Semiconductors: the cycle and the value chain

The sub-sector that plays by different rules

Everything in the software and internet articles in this guide assumes a business that, absent a major shock, compounds steadily: growth might accelerate or decelerate, but earnings do not usually swing violently from one year to the next for reasons entirely outside the company's control. Semiconductors break that assumption completely. A well-run chip company can post its best-ever earnings the year before a severe downturn, purely because of supply and demand dynamics across the industry rather than anything the company did wrong. Understanding why requires understanding the value chain first.

The value chain

Fabless, foundry, and IDM

The semiconductor industry organizes around three business models, and knowing which one a given company uses is the first thing an interviewer expects you to identify correctly.

A fabless company designs chips but does not manufacture them, outsourcing production to a third-party foundry. This model is capital-light relative to owning manufacturing: the fabless company avoids the enormous cost of building and running a fabrication plant (a "fab"), and instead focuses its capital on research and development and chip design, paying the foundry a manufacturing fee per chip produced. The tradeoff is that a fabless company does not control its own manufacturing capacity, so during periods when foundry capacity is scarce industry-wide, a fabless company may struggle to get enough production allocated to meet its own demand, regardless of how strong that demand is.

A foundry manufactures chips designed by others, essentially operating as a manufacturing service for fabless customers, and does not design or sell its own branded chips. Foundries are exceptionally capital intensive: building a leading-edge fabrication plant requires an enormous, multi-year capital commitment, and the technology inside that plant becomes progressively more expensive and complex to advance with each generation. Because of that capital intensity, the foundry business has consolidated around a small number of players capable of operating at the most advanced technology levels, while a larger number of foundries continue serving customers who need older, less cutting-edge manufacturing processes that still have plenty of real-world uses.

An integrated device manufacturer (IDM) both designs and manufactures its own chips in-house, controlling the entire process from design through production. Historically this was the dominant model in the industry; today it persists most prominently among a subset of large chipmakers, alongside memory chip manufacturers, who tend to keep manufacturing in-house because memory production benefits from tight integration between design and process technology. An IDM captures more of the value chain than a fabless company does, but it also carries the full capital burden of owning fabrication capacity, meaning an IDM's earnings are exposed to both product demand cycles and capacity utilization cycles at the same time.

The rest of the value chain: equipment, packaging, and distribution

Fabless, foundry, and IDM are the three headline business models, but the full semiconductor value chain has other links worth knowing, because interviewers occasionally hand you a company from one of these adjacent categories and expect you to place it correctly on the sub-sector map rather than lumping it in with a standard chipmaker.

Semiconductor capital equipment companies make the highly specialized machines that foundries and IDMs use to actually fabricate chips. This is arguably the most leveraged position in the entire value chain to the capacity-building cycle described below, since a fab has to buy equipment before it can add capacity, which means equipment orders tend to lead the rest of the industry's cycle rather than following it: equipment makers see demand pick up first when the industry commits to expansion, and see demand fall first when the industry pulls back on capital spending. After a chip is fabricated, it typically goes through packaging and testing, often handled by specialized companies rather than the chipmaker itself, before reaching distributors or going directly to device manufacturers. Each of these adjacent businesses inherits some version of the same cyclicality discussed below, but the timing and amplitude differ enough that treating "semiconductors" as one undifferentiated cycle, rather than a chain of related but distinct cycles, is a common and avoidable mistake.

Why the industry is cyclical

The core mechanism behind semiconductor cyclicality is a mismatch between how fast demand can change and how long it takes to change supply. Building new fabrication capacity, especially at the most advanced technology levels, takes years from the decision to build to the plant actually producing chips at scale. When demand is strong and prices are rising, the entire industry has an incentive to commit to new capacity at roughly the same time, since everyone is looking at the same demand signals. That capacity then arrives years later, often right as the demand that justified it in the first place has cooled, either because end markets slowed down or because customers over-ordered during the shortage and are now working through excess inventory rather than placing new orders. The result is a boom that seeds its own bust: capacity ordered during a shortage becomes the oversupply that causes the next downturn.

Inventory dynamics amplify this further. When chips are scarce, customers up and down the value chain (distributors, device manufacturers, and end customers) tend to over-order and stockpile, trying to secure supply and protect their own production, which makes real demand look even stronger than it is. When the cycle turns, those same customers work down their excess inventory instead of placing new orders, which makes real demand look even weaker than it is. This double amplification, on the way up and on the way down, is why semiconductor revenue swings tend to be sharper than the swings in the actual underlying end-market demand for the products chips go into.

