KLA Corporation: The Tollbooth of the Silicon Age

I. Introduction & The Tollbooth of the Silicon Age

The speck that costs a fortune

Picture a leading-edge fab in southern Taiwan at three in the morning. The air handlers hum. Robots glide along ceiling tracks carrying sealed pods of 300mm wafers. There are no workers in bunny suits leaning over microscopes. The judging is done by a machine the size of a small car, which sweeps light across a photomask. A photomask is the master stencil from which every chip in a production run is printed. The machine is looking for a flaw far smaller than a wavelength of visible light.

When it finds one, a software flag lights up on an engineer's screen. The stakes explain why this company exists. A defect on a mask is not a single ruined chip. It is a rubber stamp with a crack in it, pressing the same mistake onto every exposure of every wafer until someone notices. On a line printing AI accelerators and flagship smartphone processors, "until someone notices" can be measured in millions of dollars an hour.

Here is the cold open for this episode: what if the biggest bottleneck in making an AI superchip is not lithography, etch, or raw silicon, but finding the flaw? That is the job KLA Corporation has done, more or less, for half a century.

Sizing the titan

KLA (NASDAQ: KLAC) runs its business from Milpitas, California, with a second headquarters in Ann Arbor, Michigan¹. It closed its fiscal year in June 2026 with about $13.6bn of revenue and an operating margin of almost 42%[^2]. Those two numbers together are the whole puzzle of this story. Semiconductor equipment is heavy, cyclical, engineering-intensive hardware. Businesses like that usually earn margins in the twenties. KLA earns margins that look like enterprise software.

The reason is its position. Industry surveys have for years put KLA at more than half of the global market for semiconductor process control, which covers inspecting for defects and measuring what was built. Its nearest rival holds a fraction of that share⁷. Rating agencies make the same point in their own language. Fitch rates KLA 'A' and cites dominant technology leadership in process control as the anchor of that rating².

A simple way to picture the chip-equipment industry: ASML owns the printing press, which is lithography. Applied Materials and Lam Research lay down and carve away the layers, which is deposition and etch. KLA decides what lives and what dies. Its tools tell a fab whether the other companies' tools did their job.

The core paradox

There is a counterintuitive engine inside that position. As transistors shrink toward atomic dimensions and chips grow upward in 3D stacks, the number of things that can go wrong rises faster than the number of things being built. Each new node adds process steps and mask layers, and each added step is one more place for a defect to hide. Inspection and measurement therefore tend to take a growing share of fab equipment budgets over time, even when total spending stalls. Management has made this argument for years, most fully at its 2024 investor day[^7]. This episode tests whether the record supports it.

The questions that frame the episode

Four questions run through everything that follows:

1. Can KLA keep operating margins above 41% if China, which made up a startling share of revenue in recent years, falls back toward historical levels?
2. KLA more than doubled the receivables it sold to banks last year. Does that signal customers paying late?
3. How exposed is KLA's shareholder cash-return machine to one customer, TSMC, and the pace at which it moves between process nodes?
4. Will advanced packaging and extreme-ultraviolet mask complexity permanently enlarge KLA's slice of fab spending?

The roadmap runs from two engineers who saw a problem nobody else wanted, through a merger of rivals, a failed megamerger, a costly Israeli acquisition, a China windfall and the export-control squeeze that followed, to the cash machine as it stands tonight. First, a trip back to the mid-1970s, when chips were still inspected by people squinting into eyepieces.

---

↑ Back to Top

II. The Optical Gamble & Founding KLA Instruments (1975–1985)

Two engineers and a blind spot

In 1975, Silicon Valley was obsessed with printing. Every equipment company wanted to build a better aligner or a faster stepper, the machines that projected circuit patterns onto silicon. Kenneth Levy and Robert Anderson had worked on that side of the business and saw something the industry was overlooking⁵.

The masks themselves were being checked by hand. Technicians sat at microscopes and scanned patterned glass plates for defects, field by field, hour after hour. Human eyes get tired. Human attention drifts. And as circuit features shrank, the defects that mattered moved below what a tired inspector could reliably catch. Chipmakers were losing yield to flaws they could not see, and they often could not work out whether a bad wafer came from the mask, the process, or the tool.

Levy and Anderson founded KLA Instruments in 1975 on a single premise: automate the inspection of photomasks with optics, electronics and software that compared what was printed on a mask with what should have been printed⁵. The name was a blend of the founders' initials. The company's ambition was more distinctive than its name.

The founders

Levy was the product visionary: an engineer who thought in systems and was willing to bet a company on image processing well before computing power made it easy. Anderson was his counterweight, a disciplined operator who worried about cash, customers, and whether the next machine could actually be shipped. That pairing, a technologist who stretches and a manager who holds the line, would repeat itself for decades. Its modern version is the partnership between CEO Rick Wallace and CFO Bren Higgins.

