Advantest（6857） · AI Semiconductor Equipment

Advantest Deep Value Research Report

Jun 4, 2026·~22 min read

Quick ReadPlain-language overview · read this first

Advantest is the global leader in semiconductor automated test equipment (ATE). Its core business is providing chipmakers with test platforms, test peripherals such as handlers/probers, and related services. Its main battlegrounds are high-end SoC testers and high-performance memory testing; it is not a “chip seller.” The business benefits from rising chip complexity and the expansion of AI/HPC/HBM, but it is inherently highly cyclical, with short-term revenue swinging sharply with customer capital expenditure.

The core view is a good company at a bad price: the rating is Watch. The company’s quality is genuinely excellent, but the current share price has already paid upfront, in one lump sum, for almost flawless execution over the next ten years. The market’s most fragile assumptions are that AI testing will stay highly prosperous for a long time and that the high 44.2% profit margin can be sustained over the long term.

Supporting facts: its SoC tester share is estimated at about 66%, its FY2025 operating margin was 44.2%, and it had net cash of about 319.8 billion yen, giving it strong capacity to get through cycles. But at 27,660 yen, its trailing P/E is about 50 times and its conservative owner-earnings multiple is about 67 times, already above the upper end of the optimistic DCF range. A buying opportunity would come after a significant pullback, margin normalization, or valuation compression; the margin of safety is insufficient.

Lead
Advantest is a leading semiconductor test company, with about 66% share in SoC testers, a 44.2% FY2025 operating margin, and about JPY 319.8 billion in net cash, making it a high-quality business. Yet at JPY 27,660, the stock trades at roughly 50x P/E and about 67x conservative owner earnings, already above the upper end of the optimistic DCF range, leaving insufficient margin of safety. Research rating Watch: the core risk is that strong AI demand and elevated margins normalize while the valuation multiple compresses.

Full report

Bottom Line First

Investment rating: Watch.

If Advantest is viewed as a business one might want to "acquire and hold" for the long term, it is clearly not a poor business. The company sits in semiconductor test, a segment with high technical barriers that benefits over the long run from rising chip complexity. In FY2025, ended March 2026, revenue reached JPY 1.1286 trillion, operating profit was JPY 499.1 billion, and net profit was JPY 375.4 billion, all record highs. For FY2026, management further guides for revenue of JPY 1.42 trillion, operating profit of JPY 627.5 billion, and net profit of JPY 465.5 billion. At the same time, the balance sheet is very strong: as of the end of March 2026, cash and cash equivalents were JPY 340.0 billion, total debt was only about JPY 20.19 billion, the equity ratio was 67.9%, and the R&I long-term issuer rating was A+.

But a "good company" and a "good price" are two different questions. Using the June 3, 2026 Japan close of JPY 27,660 as the near-term reference, the market's valuation is already very high: Yahoo Finance shows a trailing P/E of about 50 to 51x and EV/EBITDA of about 34.4x. Based on the company's FY2025 free cash flow of JPY 300.6 billion and roughly 732 million shares outstanding, the current free-cash-flow/owner-earnings yield is only around 1.5%. For a semiconductor equipment-chain company, this means the purchase return already depends heavily on AI demand staying strong for the next decade, share continuing to rise, and margins remaining elevated.

Does the current price offer a margin of safety: no.

Suitable investor type: better suited to investors who can tolerate the semiconductor cycle, value quality and long-term growth, and can wait a long time for a reasonable price; less suitable for conservative value investors who put "margin of safety" and the "valuation starting point for returns" first.

Largest uncertainties: First, how long the strong test demand from AI/HPC and high-performance DRAM can last. Second, whether the company can maintain its leading share in the high-end SoC test market over the long term. Third, whether the FY2025 operating margin of 44.2% is a new normal after the moat has deepened, or a cyclical peak created by the AI upcycle and favorable product mix.

Business Understanding

From the business structure, Advantest's core is not "selling chips," but providing test platforms, test peripherals, and test services to chipmakers. Its official product lines include SoC Test Systems, Power Test Systems, Memory Test Systems, as well as handlers, probers, device interfaces, system-level test, failure analysis systems, cloud-based test software, and verification tools. Starting in FY2025, the company integrated its reporting into two major segments: Test System and Services and Others. FY2025 revenue from the two segments was JPY 1,019.4 billion and JPY 109.2 billion, respectively, with test systems still the absolute core.

Its customers are essentially capital-intensive players in the global semiconductor value chain, mainly located in Asia. The company's risk-factor page clearly discloses that in FY2024, 89.4% of total sales were shipped to Asian regions outside Japan, especially Taiwan, mainland China, and South Korea. In FY2025, revenue from overseas customers rose further to 97.8%. This shows the business is highly globalized and highly dependent on the rhythm of Asian semiconductor capital expenditure.

The monetization model is straightforward: one part is one-time equipment sales; the other is recurring services, maintenance, training, interface boards, consumables, and related solutions that grow with the installed base. FY2025 revenue from "Services and Others" rose to JPY 109.2 billion. Management specifically noted that as the installed base expands, support service revenue continues to increase, while consumables related to high-end SoC also increased. In other words, this is not a pure one-off equipment sale, but it is also not pure SaaS. It is closer to high-barrier main equipment plus moderately recurring services and consumables.

