Veeco Instruments Inc.: The Semicap Crucible

I. Introduction & Episode Roadmap (00:00 – 12:00)

Somewhere in a cleanroom in Japan, a square slab of ultra-low-expansion glass slides into a vacuum chamber. Over the next several hours, a beam of charged atoms will knock material off a target and lay it down on that glass, one ultrathin layer at a time: molybdenum, then silicon, then molybdenum again, dozens of times over. When the slab comes out it is a mirror. It does not look like much. But that mirror, once patterned into a photomask, is the stencil through which an ASML extreme ultraviolet scanner prints the most advanced chips on earth. One stray particle buried in those layers and the defect gets printed onto every wafer that passes under it.

The machine that deposited those layers was very probably built by Veeco Instruments Inc., an equipment maker headquartered in Plainview, on Long Island, New York, that most technology investors could not pick out of a lineup.¹ Veeco does not make chips. It does not make scanners. It makes the specialized process tools that sit at odd, critical junctions of the semiconductor supply chain: ion beam deposition systems for EUV mask blanks and hard disk drive heads, laser annealing tools that activate transistors at the leading edge, lithography and wet-processing tools for advanced packaging, and deposition reactors for compound semiconductors.¹

That description makes Veeco sound like a quiet niche specialist. Its history is anything but quiet. This is a company that began life in 1945 hunting for leaks in vacuum systems tied to the atomic age,² rode the Chinese LED boom to nearly $1 billion of revenue, then watched two-thirds of that revenue disappear in two years.² It responded by buying Ultratech for roughly $862 million at what turned out to be a cyclical high,³ and then posted a $407 million net loss in 2018 when the bet soured.⁴ It changed CEOs, rebuilt itself around semiconductors, and clawed its way back to a modern revenue peak of about $717 million in 2024.¹

And then, on September 30, 2025, it agreed to stop being independent. Veeco and Axcelis Technologies (NASDAQ: ACLS), the ion implantation specialist from Beverly, Massachusetts, announced a merger of equals that would leave Veeco shareholders with about 42% of a combined specialty equipment company.⁵ Shareholders of both companies approved it in February 2026 by margins that looked like a referendum in a one-party state.⁷ Yet more than a year after the announcement, the deal still has not closed. The last gate is not in Washington or Wilmington. It is in Beijing, at 国家市场监督管理总局 the State Administration for Market Regulation, known as SAMR.⁶⁸

So here is the central question this story tries to answer: what exactly is Veeco, once the cyclical noise is stripped away? Is it a holder of genuine chokepoints in the physics of chipmaking, worth a premium whether or not the merger closes? Or is it a sub-scale, customer-concentrated tool vendor that has been whipsawed by every boom it has ever ridden, and that is reaching for Axcelis because standalone life has become too hard?

The roadmap runs like this. First, a fast tour of eight decades of reinvention, from leak detectors to disk-drive heads. Then the Chinese MOCVD gold rush of 2009 to 2012, the most profitable and most dangerous period in the company's history. Then the Ultratech acquisition and the 2018 reckoning. Then the turnaround under William J. Miller, who has run the company since October 2018.⁹ Then the technology that actually matters today: EUV mask-blank deposition and laser spike annealing. Then the Axcelis merger and the regulatory wall in China. Finally, the lessons, the moat analysis, the bull and bear cases, and the questions that the next few quarters will settle.

One promise up front. Veeco has a habit of being right about the physics and wrong about the cycle. Keep that tension in mind; it explains almost everything that follows, starting with how a vacuum-leak shop survived long enough to become a chip-equipment company at all.

Subscribe on LinkedIn

↑ Back to Top

II. The Vacuum Tube to Silicon Genesis: Veeco's 80-Year Survival Reflex (12:00 – 25:00)

Helium, uranium, and a garage-scale startup

Picture Long Island in 1945. The war is ending, and a generation of engineers who had worked on the most secret industrial project in history is coming home with skills that have no obvious civilian market. Two of them, Frank Raible and Al Nerken, had worked around the vacuum technology of the Manhattan Project, where enormous enrichment systems had to hold vacuum with almost no tolerance for leaks. Their answer to the question "what now?" was the Vacuum Electronics Company, the name that later compressed into Veeco.²

The founding product was a helium leak detector. The idea is elegant. Spray helium around the outside of a sealed vacuum system; if any of it sneaks through a crack, a mass spectrometer inside detects the helium atoms and tells you exactly where the leak is. It is a tool for seeing the invisible. That phrase is worth holding onto, because Veeco has been selling variations on it for eighty years.

The conglomerate decades

What followed was the familiar path of a postwar American instrument company. Veeco expanded into high-vacuum components for industry and research, and, through the 1960s to the 1980s, diversified into power supplies and industrial electronics, including the Lambda Electronics business. The corporate structure changed more than once: Veeco was combined with other businesses, taken private, and later reassembled.² The details of that era matter less today than the lesson it left behind: being a collection of loosely related instrument businesses is not a strategy. It is a holding pattern.