The technical leap was combining high-speed optical scanning with custom digital image processing. In plain terms, the machine takes a very fast, very detailed photograph of the mask. It then compares that photograph either against a neighbouring identical pattern on the same mask or against the original design data. Anything that differs is a candidate defect. That is easy to say and extremely hard to do at production speed without drowning the operator in false alarms. The false-alarm problem, separating real killers from harmless noise, has remained the central difficulty of the business ever since. It is why the company's software and algorithms are worth as much as its optics.

The first product and the IPO

KLA's automated photomask inspection system reached the market in the late 1970s and found customers quickly among the merchant chipmakers of the era⁵. In 1980 the company went public⁵. It listed just as the personal computer was about to create an enormous new appetite for memory and logic chips.

That listing matters to the long story for one reason: the customer relationships. KLA's engineers were embedded with the process engineers who ran the fabs. When a line had a yield problem, KLA was in the room. That proximity became a data advantage, a trust advantage, and eventually a switching cost, and all of it predates any of the numbers in this story.

The garage-era lesson

The strategic insight of the first decade is easy to state and rarely copied. Do not build the generic tool everyone wants. Build the diagnostic tool without which nobody can prove their other tools work. A deposition tool makes a film. KLA's tool tells you whether the film is right. The first can be benchmarked and competed on price. The second sits at the point where the fab's money is either made or lost.

The weakness was that KLA inspected masks, a narrow slice of the process. To grow, it would have to follow the defect from the stencil onto the wafer itself. It would also have to survive the first great shock in chipmaking history, which arrived from Japan.

---

↑ Back to Top

III. Surviving the Japan Memory Shock & The In-Line Pivot (1985–1996)

When the customers disappeared

By the mid-1980s, Japanese memory makers had overrun the global DRAM market. American merchant chipmakers retreated from memory into logic and microprocessors. Many of them slashed capital spending in the process. For an American equipment vendor whose customers were those American fabs, the shock was existential. Capital equipment is the most cyclical corner of a cyclical industry: when a chipmaker's revenue falls a little, its equipment budget can fall a lot.

KLA's response, as Levy later recounted, was to follow the demand rather than wait for it to return⁵. The company pushed into Japan and put its own people inside Japanese fabs. That turned out to be more than a sales channel. Japanese memory producers ran their lines with a ruthless attention to yield. Working alongside them taught KLA what the most demanding customers in the world expected from inspection, and those expectations became the product roadmap.

This is the first test of a recurring claim in this story: that KLA's customer base is not a vulnerability but a source of learning. The Japan episode supports it in its original form. The company survived because it shifted geography quickly. Four decades later, the same reflex would be tested in reverse, as sales shifted out of China.

From stencil to wafer

The more important decision of this era was technical. KLA moved from inspecting static photomasks to inspecting patterned wafers in-line, meaning during manufacturing rather than at the end⁵.

The economics are the easiest way to understand why this mattered. A modern chip passes through hundreds of process steps over weeks. If a killer defect is created at step 10 and found only at final electrical test, every step from 11 onward was spent on a wafer that was already dead. Chemicals, tool time, electricity and weeks of cycle time were all wasted. Catching the defect right after step 10 saves everything downstream. It also tells engineers which tool caused the problem while it is still causing it. In-line inspection turned KLA from a quality checker into a real-time control system for the entire factory.

It also changed how the business grew. A mask inspection tool is bought per mask shop. An in-line wafer inspection tool is bought in proportion to the number of wafers and the number of critical steps. As fabs grew and processes lengthened, inspection spending scaled with the complexity of manufacturing itself. This is the origin of the "process control outgrows fab spending" thesis.

The competitive limit

By the early 1990s KLA was the dominant name in finding defects on masks and wafers. But process control has two halves. Inspection asks, "Is there something here that should not be?" Metrology asks, "Is what should be here the right size and thickness?" The second half belonged largely to another Silicon Valley company, Tencor Instruments, which measured thin films and surface topography with great precision.

Chipmakers increasingly wanted both answers from one partner, on integrated software. The stage was set for a courtship between two proud companies that had spent twenty years competing for the same yield engineers' attention.

---

↑ Back to Top

IV. The $1.3 Billion Equal Merger: Creating KLA-Tencor (1997)

The second attempt

Silicon Valley remembers failed deals longer than successful ones, and KLA and Tencor had a failed one behind them. Earlier talks in the early 1990s had collapsed over valuation and culture. By late 1996 the math had changed. Chip features were heading below a quarter of a micron. Customers wanted a single yield-management platform that could detect defects and measure films. The research budgets needed to stay ahead in electro-optics were growing faster than either company could comfortably fund alone.

Tencor had its own founding story. Karel Urbanek, a Czech-born engineer, had built the company around surface profiling, film-thickness measurement and laser-scanning contamination detection. Its tools were in the same fabs as KLA's, answering different questions.

The deal

In 1997 the two companies combined in an all-stock merger of equals valued at roughly $1.3bn, creating KLA-Tencor⁵. No cash changed hands. Ken Levy became chairman, and Tencor's leadership took a prominent executive role in the combined company.

