentive to go through the hassle of swapping a TI chip out once it's already being used.

4) Mission-Critical Products

The stuff that TI makes and that goes into all manner of devices may be cheap but it is also mission-critical. If these chips fail, in most cases the device fails too. Though less important than other elements that comprise the company's strength, TI's brand and reputation do infer that you as a customer are getting a well-designed, well-made and well-tested product that is likely to last for the life of the device and for some customers that will be a swaying factor.

5) Niche Markets

Though TI generates ~ $15b of revenue in aggregate, the reality is that this is made up of myriad niche markets. This means that each one of those markets, in isolation, is unlikely to be big enough for a new entrant to justify throwing serious dollars at breaking into (particularly when they factor in the switching cost point outlined above).

6) Structural Growth

TI has grown its core analog + embedded business at a 5-6% CAGR over a very long period of time. If you look at TI in 2006, the business did $6.3b of core analog + embedded revenue. That number was $14.3b in 2018 and $13.2b in 2019. If you strip out the revenue that they acquired when they bought National Semiconductor in 2011 (~ $1.5b) that implies a LT organic CAGR of 6.1% / 5.3% depending on whether you use 2018 or 2019 as an endpoint. I think 2018 had some froth in it but was closer to normal than 2019 so would pencil in normalized analog + embedded growth in the 5.5% to 6% range. This range squares with the fact that semiconductor penetration is structurally rising and so we should expect to see GDP+ growth here over time.

What does this look like in numbers?

TI generates ~ 64% gross margins and ~ 80% incremental gross margins.

TI generates ~ 43% EBIT margins.

TI generates ~ 100% returns on tangible capital (PP&E + NWC) with a tangible invested capital base of ~ $6b supporting an average of ~ $6b of NOPAT over the last three years.

What is the market missing?

TI is currently in the middle of a cyclical downturn in the analog semiconductor market. TI has seen Y/Y revenues (ex other) since Q1 2018 of +14% / +11% / +5% / -0% / -5% / -9% / -11% / -9% / -6%. The market peaked in the middle of 2018, had a pretty nasty 2019 (given the weak global industrial economy) and was starting to show some signs of an emergence from that downturn before COVID arrived.

The sell-side aggressively rebased earnings estimates after the company reported Q1 and guided for Q2 in the middle of April by (I'm paraphrasing) "assuming a repeat of their GFC experience" not because that's what they necessarily think will happen but that's the last time they had a large negative economic shock to the system so they don't have a better analog to use. The sell-side cut numbers to the new guidance which leaves the stock looking very expensive on 2020 earnings. Revenue is modelled down 12% Y/Y in 2020 after being down 8% Y/Y in 2019 while EPS is expected to be ~ $4 after ~ $5.3 in 2019, putting the stock on ~ 30x consensus 2020 estimates.

The sell-side is then growing their 2020 estimates at a modest cadence from 2021 onwards but in such a way that leaves 2022 revenue estimates (ex other) implying effectively zero growth over the four years from 2018 to 2022.

That's extremely unlikely to be correct. If you take 2018 as a base, haircut it by ~ 5% for the fact that TI core revenues were running a smidge above trend that year and then grow the 2018 haircut base at their historic mid-single digit organic growth rate, by 2023 you'll be looking at a company doing ~ $20b of revenue and almost $10 of EPS.

What's the upside?

That logic would put the stock on 12x my 2023 EPS number. That's way too low for a company with (a) this depth & breadth of moat, (b) robust structural growth drivers and (c) an exceptional management team. The universe of high-quality industrial companies that offer some combination of the strengths I've laid out above typically trade in the 20-22x range offering plenty of upside over the next 2.5 years if TI ends up on that multiple in 2022/23. If it can trade on 20x my 2023 EPS of $9.75 that will be a stock price of $195. If you factor in $12 of dividends you'll get along the way, that's $207 or +70% vs the current price in two and a half years.

What's the 300mm fab opportunity?

The analog semiconductor industry basically makes chips on two types of wafers - 200mm wafers and 300mm wafers. These numbers refer to the diameter of the circle of silicon that enters the fab to be processed into hundreds of tiny chips. TI generated their first revenue on 300mm wafers in 2010. In 2019 they generated ~ $4.5b of analog revenues (or ~ 40% of total) on 300mm with the balance on 200mm. The way this mix grows is that TI manufactures all new products on 300mm while keeping the vast majority of legacy products on 200m because it's uneconomical to switch over any one low-volume product line once it's up and running and fine tuned on the 200mm architecture.