The modern company came out of that holding pattern with its initial public offering on NASDAQ in December 1990.¹⁰ The new Veeco reorganized around precision instrumentation, surface metrology, and thin-film process equipment, the last of which would prove to be its true inheritance from the vacuum era.

The disk drive gold mine

The first great modern wave was magnetic storage. In the 1990s, hard disk drive makers moved to giant magnetoresistive read heads, microscopic sensors built from stacks of extremely thin magnetic films. Building those stacks reliably required depositing and etching material with atomic control. Veeco's ion beam etch and ion beam deposition tools became the manufacturing standard for those heads, and Seagate and Western Digital became the kind of customers that define an equipment company's identity.²

The disk-drive business taught Veeco two things that still shape it. First, ion beam technology is unusually good at laying down extremely smooth, extremely pure films. Second, a business tied to a handful of giant customers is a business that lives and dies by their capital-spending calendars. Data storage is still a Veeco segment today, and it was one of the businesses that softened in 2025.¹

Selling the microscope to bet on the furnace

The most consequential pruning came in 2010. Veeco had built a respectable business in atomic force microscopes and optical metrology, tools that measure surfaces at the nanometer scale. But metrology was a fight against larger, more focused players, and the company had an opportunity booming elsewhere. It sold the metrology unit to Bruker for about $229 million in cash.² The proceeds and management attention went toward deposition, and specifically toward a product that was about to become the hottest tool in the industry.

What eighty years actually proves

It is tempting to romanticize all this as a story of brilliant adaptation. The record supports a narrower verdict. Veeco did prove that a core of vacuum physics, ion optics, and thermal control can be redeployed across very different end markets: nuclear, industrial, data storage, lighting, and now advanced chips. Few instrument makers from 1945 still exist in any recognizable form, so survival itself is evidence of something.

But the same record shows the cost of that flexibility. Every pivot meant buying into a new end market's cycle, often near its peak, with customers who were few and powerful. The platform survived; the capital deployed in each era did not always survive with it. The next chapter is the clearest proof, and it starts with a Chinese government subsidy program and a reactor called TurboDisc.

↑ Back to Top

III. The LED Faustian Bargain: John Peeler and the Great Chinese MOCVD Boom & Bust (25:00 – 45:00)

The reactor everyone wanted

In 2010, the most coveted machine in a certain corner of the electronics world looked like an industrial pressure cooker. It was a metal organic chemical vapor deposition reactor, or MOCVD. Inside, wafers of sapphire spun on a heated platter while precursor gases flowed over them and decomposed, leaving behind perfect crystalline layers of gallium nitride. Those layers are what make a blue or white LED glow. Every flat-panel TV backlight and every LED bulb started its life inside a reactor like this.

Veeco's version was the TurboDisc line, and the company running Veeco when it took off was John R. Peeler, CEO from 2007 until October 2018.⁹ Peeler was a telecom-equipment veteran rather than a vacuum physicist, and his instinct was commercial: find the biggest wave and get on it.

The wave was China. Beijing had identified solid-state lighting as a strategic industry, and local governments competed to build LED champions. Subsidies for buying MOCVD tools were widespread and, by industry accounts, generous enough to cover a large share of a reactor's price. That made the customer's economics oddly insensitive to the tool price, and the order books of the two dominant suppliers, Veeco and Germany's Aixtron, exploded.

The numbers of a bonanza

The scale of what happened is easier to grasp as a story than as a ledger. In 2009, Veeco's operating income was about $11 million. Two years later, in 2011, it was about $276 million on revenue of roughly $979 million.² Operating income multiplied by about twenty-five times in two years. That is not the profile of a business winning share in a healthy market. It is the profile of a business standing in front of a firehose.

For a while, the duopoly framing made it all look durable. Two suppliers, high switching costs, deep process know-how: on paper, that is a moat. Investors priced it accordingly, and the stock traded above $50.

The firehose shuts off

Then the subsidies did what subsidies do. By 2012, Chinese LED capacity had overshot demand. Factory utilization fell, new orders froze, and the market for reactors went from shortage to glut almost overnight. Veeco's revenue fell from about $979 million in 2011 to about $332 million in 2013, a drop of roughly two-thirds, and the company swung to an operating loss of about $72 million that year.² It was not a one-year hangover. The operating losses persisted for years, reaching about $120 million in 2016.¹

Myth vs reality: the MOCVD moat

The myth was that Veeco and Aixtron held a durable duopoly. The reality was that the duopoly lasted only as long as the customers were willing to buy Western tools. The policy that created the demand also aimed to build domestic supply. 中微公司 Advanced Micro-Fabrication Equipment Inc. (AMEC), based in Shanghai, developed competitive MOCVD reactors and won significant share in Chinese LED capacity, with the commodity end of the market moving toward domestic suppliers and prices coming under heavy pressure.