The strategic logic was almost unusually clean:

• Zero product overlap. KLA found defects; Tencor measured structures. There was little duplicated hardware to rationalise and few product lines to kill.
• Shared research economics. Both companies were spending heavily on optics, lasers, sensors and image-processing algorithms. A combined revenue base let those costs be spread over far more tools.
• One customer workflow. A yield engineer could now get defect data and measurement data from one vendor on software designed to connect them.

That last point is the heart of what the combined company called yield management. The defect map tells you where something went wrong. The metrology data tells you why: a film too thick, a line too wide, a layer misaligned. Together they close the loop. The engineer can adjust the upstream tool and confirm the fix.

Integration and the Asian crisis

The merger closed into one of the worst possible windows. The Asian financial crisis of 1997–1998 crushed Korean and Southeast Asian chip spending just as two sales forces, two software stacks and two engineering cultures were being merged. The combined company survived it, and over the following years the "yield management" category it created became an industry standard.

It is tempting to call the merger flawless. The more careful verdict is narrower. It worked because it satisfied a condition that most mergers of equals do not: the two businesses were complements, not substitutes. The company's later history would test the same M&A instinct twice more, once as a target and once as a buyer, with very different results.

The practical outcome for the industry was that competitors trying to enter inspection now faced an integrated incumbent whose tools and software were already woven into every major fab's workflow. For the next leader, the task was no longer invention. It was turning a technology franchise into a financial one.

---

↑ Back to Top

V. The Wallace Era, The 2006 Options Storm, and Platform Scaling (1998–2018)

A new CEO inherits a storm

Rick Wallace became chief executive in January 2006⁸. He was not a newcomer. He had joined KLA in 1988 and rose through application engineering, sales and general management, spending years close to the customers' fabs⁸. His style has consistently been described as calm and operator-first: less a visionary founder than a steward who tightens the machine one bolt at a time.

He took over just as the company was pulled into the stock-option backdating investigations that swept Silicon Valley in 2006. Those led to restatements and leadership changes at many firms, including KLA-Tencor, where Ken Levy's long role at the top ended that year. For the new CEO, the first task was not strategy. It was rebuilding governance credibility.

The legacy of that storm is visible today in unglamorous places. The chair is now an independent director, Robert Calderoni, rather than the CEO, and eight of nine board members are independent[^3]. Directors and executives are barred from hedging or pledging their shares[^3]. Shareholders approved executive pay with about 92% support at both the 2024 and 2025 meetings[^3]. None of these facts prove good behaviour. Taken together, they suggest that the board that emerged from the 2006 scandal has not produced a new one in the two decades since.

Twenty years at the top

Wallace has now led KLA for more than twenty years. Bren Higgins has been CFO since 2013⁸. That longevity is rare in capital equipment, and it is the right lens for the long-run numbers. Over the ten years to June 2026, revenue compounded at about 16% a year and net profit at about 21%[^2]. Operating margin rose from the low twenties in FY2015 to above 40%[^2]. The whole of that record belongs to one management regime, so it is a fair test of that regime.

The test is not all flattering. The margin record includes FY2019 and FY2020, when operating margin fell back to about 30%[^2]. That was the period that absorbed the Orbotech acquisition, which comes next. The long-run trend is up, but it is not a smooth staircase. The regime has made at least one expensive mistake, and it has been reversed only recently.

Pay is large. Wallace's reported compensation was about $30m in FY2026[^3]. It has risen alongside operating profit and shareholder returns. The SEC's "compensation actually paid" measure, inflated by the rise in the stock price, came to more than half a billion dollars[^3]. Investors can reasonably debate the size. The structure at least ties the CEO's fortune to the share price.

The tuck-in years and the 2008 test

Through the 2000s, KLA-Tencor extended its product range mostly with small acquisitions of specialised measurement technologies rather than big bets. The Great Financial Crisis was the real stress test. Chip spending collapsed, fab utilisation fell, and equipment orders nearly vanished for a few quarters. The company cut costs but kept funding its core optics research. The later rebound confirmed an old lesson from the Japan era: in process control, the vendor that keeps investing through a downturn owns the next node.

The megamerger that never happened

In October 2015, Lam Research agreed to acquire KLA-Tencor in a cash-and-stock deal valued at about $10.6bn[^11]. On paper it made sense for Lam: combine the tools that etch and deposit layers with the tools that inspect them, and sell a more complete package to chipmakers.

That was exactly what worried the U.S. Department of Justice. Regulators feared that a combined company could favour its own process tools, and the chipmakers that relied on KLA's independent measurements could lose their neutral referee. In October 2016 the two companies abandoned the deal in the face of that opposition[^11].

For KLA, the failure was arguably the best thing that could have happened. It stayed independent, and independence is part of the product. A company that grades every other vendor's tools cannot also be one of those vendors without inviting suspicion. The DOJ's objection quietly confirmed what the company's customers already believed: KLA's neutrality has economic value. This point returns in the moat discussion.