Why semis trade differently: the mean-reversion multiple

This cyclicality produces a valuation pattern that looks backward from how most businesses get priced. For a steadily compounding business like a mature software company, rising earnings generally support a rising, or at least stable, multiple, since the market has reasonable confidence the higher earnings level will persist. For a cyclical semiconductor company, the market often does close to the opposite: multiples compress right as earnings peak, because sophisticated investors know a peak in a cyclical industry tends to be followed by a trough, and they price the stock off some normalized, mid-cycle earnings level rather than the current peak number. Conversely, multiples often expand when earnings are depressed at a cyclical trough, because the market is pricing in an eventual recovery.

This is precisely why comparable companies and precedent transactions typically carry more analytical weight than a DCF for a semiconductor name: a DCF requires picking a "normal" earnings level to project forward, and getting that normalization wrong (using a peak year as if it were sustainable, for instance) produces a badly distorted valuation. An experienced analyst looks at where a semiconductor company sits in its cycle, often using metrics like the book-to-bill ratio, before applying any multiple at all.

Book-to-bill and other cycle-tracking metrics

The book-to-bill ratio, the dollar value of new orders booked in a period divided by the dollar value of product shipped (billed) in that same period, is one of the most closely watched real-time indicators of where the semiconductor cycle stands. A ratio above one means new orders are coming in faster than product is shipping, generally a sign of strengthening demand or tightening supply; a ratio below one suggests the opposite. Gross margin is another useful cycle indicator, since semiconductor gross margins tend to expand when the industry is capacity-constrained (pricing power favors the manufacturer) and compress when there is excess capacity chasing scarce demand (pricing power shifts to the customer). Neither metric on its own tells the whole story, but together with unit shipment trends, they give a much better read on cycle position than a single quarter's reported earnings.

How deals differ in semiconductors

Semiconductor M&A carries a distinct regulatory risk that shows up far less often in a plain software or internet deal: national security review. Chips are considered critical infrastructure by governments around the world, and a cross-border semiconductor acquisition, or even a domestic one that would meaningfully consolidate a critical part of the supply chain, routinely draws intense antitrust and national-security scrutiny. Broadcom's attempted hostile takeover of Qualcomm is the standard reference point here: the deal was ultimately blocked by the U.S. government specifically on national-security grounds, not ordinary antitrust economics, which is a distinct category of deal risk a candidate should be able to name when discussing semiconductor M&A specifically. This regulatory overlay, combined with the sector's cyclicality, is why semiconductor deal structures and diligence differ meaningfully from the take-private and carve-out patterns common in software, covered in TMT deal structures.

How this differs from the rest of the map

Every other technology sub-sector in this guide, software especially, gets easier to value the more predictable and recurring its revenue becomes. Semiconductors invert that logic: even a well-run, well-positioned chip company remains cyclical because the cyclicality comes from the industry's supply-and-demand structure, not from any flaw in the company's execution. That is the single fact worth holding onto across every semiconductor question an interview throws at you, whether it is about valuation, about a deal, or about pitching a semiconductor stock, covered in pitching a tech stock in a TMT interview. It is also the sharpest possible contrast to the software valuation logic covered elsewhere in this guide, and interviewers frequently ask candidates to articulate that contrast directly, so it is worth being able to state in one sentence: software is valued on the assumption that recent growth will continue, while semiconductors are valued on the assumption that recent results will not.

Practice question

A semiconductor company just reported its highest quarterly earnings in its history. Is now a good time to buy the stock?

Not necessarily, and the fact that it's a record quarter is actually a reason to be more careful, not less, because semiconductors are a deeply cyclical industry. Chip supply takes years to build, since new fabrication capacity has a long lead time from the decision to invest to the plant actually producing at scale, while demand can shift much faster. That mismatch means the industry tends to overbuild capacity right as a period of strong demand starts to fade, and inventory dynamics make it worse: customers over-order during shortages, which makes demand look stronger than it really is, then work down that excess inventory during a downturn, which makes demand look weaker than it really is. A record quarter could reflect the peak of that cycle rather than a new sustainable level, which is exactly why sophisticated investors often compress the multiple right when earnings peak, pricing in the next downturn rather than extrapolating the current number forward. Before I'd get comfortable buying, I'd want to check the book-to-bill ratio to see whether new orders are still outpacing shipments, look at gross margin trends for signs of pricing pressure creeping in, and understand where end-customer inventory levels stand across the value chain. If those signals suggest the company is near a cyclical peak rather than early in a sustained upswing, I'd be cautious regardless of how good the headline number looks.

What the interviewer is listening for: Whether you instinctively distrust a single strong data point in a cyclical industry and reach for cycle-position indicators (book-to-bill, gross margin trend, channel inventory) instead of taking reported earnings at face value.