That is the historical falsification at the heart of this section. Process know-how and installed base, the classic switching-cost arguments, did not protect Veeco when the buyer was a policy-driven industry whose stated objective was self-sufficiency. When the product is a commodity output like LED chips, the customer can afford to qualify a cheaper, local tool. The moat narrowed to almost nothing in the segment where the volume was.

What the boom actually left behind

There are two ways to score this era. In cash terms, Veeco did harvest enormous profits in 2010 and 2011, and the boom funded a strong balance sheet. In strategic terms, it trained management and investors to believe that the company's destiny lay in a single, high-volume, policy-driven market, and it left the company with a cost structure and an identity built for a market that no longer wanted it.

For investors today, the MOCVD era matters for a very specific reason: China still represented about 27% of Veeco's sales in 2025, and more than a third in 2024.¹¹¹ The company has not left China. It has shifted the mix toward mature-node semiconductor and compound-semiconductor buyers, but the lesson of 2012 is that Chinese demand can be both enormous and fleeting, and that Chinese industrial policy is designed, eventually, to replace the vendor.

By 2016, Peeler faced a stark problem: a shrinking LED franchise, a strong but idle balance sheet, and a stock price far below its peak. His answer would be the biggest bet in Veeco's history.

↑ Back to Top

IV. The $862 Million Ultratech Scramble: Buying Annealing and Writing Off Half the Company (45:00 – 1:05:00)

The deal

In early 2017, Veeco agreed to buy Ultratech, Inc., a San Jose equipment maker founded on lithography and best known in its later years for laser annealing. The deal, paid in a mix of cash and stock, closed in May 2017 at a value of roughly $862 million.³ To fund it, Veeco issued $345 million of 2.70% convertible notes due 2023 in January 2017.¹

What did Veeco buy? Two franchises. The first was Laser Spike Annealing, or LSA. When chipmakers implant dopant atoms into silicon to create transistors, the crystal lattice gets damaged and the dopants are not yet electrically active. Heating fixes both, but heating too long lets the dopants wander, blurring the tiny structures. LSA solves that by sweeping a laser across the wafer surface, heating it to over 1,000 degrees Celsius for a fraction of a second, so the top layer gets the heat it needs while the bulk stays cool. Think of searing a steak with a blowtorch so the outside is crusted while the inside barely warms.

The second franchise was advanced packaging lithography: steppers used to pattern bumps and redistribution layers that connect chips to their packages, a market tied in 2017 largely to smartphone application processors.

The logic, and the trap

Strategically, the logic was sound. Veeco needed exposure to front-end semiconductor manufacturing, the deepest profit pool in equipment, and LSA was a real technology with real customers at leading-edge logic fabs. The trouble was price and timing. Semiconductor equipment was near a cyclical high in 2017. Smartphone unit growth was flattening. And Veeco was paying a full multiple for a business whose packaging half was tied to that slowing phone market.

The convertible notes added a second problem. Convertibles look cheap because the coupon is low, but they are only cheap if the stock rises enough for holders to convert. If it does not, the company owes the principal back. Veeco had taken a business with lumpy, sometimes negative operating cash flow and attached a $345 million maturity to it.

The 2018 reckoning

In 2018, the cycle turned. Equipment demand cooled, LED orders kept weakening, and the acquired businesses underperformed the plan. Veeco wrote down goodwill and intangible assets and reported a net loss of about $407 million for the year.⁴ On a company whose revenue that year was about $542 million, that loss was a brutal statement: the market value of what Veeco had bought was far below the price it paid, and a large share of shareholders' book equity went with it.

In October 2018, the board replaced Peeler with William J. Miller, then the company's president, a Veeco insider who had spent many years running operations and engineering teams.⁹ The handover was quiet by Wall Street standards. Its meaning was not. The board was ending the era of big bets and choosing someone who knew the tools rather than the deal market.

The forensic verdict

Was Ultratech a mistake? The honest answer has two halves. In financial terms, yes: the write-downs say Veeco overpaid, and shareholders absorbed the loss through impairment and, later, through dilution as convertibles were refinanced and settled in stock. In strategic terms, the answer is more complicated. LSA is today one of the two technologies that define Veeco's front-end semiconductor story, and its expansion into gate-all-around logic and advanced DRAM is the company's best growth vector. The deal bought the right asset at the wrong price.

That distinction matters for how investors judge the current merger. Veeco's management has a documented history of making large strategic moves near cycle peaks. The Axcelis combination is structured very differently, as an all-stock merger of equals with no new debt, but the record is a reason to ask hard questions about synergy claims rather than accept them.

Miller inherited a damaged balance sheet, a fading LED franchise, and a portfolio of promising but sub-scale semiconductor tools. What he did with it is the next chapter.