The capital return formula

The post-Lam years established the financial template that defines KLA today. The company buys back shares steadily and raises the dividend, rather than piling up cash. The balance sheet tells the story in one number. Shareholders' equity has been kept small relative to profits because so much cash has been returned. That is why reported return on equity is in the high double digits, and why it should not be read as purely operational excellence: part of it is simply a thin equity base[^2].

That discipline would soon face its sternest test, because in 2018 the company did something it had not done in decades: it made a big, expensive acquisition outside its core.

---

↑ Back to Top

VI. The Orbotech Acquisition and the Advanced Packaging Frontier (2018–2024)

A bet announced, a year of waiting

In March 2018 KLA-Tencor agreed to buy Orbotech, an Israeli maker of inspection and manufacturing tools for printed circuit boards and flat-panel displays, for about $3.4bn in cash and stock[^10]. Orbotech also owned SPTS Technologies, a UK-based maker of etch and deposition tools for specialty semiconductors and packaging. Closing required approval from regulators around the world, including China's market regulator. The deal finally completed on February 20, 2019[^10].

The same year, the company dropped "Tencor" and renamed itself KLA Corporation. It later opened a second headquarters in Ann Arbor, Michigan, to draw on engineering talent in the Midwest and bring the company closer to automotive customers⁶.

The strategic logic

The argument for Orbotech was about where chip performance would come from next. For fifty years, performance improved mainly by shrinking transistors on a single die. As that got harder and more expensive, the industry began to gain performance by packaging: placing several chips side by side or stacked on top of each other, connected by dense wiring. TSMC's CoWoS technology, which joins an AI processor to stacks of high-bandwidth memory, is the most famous example today.

Packaging creates new inspection problems. Bumps, micro-joints, through-silicon wires and substrate traces all need checking, and a single bad connection can kill a multi-thousand-dollar AI module. Orbotech brought inspection capabilities for boards and substrates. SPTS brought process tools used in packaging and in compound semiconductors such as silicon carbide and gallium nitride.

The trap inside the deal

Orbotech also brought a business that was much less attractive: equipment for flat-panel display makers. That market was dominated by large panel producers, increasingly Chinese, whose spending swings wildly and who were steadily adopting cheaper domestic suppliers. It was the opposite of KLA's core: a buyer-driven market with no durable technical moat.

Here is the historical falsification test for the claim that Wallace and Higgins are disciplined capital allocators. The record says the display half of Orbotech did not work. KLA recorded impairments of goodwill and intangible assets in its PCB, display and component inspection unit, including about $290m in FY2024[^5], with further charges in FY2025 as it wound down display manufacturing[^4]. Across the two years the write-downs came to roughly half a billion dollars. The segment lost about $281m in FY2025[^2].

In FY2026, with display largely behind it, the segment swung to a profit of about $108m[^2]. The auditor removed the impairment of that unit from its list of critical audit matters, leaving revenue recognition as the only one[^2].

Weighing the verdict

The honest verdict has two halves. On the strategic question, Orbotech looks right in hindsight. Packaging has become one of the fastest-growing parts of the chip equipment market, and KLA now has product lines there. On the price and scope question, the record rejects any claim of flawless underwriting. A meaningful part of what KLA bought turned out to be worth much less than it paid. The mitigating fact is that management eventually wrote it down and exited rather than defending it indefinitely. It took about five years to do so.

For investors, this narrows the "disciplined allocator" claim to a smaller, more accurate version. KLA's organic engine and its buyback discipline have been excellent. Its record on large acquisitions outside its core is mixed. That matters because the company's growth ambitions in packaging could tempt it into another one. The KPI to watch is simple: whether the packaging and specialty businesses grow profitably without further write-downs.

Meanwhile, the biggest swing in KLA's recent history had nothing to do with M&A. It came from a single country and a single regulatory document.

---

↑ Back to Top

VII. The $4 Billion China Windfall and the Export Control Squeeze (2020–2026)

October 7, 2022

On October 7, 2022, the U.S. Commerce Department's Bureau of Industry and Security published sweeping new export controls on advanced computing and semiconductor manufacturing items destined for China⁴. The rules restricted the sale of advanced tools to Chinese fabs making leading-edge logic, advanced DRAM and high-layer 3D NAND. They also restricted U.S. persons, including field service engineers, from supporting production at those facilities without a licence⁴.

For an equipment maker whose engineers live inside customers' fabs, that second provision was as disruptive as the first. Service is not optional in process control. Tools drift, recipes need tuning, and hardware needs maintenance. Withdrawing people is, in practice, withdrawing the tool.

The windfall

What happened next looks paradoxical. KLA's China revenue rose. Chinese chipmakers, locked out of the leading edge but facing the risk of still tighter rules, poured money into mature-node and specialty fabs for automotive, power and industrial chips, and bought equipment ahead of any further restrictions. By FY2024, China accounted for about 43% of KLA's revenue, roughly $4.2bn, on a ship-to basis[^5].

That is an extraordinary number for a company whose sales to China's most advanced fabs were constrained. It meant that for a period, the single largest source of KLA's revenue was the part of the market most exposed to geopolitics and most likely to prove temporary.