The cost differential in moving from 200mm to 300mm is ~ 40% of silicon and ~ 20% of total cost of goods. That means that a 60% GM product on 200mm can become a 68% GM product on 300mm (as the 40% cost of revenue gets chopped by 20%, i.e. 8%). 

The gradual transition of TI's mix of products from 200mm to 300mm is a multi-year gross margin tailwind and part of the reason why - in addition to standard operating leverage - their incremental gross margins are ~ 80% while their reported gross margins are ~ 64%. The pace of transition to 300mm is dictated (i.e. constrained) by the slow pace with which legacy products ultimately die so the impact in any given year is relatively modest. I estimate that the transition is likely to generate 30-40 bps of GM expansion for many many years to come.

Why will this benefit not be competed away for a long time? You probably need $5b of revenue flowing through a newly-built 300mm fab today to generate an economic return on the CapEx. It has taken TI almost a decade to get their first 300mm fab up to $4.5b of revenue. The next biggest player in this space is ADI and they generate less than half TI's revenue. If it took TI a decade, it would probably take ADI at least two. That's just too far out to underwrite. If a foundry partner did this instead that could be more viable from a scale perspective but that would likely involve paying away a significant share of the economic benefit to that partner ( TSMC makes ~ 50% gross margins) which might render it uninteresting to the potential customers and the foundry would still face the issue of it taking a long time to fill-up the fab given the dynamic around long lifecycles and limited incentive to switch legacy products.

This is all to say, TI has an idiosyncratic driver of margin expansion that is likely to continue to drive modest but consistent excess gross margin expansion for many many years to come and is unlikely to be accessible to other market participants or competed away.

What's the distribution opportunity?

TI has historically relied (like the rest of the semiconductor industry) on the distribution channel to help get their products into the hands of as many customers as possible. They have relied on the channel for two things - help selling and help distributing. They gradually cut out the first of these two activities over the last decade as they pulled back the incentives they offered distributors for what they called "demand creation" realising that in many instances, the customer knew what they wanted so they were effectively paying something for nothing. That role of the channel for TI is now effectively gone. That said, until recently they have still relied on the channel for a large chunk of their distribution. In 2019 they generated ~ 65% of revenues from distributors vs ~ 35% from end customers. This year they are executing a pretty major change here. They announced in October 2019 that they would be consolidating their distributor relationships from 8 to 1 (the one being Arrow). They put in place transition agreements with the other 7 in Q4 2019, started to transition from 8 to 1 in January 2020 and hope to have completed the transition by the end of the year.

The distributors make a 3-5% gross margin. They were probably on the low-end of that range with TI products given TI's scale and their distribution-only role for TI products. There will likely be some incremental costs to facilitate direct sales that used to be carried out by distributors but TI already has an ecommerce-like website where customers can research and order products, they already have the infrastructure in place to manage a $5b direct business so it's a question of scaling something up rather than building something new and a lot of the inventory that was sitting on distributor's shelves was there on consignment so the impact to NWC is likely to be modest.

It's hard to get precise with the $ opportunity here as it's a function of the portion of revenue that shifts to direct, the improvement in terms TI got from Arrow to become their exclusive partner (not disclosed) and the incremental costs that TI ends up having to incur to support whatever demand moves from the channel to direct. That being said, my rough math suggests you could be looking at a low-single digit uplift to total gross profit $s out of this. It's not game-changing but it's nice-to-have.

What about management?

This is one of the best and most disciplined management teams you are likely to find in the world. They act like owners. They spell out in crystal clear terms their operational and capital allocation strategy at every single opportunity they get and they are lased-focused on executing those strategies. I've spoken to many formers and if you ask any one of them to use a single word to describe the CEO, it is almost always focus.

This slide deck is mandatory reading for anyone interested in TI (and useful to point other management teams towards if you ever get asked, who do you think does capital allocation well). The crux of the capital allocation piece is that they have a very high hurdle for M&A and will return 100% of FCF to shareholders via a mix of dividends and buybacks. That's not 50-100%... that's not 75-100%... that is an unwavering commitment that every cent of FCF is coming back to you.

If we look at the last decade, TI generated $49b of CFFO and incurred $8b of CapEx. Management spent $6b on M&A (all in 2009/10/11) and returned $22b via share repurchases and $16b via dividends.

Catalyst

I don't think you really need catalysts while owning a business that is this good & this well-run but I do think street estimates for the next few quarters are way way too low so you have a nice combination of a reasonably cheap, very high quality business that I think is poised to meaningfully beat estimates for the next few quarters.