↑ Back to Top

V. The Bill Miller Pivot: Stripping the Fat to Build a Specialized Semicap Machine (1:05:00 – 1:25:00)

An engineer's mandate

Miller's style, as it shows up in Veeco's filings and calls, is the opposite of Peeler's. He speaks in process steps and customer qualifications rather than in total addressable markets. His mandate was simple to state and painful to execute: stop funding commodity businesses, concentrate R&D on process steps where Veeco could be one of the top suppliers, and pull the company toward front-end semiconductor manufacturing.⁹

The first moves were subtractions. Veeco de-emphasized and wound down commodity LED MOCVD work, the business that had defined it a decade earlier, and refocused the compound-semiconductor effort on more specialized applications.⁹ Facilities and engineering consolidated around Plainview and San Jose. Instead of shrinking, headcount grew, from about 990 in 2020 to about 1,265 at the end of 2025, with a quarter of employees in R&D.⁹¹ That is the signature of a company reinvesting in engineering, not one cutting to the bone.

How Veeco makes money

The revenue model is classic capital equipment. Most of Veeco's sales, roughly three-quarters to four-fifths, are system sales: individual tools, often customized, priced from several million to tens of millions of dollars, with revenue tied to shipment and customer acceptance milestones.¹ The rest is aftermarket: spare parts, upgrades, and service contracts. In 2025, parts and upgrades were about 19% of sales and service about 6%.¹

The aftermarket matters more than its share suggests. Every tool Veeco ships becomes a long-lived installed base that needs parts and upgrades, and that revenue does not depend on the customer's next capacity decision. It is the closest thing Veeco has to a recurring business, and it rose as a share of sales in 2025 even as system revenue softened.¹

Capital-light, R&D-heavy

Veeco's other structural choice is to keep manufacturing light. Capital spending has generally run around 2.5% of sales, with a temporary bump to about 4% in 2023 for San Jose cleanroom expansion.¹²¹ Research and development, by contrast, runs at roughly 17% to 18% of revenue, about $120 million in 2025.¹ The company assembles and tests its tools in its own cleanrooms but buys many subassemblies from suppliers. That model protects cash in downturns, but it also creates a dependence on single- and limited-source suppliers for lasers, optics, and vacuum components, a risk Veeco itself flags.¹

The comeback numbers

The recovery is real, if not spectacular. Revenue grew from about $419 million in 2019 to about $717 million in 2024, an annual growth rate of roughly 11% over five years.⁹¹ Gross margin rose from about 37% in 2015 to about 42% in 2024,²¹ a meaningful improvement that reflects the shift toward semiconductor tools with more pricing power. Operating income reached about $67 million in 2024.¹

Then it slipped. Revenue fell about 7% in 2025, to roughly $664 million, as data storage and compound-semiconductor demand softened, and gross margin fell to about 40%.¹ Operating income nearly halved, to about $36 million.¹ In the first half of 2026, revenue grew again, by about 5.5%, but gross margin fell to about 37%, back to where it was in 2015 before the turnaround began.⁶

That last fact deserves a direct verdict. The claim that Miller structurally lifted Veeco's margins is not rejected by the record, but it is narrowed. The improvement held for several years and peaked in 2024; it has since reversed most of the way, driven by mix, tariffs, and component costs that management describes as headwinds.⁶ Whether 42% was a new floor or a cyclical high is one of the central questions in this story, and the 2026 annual results will settle a large part of it.

Cleaning up the balance sheet

The quieter achievement was financial. Veeco worked steadily through the debt it inherited from the Ultratech era. It issued new convertibles in 2020 and, in May 2023, raised $230 million of 2.875% convertible notes due 2029, using most of the proceeds to retire older notes.¹ That refinancing came with a price: a $97 million non-cash loss on debt extinguishment in 2023, which helped push that year's net result to a loss.¹² The last of the 2025 notes were settled in January 2025 with about 1.1 million shares, and the 2027 notes were cleared in May 2025 through an induced conversion for about 1.6 million shares and some cash.¹

The trade-off is plain. Veeco removed near-term maturities and now holds more cash than debt, about $429 million of cash and short-term investments against $230 million of notes at June 30, 2026.⁶ But it paid partly in stock: diluted shares rose from about 54 million in 2022 to about 61 million by mid-2026.¹⁶ And share repurchases have been limited to withholding shares for employee taxes.¹ The balance sheet is clean; existing shareholders funded part of the cleaning.

The turnaround left Veeco smaller in ambition than in the LED days, but sharper. To understand whether that sharpness is a moat, it is worth stepping inside the two technologies that define it.

↑ Back to Top

VI. The Hidden Monopoly: EUV Mask Blanks, Laser Spike Annealing, and the AI/HBM Supercycle (1:25:00 – 1:45:00)

The mirror that must be perfect

Go back to the slab of glass from the opening. EUV light, at 13.5 nanometers, is absorbed by almost everything, including glass and air. That is why EUV lithography uses mirrors rather than lenses, and why the photomask itself is a mirror. The reflective coating on that mask is a stack of dozens of alternating molybdenum and silicon layers, each a few nanometers thick, tuned so that reflections from every interface add up.¹⁵

The demand on that coating is merciless. A single buried defect becomes a printing defect. Ion beam deposition is the method used for this job because it produces extremely smooth, dense films with very low particle contamination: the ion source sits apart from the substrate, so the glass is not bathed in plasma the way it would be in conventional sputtering. Veeco built this capability in the disk-drive era and carried it into EUV.¹

Mask-blank makers such as Hoya and AGC are the main buyers, and Veeco is widely regarded as the leading, and by many estimates near-sole, supplier of the ion beam deposition tools they use. Veeco itself does not publish a market-share figure for this product, so the commonly cited "more than 90%" should be treated as an industry estimate rather than a disclosed number. As High-NA EUV arrives, with tighter defect budgets and new mask materials, the case for continued investment in these tools strengthens.