The normalisation

Since then, China's share has fallen: to about a third in FY2025 and about 30% in FY2026, or roughly $4.0bn[^2]. The absolute dollars barely moved. What changed is that everything else grew around them. Taiwan climbed to about 27% of revenue and Korea to about 14%[^2], carried by leading-edge logic for AI and by high-bandwidth memory.

Now to the first central question. Did margins suffer as China shrank from 43% to 30% of sales? No. Operating margin rose from about 37% in FY2024 to about 42% in FY2026[^2]. The mix shifted from mature-node Chinese tools toward leading-edge tools for Taiwan and Korea, which are more complex, more expensive, and carry more KLA intellectual property. The core process-control segment earned operating margins of about 45%[^2].

Myth versus reality

Myth: KLA's high margins were a China windfall that will disappear with China.

Reality: The margins expanded as China's share fell. The evidence rejects the strongest version of the bear claim. Margins were not propped up by China pricing.

But a narrower version survives. China still spends about $4bn a year with KLA, which is more than its entire revenue from Korea. That spending is mostly on mature-node capacity. If Chinese mature-node fab construction slows, or if the U.S. extends controls to trailing-edge tools or service, KLA would lose a large block of revenue that currently absorbs fixed costs. The company's own history shows what reduced absorption does: operating margin was about 30% as recently as FY2020[^2]. A fall in China revenue would not be fatal, but a sharp one would test operating leverage in a way the last two years have not.

Can China build its own KLA?

China's domestic equipment industry is growing, and inspection is a policy priority. Local suppliers have made progress with less demanding tools and older nodes. The hardest part of KLA's business, broadband optical and e-beam inspection at the leading edge, depends on deep-ultraviolet light sources, precision optics and decades of defect data. Those are much harder to replicate than a deposition chamber. KLA's own risk disclosures nonetheless treat competition from Chinese domestic suppliers, supported by state policy, as a real and rising threat[^2].

The calibrated verdict is this: substitution is a slow erosion at the trailing edge, not an imminent threat at the leading edge. The settling figure is China's share of revenue. If it falls below about 20% while group operating margin holds near 40%, the China question is largely answered. If margin slips as it falls, the bears were right in a way that matters.

With China flat and Taiwan booming, the next question moves from geography to cash. Record shipments create record receivables, and last year KLA sold more of them than ever before.

---

↑ Back to Top

VIII. The Cash Machine, Factoring Levers, and TSMC's Node Pacing (2024–2026)

The June quarter

The June 2026 quarter was the biggest in KLA's history: about $3.7bn of revenue and an operating margin of about 42%[^2]. On the earnings call at the end of July, management described demand driven by leading-edge logic, high-bandwidth memory and advanced packaging for AI[^6].

The cash flow statement tells a more complicated story than the income statement. Net profit for the year was about $4.8bn. Cash from operations was about $4.1bn[^2]. The gap of roughly $700m was working capital: about $1.4bn tied up in receivables, inventory and other assets as the business grew[^2].

The long view on cash

Start with the long record before the anomaly. Over twelve years KLA reported about $25.6bn of net profit and produced about $27.3bn of operating cash flow[^2]. Every dollar of reported profit arrived as more than a dollar of cash, over a full cycle and a large acquisition. That is the strongest evidence that the accounting profits are real.

The reason is the business model. Capital expenditure has run below 3% of revenue: about $376m in FY2026[^2]. KLA designs optics and software and assembles tools in modest cleanroom facilities. It does not operate giant factories. Almost all operating cash flow becomes free cash flow: about $3.8bn in FY2026[^2].

The receivables puzzle

Now the anomaly. Receivables as days of revenue rose from 73 in FY2025 to 86 in FY2026[^2]. That alone is the highest level since FY2019. The footnotes reveal something more. In FY2026 KLA sold about $516m of customer receivables and promissory notes to banks without recourse, up from about $231m a year earlier, plus about $86m of customer letters of credit[^2]. Once sold, those receivables leave the balance sheet. Put them back, and debtor days would have been about 100[^2]. KLA also recorded, for the first time, about $181m of receivables due after more than a year[^2].

In plain English: some customers are taking longer to pay, and KLA is converting part of that delay into cash by selling the invoices.

Is it a warning sign? The case for concern is real. Longer terms mean KLA is financing part of its customers' equipment purchases, and factoring flatters reported operating cash flow and receivables. The case against is also strong. KLA's allowance for bad debts fell slightly, to about $31m, roughly 1% of receivables[^2]. Its standard contracts still collect most of the price shortly after shipment, with the balance due after the customer accepts the tool[^2]. And its largest customers are some of the most creditworthy companies in the world.

The verdict is that this looks like commercial terms negotiated on very large orders, not credit deterioration. That judgement is not proven. Factoring rose while the company was returning more cash than ever to shareholders. A skeptic would say KLA is choosing to fund buybacks with its customers' invoices. The settling data are next year's factoring volume and the credit-loss allowance. If factoring keeps climbing and the allowance starts to rise, the benign reading breaks.