The limitation is size. This is a tiny market. The world needs only a modest number of mask-blank deposition tools, and purchases come in lumpy waves. A chokepoint that sells a handful of systems a year is strategically important, and it supports pricing and service revenue, but it cannot carry a company with $700 million of revenue by itself.

The laser that activates the transistor

Laser Spike Annealing is the larger growth story. As logic moves from FinFET transistors to gate-all-around designs at the 2-nanometer generation, transistors become three-dimensional stacks of nanosheets where dopant placement must be controlled with extreme precision. Fast, surface-only heating becomes more valuable as the structures shrink. Veeco says LSA has won positions at leading logic customers and, more recently, has expanded into advanced DRAM, with tool wins at Tier-1 DRAM makers in 2024 and 2025.¹¹⁶

That second leg matters because memory is where AI demand is currently most intense. High-bandwidth memory stacks DRAM dies vertically, and the DRAM itself is moving to more complex peripheral circuitry where thermal budgets are tight. If LSA becomes process-of-record across more memory makers, it widens the market well beyond logic.

Here the customer concentration story arrives. Veeco's largest customer, labelled Customer A, accounted for about 17% of 2025 sales and nearly a quarter of receivables, up from 11% of sales a year earlier.¹ Veeco does not name the customer. Taiwan revenue jumped from about $115 million in 2024 to about $179 million in 2025,¹ which makes the identity of a leading-edge Taiwanese foundry a reasonable inference, but it remains an inference. Either way, one account's spending calendar now moves Veeco's quarterly numbers.

Packaging and wet processing

The third and fourth legs are advanced packaging lithography and single-wafer wet processing. Both benefit from AI accelerators, whose packages are larger and denser than smartphone chips and require more redistribution layers and more cleaning steps. Veeco reports continuing order activity here,¹⁷ but these are competitive markets with several capable rivals, and they carry less pricing power than the IBD and LSA franchises.

Historical falsification: Epiluvac

The cleanest test of how Veeco converts technology into revenue is a recent one. In 2023, Veeco bought Epiluvac AB, a Swedish developer of silicon carbide epitaxy reactors, to enter the power-semiconductor equipment market tied to electric vehicles.¹² By the end of 2024, electric vehicle demand had cooled, silicon carbide capacity plans were shelved, and Veeco took a roughly $28 million impairment while reversing about $21 million of earnout liabilities that would have been paid had the business hit its targets.¹¹

The earnout reversal softened the hit to profit, which is exactly what earnouts are designed to do. But the episode is a reminder that a working technology and a qualified customer base are two very different things, and that Veeco's timing on adjacent markets has repeatedly been early or late.

The competitive threat

The final limitation is competition. Applied Materials sells thermal processing tools alongside deposition and etch, and its scale lets it bundle offerings and fund roadmaps at a level Veeco cannot match. SCREEN Holdings, the Japanese equipment maker, competes in annealing and wet processing.¹ Veeco's defense is that its tools are qualified into specific process steps where requalification is expensive. That is real, but it is a defense measured one node transition at a time.

The verdict on this section: Veeco owns at least one genuine chokepoint, in EUV mask-blank deposition, and a growing but contestable position in laser annealing. Neither, on its own, solves the problem of scale. That problem is what sent Bill Miller to Beverly, Massachusetts.

↑ Back to Top

VII. The $4.5 Billion Escape Hatch: The Axcelis Merger and Beijing's Regulatory Wall (1:45:00 – 2:05:00)

The announcement

On September 30, 2025, Veeco and Axcelis Technologies announced an all-stock merger of equals.⁵ Each Veeco share would become 0.3575 Axcelis shares, leaving Veeco holders with about 41.6% of the combined company and Axcelis holders with the rest.¹³¹ The combined business was pitched as a specialty semiconductor equipment company with roughly $1.7 billion of revenue.⁵

Axcelis is a different animal from Veeco. It is the leading independent maker of ion implanters, the machines that fire dopant atoms into silicon. Its Purion product line found an enormous market in the 2020s among silicon carbide power-chip makers and mature-node fabs, many of them in China.

The logic: implant, then anneal

The technical story is genuinely neat. Implantation and annealing are sequential steps. Axcelis drives dopant atoms into the silicon lattice and damages it in the process; annealing activates the dopants and heals the damage. A combined company can engineer the two steps together and sell them to the same process engineers.⁵ There is very little product overlap, which the companies emphasized.⁵

The financial story is about scale. The largest equipment makers, Applied Materials, ASML, Lam Research, Tokyo Electron and KLA, spend more on R&D in a year than Veeco earns in revenue. A mid-sized company faces rising compliance costs from export controls, customers consolidating their vendor lists, and R&D programs that get more expensive with every node. The merger is, in part, an admission that sub-$1 billion independents struggle to keep pace.

Shareholders say yes

On February 6, 2026, both companies' shareholders voted. Veeco's tally was about 53.4 million shares for and about 0.47 million against.⁷ Governance was handled cleanly in at least one visible respect: Keith D. Jackson, who sat on both boards, recused himself from Veeco's deliberations.¹⁴ The proxy also shows how management alignment works here: CEO Miller owned about 583,000 shares, under 1% of the company, and the five largest institutional holders, led by BlackRock and Vanguard, controlled almost half of the shares.¹⁴

The wall

Then the deal stalled at the last gate. The merger requires clearance from China's SAMR, and by the second quarter of 2026 that was the remaining regulatory condition.⁶ This is not a classic antitrust problem; the two companies barely overlap. It is a geopolitical one. Both companies sell a lot into China: Veeco about 27% of its 2025 sales, roughly $182 million, down from about 36% in 2024,¹ and Axcelis even more heavily. Both are American suppliers of technology that Washington restricts and Beijing wants to replicate.

The precedents are not encouraging. Applied Materials abandoned its planned acquisition of Kokusai Electric in 2021 after Chinese approval failed to arrive in time, and Intel walked away from its purchase of Tower Semiconductor in 2023 for the same reason. In both cases, SAMR did not have to say no. It only had to not say yes.

The merger agreement set an outside date of September 30, 2026, extendable to June 30, 2027.⁶ That original date has now passed without a termination announcement, which places the transaction in its extension period. If the deal breaks under certain conditions, Veeco could owe a $77.5 million termination fee, such as after a change in its board's recommendation, while Axcelis faces a larger fee of about $108.7 million in specified circumstances.¹¹³ The more likely scenario, if SAMR runs out the clock, is simply that both companies absorb their transaction costs and go back to being independent. Veeco has already spent meaningfully on deal expenses, which helped depress 2026 operating income.⁶

The verdict on the deal

Is the merger a masterstroke or a concession? On the evidence, it is mostly the latter, dressed in sensible strategic logic. The product fit is real. But the deeper motive is scale, and scale is the thing Veeco has repeatedly lacked when cycles turn. The problem is that the decision now rests with a regulator whose incentives are tied to trade policy rather than market structure. Veeco built tools that China needs, and China now holds the key to Veeco's corporate future.

What does the full story teach? Before the moat analysis and the valuation, it is worth distilling what eight decades of Veeco say to founders and investors.

↑ Back to Top

VIII. Playbook: Business & Investing Lessons (2:05:00 – 2:20:00)

Lesson 1: A subsidy is a customer who plans to leave

In 2011, Veeco's reactors were selling into a market where local governments helped pay for them. The profits were real, and they were gone within two years. Then the same policy that financed the orders financed a domestic competitor, AMEC, which went on to win much of the commodity LED tool market. The wider lesson is not "avoid China." It is that demand created by industrial policy carries an expiry date written by someone else, and that the more strategic the product is to the buying country, the faster that country will try to make it at home.

"When a government pays your customer's bill, you are not selling a tool. You are renting a policy that was written to replace you."

Lesson 2: The right asset at the wrong price is still the wrong deal

Ultratech brought Veeco the laser annealing technology that today anchors its growth story. It also brought a $407 million net loss in 2018 and a balance sheet that took six years and a lot of stock to repair. Both statements are true at once. For investors, the lesson is to separate the strategic merit of an acquisition from its return on capital: shareholders who held through 2018 paid for an asset that management only later learned how to use.

"In semicap, you buy at the top with borrowed money and pay for it at the bottom with your own equity."

Lesson 3: Find the step where physics allows no substitute

The strongest part of Veeco's business today is not its biggest. It is the narrow job of depositing near-perfect mirrors on EUV mask blanks, where the demands of the physics leave few practical alternatives. That position survived the LED bust, the Ultratech write-down and the 2025 slowdown. The general lesson for founders is that a tiny market with no substitute is worth more, per dollar, than a large market with many.

"The best niche in technology is the one ASML cannot shine light through without you."

Lesson 4: Technology is not a customer

Epiluvac worked. Its reactors deposited silicon carbide. But by the time Veeco owned it, the electric-vehicle buildout that would have needed those reactors had stalled, and Veeco booked a $28 million impairment. The earnout cushioned the blow, which is precisely why earnouts are wise. But the lesson stands: a technical capability without qualified customer pull is an option, not a business.

"A tuck-in without a customer pulling on it is just a slow-motion impairment."