The TSMC question

TSMC accounted for about 19% of KLA's revenue in both FY2025 and FY2026, roughly $2.6bn last year, up from 13% in FY2024[^2]. No other customer exceeded 10%[^2].

The dependence has a specific rhythm. TSMC's equipment purchases peak when it builds capacity for a new node and slow while it digests that capacity before the next transition. KLA feels those pauses directly in orders. As TSMC moves from its 3nm process to 2nm, which uses a new transistor design called gate-all-around, and then to the A16 node with power delivered from the back of the wafer, both the size and the timing of its budgets will set KLA's cadence[^6].

The buffer is services. KLA earned about $3.1bn from servicing its installed base in FY2026, about 23% of revenue, growing steadily each year[^2]. When fabs stop buying new tools, they run their existing ones harder, and those tools still need maintenance, upgrades and support. Services do not eliminate cyclicality. They put a floor under it.

Where the cash goes

In FY2026 KLA paid about $1.1bn in dividends and bought back about $2.3bn of stock, returning roughly 89% of free cash flow[^2]. The board added $7bn to the buyback authorisation during the year, leaving about $9.7bn[^2]. Shares outstanding fell from about 135m two years earlier to about 131m[^2].

The balance sheet supports the programme. Debt consists of about $6bn of unsecured senior notes with long maturities and no amounts due within the year[^2]. Cash and investments are about $4.9bn[^2]. A new $1.5bn credit line is undrawn[^2]. Moody's moved its outlook to positive in 2026³. Leverage is low enough that the only real constraint on buybacks is the cycle itself.

The answer to the third central question is therefore nuanced. TSMC's pacing can swing a quarter or two. It is unlikely to break the capital-return programme, because services, cash on hand and an untouched credit line cover a normal digestion period. A prolonged TSMC pause coinciding with a China decline would be the scenario that forces a real reduction in buybacks.

All of this rests on the assumption that the moat is real. That deserves a full cross-examination.

---

↑ Back to Top

IX. Competitive Moat & Strategic Powers Analysis

Nine months for a lens

In KLA's optical research labs, some of the most important parts of a tool are not electronic at all. They are lenses and mirrors ground and polished to tolerances measured in fractions of a wavelength, lasers producing deep-ultraviolet light, and sensors that capture billions of pixels a second. Many come from single or limited suppliers, and lead times are long[^2]. KLA spent about $1.5bn on research and development in FY2026, roughly 11% of revenue, and grew total headcount to about 17,000[^2].

The question for this section is why, after thirty years of dominance and a period of very high profits, no well-funded rival has taken KLA's share. Applied Materials, a much larger company, has competed in inspection and e-beam review for decades. Hitachi High-Tech is strong in electron-beam measurement. Onto Innovation and others compete in niches. None has come close to KLA's position in broad optical inspection⁷.

Hamilton Helmer's 7 Powers

Switching costs. This is the strongest power. A fab does not simply buy an inspection tool. It builds recipes, the settings that tell the tool what to look for and what to ignore, over months or years of a node's ramp. It trains engineers on KLA software and connects defect data to the rest of its yield systems. Switching vendors mid-ramp risks missing a defect that costs far more than the tool. The evidence is behavioural: services revenue has grown every year, and management says most of the installed base is under service contracts[^2].

Scale economies. Process control is a research-intensive business with high fixed costs. KLA's $1.5bn research budget is spread across the industry's largest installed base. A competitor would have to spend similar sums on optics and algorithms while selling far fewer tools, which is a losing trade unless it can take share quickly. That explains why rivals attack niches rather than the core.

Cornered resource. KLA's data, patents and optical supply relationships are hard to replicate. The data point deserves care. Every inspection tool generates images of defects, and decades of classifying them is training material for the machine-learning systems that separate real defects from noise. This looks like a genuine advantage. Its size is hard to measure from outside, and KLA does not publish a figure for it.

Counter-positioning. KLA's independence from process tools is a structural advantage. Applied Materials and Lam sell the tools whose output KLA grades. Chipmakers trust the independent referee more, and the DOJ's handling of the Lam deal shows regulators see it the same way[^11].

Network effects in the strict sense are weak. More customers do not directly make the tool better for each customer. The data advantage described above is sometimes called a network effect, but it is better understood as a learning effect tied to scale.

So KLA clearly has three to four of the seven powers, not five. Brand and process power exist in weaker forms.

Porter's Five Forces

• New entrants: very low threat at the leading edge. The physics, optics and software barriers take decades to build. At the trailing edge, state-backed Chinese entrants are the exception.
• Buyer power: high on paper. TSMC is almost a fifth of revenue, and the top chipmakers are few, concentrated and sophisticated. In practice, their power is limited by the cost of a yield disaster relative to the price of a tool, and by the lack of an equivalent alternative. The fact that margins have held above 40% while TSMC's share rose supports that. KLA's pricing power is real. It is not infinite, and TSMC's negotiating leverage shows up in payment terms rather than prices.
• Supplier power: meaningful. Single-source optics and lasers can limit how many tools KLA can build[^2]. This is the most underappreciated risk to the moat, because it caps growth in boom years.
• Substitutes: low. Electron-beam inspection sees smaller features but is far slower than optical inspection. KLA sells both. The more plausible substitute is data: fabs using more electrical test data and software to infer defects instead of inspecting every layer. There is little evidence yet that it reduces inspection spending.
• Rivalry: concentrated and stable. KLA's peers are large and competent, but they compete at the edges of its franchise.

Testing the moat against the record

The strongest disconfirming evidence is Orbotech's display business, where KLA's brand and technology did not protect it from commoditisation. That shows the moat is not a general property of the company. It is specific to process control at the leading edge. It also shows the risk of stretching that moat into adjacent markets.

A second test is market share. If rivals were gaining, KLA's process-control revenue would lag total fab equipment spending during upcycles. Management says the opposite has happened across recent cycles[^7], and industry share data support KLA's continued lead⁷. The claim stands, with one caveat: the growing packaging and specialty markets are more competitive, with more Asian suppliers.

The fourth central question, whether packaging and EUV complexity will permanently expand KLA's share, therefore remains intact but unproven. The mechanism is plausible, and the recent evidence is supportive. The settling KPI is whether process-control revenue keeps outgrowing total fab equipment spending over the next two years.

The moat explains the margins. It does not explain the stock price, which is where the debate now moves.

---

↑ Back to Top

X. Bear vs. Bull Case

A 35% drawdown

On September 30, 2026, KLA's shares closed at about $197, roughly 35% below their 52-week high of about $302¹. The stock has been volatile, and its largest fall over five years was about 45%. That is a reminder that a franchise business can still carry a cyclical stock.

What the price appears to assume

The valuation data Empor uses are unusual and should be read with caution. On those data, KLA's market value is about $26bn, its trailing price-to-earnings ratio is about 5x, and its free cash flow yield is about 14%¹. The same data show a five-year median P/E of under 3x. These multiples are far below the levels typically quoted for large semiconductor equipment companies, including peers such as Applied Materials and Lam Research. The market value in the data set implies a price investors would normally associate with a business in structural decline, not one earning 40%-plus margins with rising revenue. Readers should check the market values against current quotes before drawing conclusions. The analytical question is the same either way: what would have to go wrong for today's cash generation to be unsustainable?

The bear case

1. China normalisation. About $4bn of revenue depends on Chinese mature-node spending and U.S. export policy. A sharp decline, combined with lower absorption of fixed costs, could pull operating margin back toward the mid-thirties seen in FY2023 and FY2024[^2].
2. Working capital stress. Rising debtor days, record factoring and new long-dated receivables suggest customers are extracting better terms. If that is the start of a trend, cash conversion falls and the buyback programme is partly financed by customers.
3. Concentration. TSMC is almost a fifth of revenue and Taiwan more than a quarter. A pause between nodes, an earthquake or a geopolitical crisis in the Taiwan Strait would hit KLA hard.
4. Supply constraints. Single-source optical and laser components could prevent KLA from converting its roughly $1.8bn of contract liabilities into revenue on schedule[^2].

The bull case

1. Complexity keeps rising. Gate-all-around transistors, backside power and high-NA EUV lithography all multiply the ways a chip can fail. More failure modes mean more inspection and measurement per wafer. This is management's central argument[^6], and it has been right across the past decade.
2. Packaging growth. AI hardware depends on advanced packaging and high-bandwidth memory. KLA now has products there, and the post-Orbotech clean-up has made that business profitable[^2].
3. The services annuity. About $3.1bn of services revenue, growing each year, provides a large, stable base that makes each downturn shallower than the one before[^2].
4. Shareholder yield. At the valuation in the data set, KLA's free cash flow and remaining buyback authorisation would allow it to retire a large share of its stock over a few years if the price stays low.

The activist's stress test

A skeptical activist would focus on three points. First, why is the company returning almost 90% of free cash flow while factoring receivables at record levels? Second, what is the plan to avoid another Orbotech-style write-down as it expands in packaging? Third, is the CEO's pay, already very large, too dependent on a share price that depends partly on buybacks? None of these points threatens the franchise. All of them are fair questions of governance and capital allocation.

The three KPIs that matter

• Process-control revenue growth against total fab equipment spending. This is the direct test of the moat and the complexity thesis. Recent readings are favourable, according to management[^6].
• China's share of revenue. Currently about 30% and falling[^2]. The question is where it stabilises and whether margins hold.
• Factoring volume and debtor days. Currently about $516m of receivables sold and 86 debtor days, both rising[^2]. A reversal would settle the cash-quality debate.

Those numbers are what investors should watch. The next section asks what founders and investors should learn.

---

↑ Back to Top

XI. Playbook: Business & Investing Lessons

The long game

KLA's history reads like a single strategy executed across fifty years and four generations of chips. Its revenue has compounded at about 16% a year for a decade, and its profits at about 21%[^2]. The lessons belong to this company specifically.