Lesson 5: Own the recipe, rent the factory

Veeco keeps capital spending around 2.5% of sales, spends about seven times that on R&D, and ended June 2026 with more cash than debt.⁶ That structure is why the company survived the post-2011 collapse and the 2018 write-downs without a liquidity crisis. The price of this model is dependence on single-source suppliers and limited scale economies. But in an industry where demand can fall by two-thirds in two years, the right to survive the downcycle is the most valuable option of all.

"Own the patents and the cleanroom, outsource the metal, and always keep more cash than your convertibles."

↑ Back to Top

IX. Analysis & Bear vs. Bull Case (2:20:00 – 2:40:00)

The question in the room

Imagine the scene at an investor conference this fall. An analyst asks the obvious question: why did gross margin fall from about 42% in 2024 to about 37% in the first half of 2026,¹⁶ while inventory climbed to a record of about $293 million?⁶ Management's answer on recent calls points to product mix, tariffs, and component costs, along with inventory staged for future shipments backed by customer deposits.¹⁷ That answer is plausible. It is also exactly the answer an equipment company gives at the start of a downturn. The data will have to decide.

Valuation as evidence

At around $50 a share, Veeco is worth about $3.1 billion in equity, and about $2.9 billion after subtracting net cash.⁶ Against trailing revenue of roughly $683 million, that is an enterprise value of about 4.2 times sales.¹⁶ Trailing operating income, depressed by deal expenses and tariffs, is tiny, so the price-to-earnings ratio on trailing results is meaningless.

The more useful anchor is the merger itself. Because Veeco shareholders receive a fixed 0.3575 Axcelis shares, Veeco's price largely tracks Axcelis's price multiplied by that ratio, minus a discount reflecting the risk that the deal fails. Put differently, the market is not valuing standalone Veeco at 4.2 times sales. It is valuing a probability-weighted claim on the combined company. If the deal breaks, the relevant comparison shifts to Veeco's own history, where the stock has often traded closer to 2.5 to 3 times sales, and the gap between those multiples is a rough measure of what the market is paying for the merger.

Hamilton Helmer's 7 Powers

Cornered Resource (strong, but narrow). The ion beam deposition know-how behind EUV mask blanks is the most defensible asset Veeco owns. It is built on decades of ion source design and process data, and the customer base is small and conservative. Its limit is the size of the market it protects.

Process Power (moderate). LSA rests on years of temperature-control and laser-scanning recipes calibrated in customer fabs. That is real, accumulated knowledge. But it is renewed one node at a time, and each node is an opportunity for a competitor to bid.

Switching Costs (moderate to high). Once a tool is qualified as a customer's process of record, replacing it means requalification, yield risk and delay. The MOCVD era, however, proved that switching costs collapse when the buyer is policy-driven and the output is a commodity.

Scale Economies (weak). Applied Materials and Lam Research operate at revenue scales roughly twenty to forty times Veeco's. Veeco cannot match their R&D budgets, purchasing power or service networks. The Axcelis merger is the response.

Counter-Positioning, Network Economies, Brand (absent or negligible). Veeco does not have a business model that incumbents cannot copy without self-harm; it operates in niches that larger players have chosen not to prioritize, or competes head-on where they have.

Porter's Five Forces

Buyer power is extremely high. The top ten customers made up about 68% of receivables at the end of 2025, and the largest single customer about 17% of sales.¹ Some Asian customers, notably in China, receive payment terms of up to 150 days.¹ These are customers who can delay orders, split purchases, and demand concessions.

Supplier power is moderate to high. Veeco relies on single- or limited-source suppliers for lasers, optics and vacuum hardware, and had about $151 million of purchase commitments at the end of 2025.¹

The threat of substitutes is moderate. Rapid thermal processing and flash lamp annealing compete in mature nodes. At the leading edge, laser annealing's case is stronger, but not untouchable.

The threat of new entrants is low in the West, given the physics, qualifications and patents involved. In China, it is higher, because the state funds entrants and buyers want them.

Rivalry is high, especially against Applied Materials and SCREEN in thermal processing and against several specialists in packaging and wet processing.

The money trail: three clues

Profit into cash. From 2023 to 2025, Veeco's cumulative operating cash flow was about $195 million against net income of about $79 million.¹ The gap comes mostly from non-cash charges: share-based compensation of about $100 million over three years, the 2023 debt-extinguishment loss, and the Epiluvac impairment, partly offset by inventory build.¹ The share-based compensation line is the one an activist would circle. At roughly $37 million a year, it consumes much of what the company earns in a weak year, and it is a reason share count keeps creeping up.

Receivables. Credit quality is not a concern on the evidence: the bad-debt allowance is about $1 million and write-offs have been negligible.¹ The issue is concentration and timing, not collectability.

Interest income. Because cash exceeds debt, Veeco earns more in interest than it pays. In 2025, interest income was about a third of pre-tax income.¹ That is a cushion, but it also means the operating business earned less than the headline profit suggests.