Lesson 1: Own the verdict, not the tool

Levy and Anderson did not try to beat the industry at printing. They built the machine that tells the printer whether it worked. The economics follow from that choice. The cost of a missed defect is so much larger than the price of the tool that finds it that customers argue about payment terms, not about whether to buy. Founders looking for durable pricing power should ask where the diagnostic bottleneck is in their industry.

"In a factory of a thousand machines, the richest one is the machine that grades the others."

Lesson 2: Mergers of equals work only when they fill a gap

The 1997 merger worked because KLA found defects and Tencor measured structures: two halves of one customer workflow. The Lam deal failed because it would have joined the referee to one of the teams. The contrast between those two deals is a better guide to M&A than any spreadsheet of cost synergies.

"A great merger completes a customer's workflow; a bad one merges the referee with a player."

Lesson 3: Put capital in the optics, not the cleanroom

KLA spends about four times as much on research as on capital equipment[^2]. It owns the hard intellectual property and keeps the physical footprint modest. The result is return on invested capital above 60%[^2]. The same model does come with a price: dependence on a few specialised suppliers for the most critical parts.

"Own the physics; rent the floor space."

Lesson 4: The installed base is the lifeboat

Every tool sold becomes a source of service revenue for a decade or more. When chip spending falls, fabs keep their existing tools running harder. Services at KLA have grown every year, even when tool sales fell[^2]. That turns a cyclical hardware maker into something closer to a subscription business at the bottom of each cycle.

"Sell the machine once; service the recipe for a generation."

Lesson 5: Admit the bad deal, eventually

The display business inside Orbotech destroyed about half a billion dollars of value. KLA took the write-downs and left the business rather than trying to justify it[^2]. That took too long, and it does not erase the mistake. It is still more than many acquirers ever do.

"Discipline is not never buying the wrong asset; it's refusing to keep paying for it."

---

↑ Back to Top

XII. Epilogue

Where KLA stands tonight

As of September 30, 2026, KLA sits at the centre of the AI hardware boom. It has record revenue, margins above 40%, a clean balance sheet and a shareholder-return programme on a very large scale[^2]. Its stock trades well below its recent high, and investors are divided on whether the cycle has peaked.

The next moments that decide it

TSMC's 2nm ramp. TSMC's move to 2nm and then A16 will show how much inspection and measurement the new transistor and power-delivery designs require. If inspection intensity per wafer rises as management expects, KLA's claim that process control outgrows fab spending gains another cycle of proof. If TSMC's budgets pause between nodes, KLA will feel it within a quarter or two[^6].

China's floor. The next few quarterly reports will show whether China revenue settles near its current $4bn level or falls further under tighter controls or domestic substitution. A decline to below a fifth of revenue with margins intact would answer the first central question in KLA's favour. A decline with falling margins would answer it the other way.

The factoring footnote. KLA's next annual report, for the year ending June 2027, will show whether factored receivables and debtor days come down as shipment timing normalises. That footnote is the simplest test of whether the cash is as good as the profits.

The tension that remains

The deepest open question is structural. The chip industry is splitting into two systems: leading-edge fabs in Taiwan, Korea, the United States and Japan building AI chips, and a Chinese system building mature-node chips under export controls. KLA dominates the first. It is steadily being limited in the second. Whether its margins and valuation can survive that split, with a smaller, more concentrated but more complex customer base, will define the next decade.

---

↑ Back to Top

XIII. Outro

In 1975, Ken Levy and Bob Anderson were not thinking about personal computers, smartphones or artificial intelligence. They were thinking about a technician at a microscope, tired and squinting, missing a flaw on a photomask that would ruin a production run. They understood something that has proved remarkably durable: whatever people build on silicon will contain mistakes, and someone will need to find them.

Fifty years later, chips from TSMC, Samsung, Intel, SK Hynix and Micron still pass through KLA's optics and electron beams many times before they reach a data centre or a phone. In an industry that celebrates the companies that print the future, much of the lasting profit has gone to the company that checks the print for errors.

KLA does not make the chips that change the world. It makes the eyes the world needs to make them.

References

1. KLA Corporation Investor Relations — KLA Corporation, 2026-09-30 ↩↩↩

2. Fitch Affirms KLA at 'A'; Outlook Stable — Fitch Ratings, 2023-12-12 ↩

3. Moody's Ratings Affirms KLA Senior Unsecured Debt at 'A2', Outlook Positive — Moody's Investors Service, 2026-06-15 ↩

4. Implementation of Additional Export Controls on Advanced Computing and Semiconductor Manufacturing Items — US Bureau of Industry and Security, 2022-10-07 ↩↩

5. Oral History of Kenneth Levy — Computer History Museum, 2002-11-20 ↩↩↩↩↩↩↩

6. KLA Corporation Officially Opens Second Headquarters in Ann Arbor — Michigan Economic Development Corporation, 2021-11-04 ↩

7. Semiconductor Process Control Market Share Analysis — Gartner, 2025-04-15 ↩↩↩

8. Management Team — KLA Corporation, 2026-09-30 ↩↩↩