The bull case

The merger closes. SAMR clears it, perhaps with commitments to keep supplying Chinese customers. The combined company sells integrated implant-and-anneal solutions, captures cost synergies, and earns a valuation closer to specialized peers rather than standalone history.

High-NA EUV drives a new round of mask-blank deposition tool purchases as defect budgets tighten. LSA wins more memory makers as HBM and advanced DRAM ramp. And the inventory build turns into shipments: customer deposits jumped by about $61 million in the first half of 2026, to about $124 million,⁶ a sign that customers are paying cash for tools still being built. If those tools ship in the second half of 2026, inventory falls, margins recover and free cash flow surges.

The bear case

Beijing lets the clock run out, as it did with Applied-Kokusai and Intel-Tower. Veeco is left with deal costs, management distraction, and no path to scale.

Margins prove structural at 37% to 39%, as tariffs, component inflation and a less favorable mix persist. Customer A delays or dual-sources, and a single account's spending pause drives a violent revenue drop. And the inventory build becomes a write-down: KPMG has made the valuation of excess and obsolete inventory its sole critical audit matter,¹ a signal of how much judgment sits in that balance sheet line. In a mature-node China slowdown, those judgments can move against the company.

The activist's stress test

A skeptical activist would ask three questions. Why, after six years of a focused strategy, are margins back to 2015 levels? Why does stock compensation consume so much of profit while buybacks are absent? And why did management pursue Epiluvac, a cyclical adjacency, after the board had specifically chosen Miller to end the era of ill-timed acquisitions? These are fair questions, and the 2025 annual bonus outcome offers a partial answer: executives received no annual cash incentives because financial targets were missed.¹⁴ Pay, at least, moved with performance.

The three KPIs that matter

Gross margin. It peaked at about 42% in 2024, fell to about 40% in 2025, and stood at about 37% in the first half of 2026.¹⁶ The direction is down. A full-year 2026 figure back above 41% would support the claim that the turnaround's gains were structural.

Inventory against customer deposits. Inventory rose from about $238 million at the end of 2023 to about $293 million in mid-2026, while customer deposits rose to about $124 million.¹²⁶ Both are rising. The test is whether inventory falls as deposits convert to revenue.

Customer A's share of sales. It rose from 11% in 2024 to 17% in 2025.¹ The direction is toward more concentration. Above 20% would signal dependency; back under 15% would signal broadening.

↑ Back to Top

X. Epilogue (2:40:00 – 2:50:00)

Tonight, Veeco is a company in suspended animation. It holds more cash than debt, about $203 million net.⁶ It employs about 1,265 people.¹ It builds tools that sit at chokepoints in the most advanced chip manufacturing on earth. And its share price is pinned to a fixed exchange ratio with another company, waiting for a regulator on the other side of the world.

Three moments will decide the story.

The first is the decision from SAMR. There are three plausible outcomes. Clearance with conditions, perhaps commitments to keep supplying Chinese fabs, would let the merger close and turn Veeco into the annealing and deposition arm of a larger specialist. Unconditional clearance would be the best case and, given the geopolitical climate, the least likely. Silence until the June 30, 2027 extension date expires would be the worst case, and the most consistent with the precedents.

The second is the third and fourth quarter results for 2026. Customer deposits have piled up and inventory is at a record. If tools ship and margins recover, the inventory build will look like a runway. If shipments slip and margins stay near 37%, it will look like the beginning of a cyclical trap, and the critical audit matter on inventory will start to read like a warning.

The third is the standalone test. If the merger fails, can Bill Miller defend Veeco's valuation on its own? The answer depends almost entirely on the first two outcomes. A company with recovering margins and converting inventory could live with the multiple of a specialist. A company with neither would likely drift back toward its historical range.

Behind all three sits the tension that has defined Veeco for fifteen years. The company is good at the physics. It has repeatedly been on the wrong side of forces it does not control: subsidy cycles, smartphone cycles, electric vehicle cycles, and now the industrial rivalry between Washington and Beijing. In the modern semiconductor cold war, building the best atomic-scale tools in the world is necessary. It is no longer sufficient.

↑ Back to Top

XI. Outro (2:50:00 – 2:55:00)

Go back once more to Long Island in 1945. Two engineers fresh from the Manhattan Project are spraying helium around a vacuum chamber and watching a needle twitch, finding leaks too small to see. The tool they built was a way of making the invisible visible, and of controlling what happens inside an empty space.

Eighty years later, the empty space is smaller and the stakes are larger. Veeco's machines still live in vacuum, still steer beams of charged atoms, still use heat with exquisite precision. The company outlived the decline of the disk drive, the collapse of the Chinese LED boom, and the most expensive acquisition in its history. It learned to build mirrors for light that glass cannot bend and lasers that heat a wafer's skin without warming its bones.

What it never learned to control is the world outside the chamber. Veeco is the vacuum survivor that mastered the nanometer and still answers to geopolitics: a company whose future, after eight decades of physics, will be decided by a regulator's silence.