Revenue stability therefore has a "two-layer structure": the underlying demand comes from rising global semiconductor test intensity, a relatively stable long-term trend; but short-term revenue is still affected by customer capex, product transitions, and the cycle, making it highly cyclical. The company says this plainly in its medium- to long-term management policy: semiconductor test-related markets will grow over the medium to long term, but will still experience downturns in the short term. This statement matters because it determines the investment method. You cannot treat it as a utility, and you cannot view it only as a short-term cycle trade.

On cost structure, this company does not depend on heavy fixed-asset investment to drive growth. What truly matters is R&D and engineering capability. In FY2025, R&D expense was JPY 78.1 billion, capital expenditure was JPY 34.3 billion, and depreciation and amortization was JPY 25.6 billion. Capital expenditure was only about 3% of annual sales, while R&D was 6.9% of sales. This kind of structure usually means that as long as the products and technology remain ahead, profit and cash-flow elasticity can be large; but once the technology lead is lost, profits can fall quickly.

On dependencies, this is clearly not a business with "no weaknesses." The official risk map classifies "Industry," "Sensitivity," and "Currency" as high-probability plus high-impact risks, and states clearly that changes at one or more major customers, capex volatility, and failure of customers' main products could materially hurt the company. Competitors such as Teradyne and Cohu, as well as domestic Chinese and Korean vendors, are also advancing cheaper or stronger test solutions. In other words, this business is understandable, but it cannot be called "simple and risk-free."

My judgment: this is a business I can understand, and it is a relatively clear category within technical equipment. If the stock market were closed for five years, I would be willing to hold this business; the condition is that I did not buy it at an extremely high valuation. Business understandability score: 4.5/5.

Industry and Competitive Landscape

The semiconductor test industry is not a declining industry. On the contrary, it benefits over the long term from two structural variables: rising semiconductor performance and rising semiconductor complexity. Advantest's own medium- to long-term policy emphasizes that the company's future growth opportunities come not only from semiconductor volume growth, but also from the industry's structural challenge of "complexity management." One of its core strategies is to "Outpace the growth in our core market." This view matches current trends in AI/HPC, HBM, advanced packaging, and system-level test.

Even so, this remains a cyclical industry with growth. FY2025 was explosive because demand for AI-related high-performance SoCs and high-performance DRAM was very strong, while the company successfully improved supply capability and delivered on time. At the same time, the company clearly stated that test demand for mature applications such as automotive and industrial remained soft. In other words, the long-term direction is sound, but the short- to medium-term rhythm is highly uneven.

In competition, Advantest faces a small number of strong rivals plus a group of regional challengers. The official risk factors directly list Teradyne, Cohu, CCTech, Accotest, UniTest, and EXICON as major competitors. Teradyne also publicly said in 2026 that its semiconductor test business was being supported by strong AI-driven compute demand, indirectly confirming that the whole industry is currently in an AI upcycle.

Advantest has a strong position in high-end SoC test, the most important profit pool. The company's April 2026 FY2025 presentation stated that its estimated share of the SoC tester market reached 66%, up about 10 percentage points year over year. Another official document also said the company had achieved "majority market share" in the core SoC test market. This means that at least in the segment with the highest technical complexity, highest customer stickiness, and fattest profit pool, it is not an ordinary participant. It is a clear leader.

Is the industry's profit pool concentrated? Broadly yes, especially in high-end ATE, SoC testers, and the matching interface/service chain, where leading vendors' technology and customer validation accumulation create strong clustering effects. At the same time, the official risk document also points out that "operating cost" businesses such as device interfaces face continuous cost-reduction pressure, while competitors keep launching new products with better price/performance, and customers may even develop internal test solutions. This means the industry is not a quasi-monopoly where companies can earn money passively. It still requires continuous innovation and engineering execution.

Pricing power therefore deserves a "present locally, limited overall" assessment. In high-end SoC and leading-edge memory test platforms, plus supporting hardware and software solutions, switching is difficult for customers, and Advantest's bargaining power is clearly stronger. But in mature nodes, interface boards, and some peripheral equipment, cost pressure is greater. The FY2025 operating margin surged to 44.2% largely because of a "higher mix of high-margin products." This reminds us that the high margin contains moat elements, but also cyclical and product-mix benefits.

My judgment: this is a "good company in a good industry," but the industry is inherently cyclical and should not be linearly extrapolated. Industry attractiveness score: 4/5.

Moat and Management

Moat

If the moat is broken down in a Buffett-style framework, Advantest's strongest advantage is not "consumer brand mindshare," but a compound moat of technology, scale, switching costs, and service system.

Start with technology and scale. The company's estimated SoC tester share reached 66%, a very strong figure in high-end test equipment. The company also lists "continue to outpace core market growth" as a medium-term target. Then look at service and brand. In TechInsights customer satisfaction surveys, the company has ranked first among global semiconductor test equipment suppliers for many consecutive years, and in both 2025 and 2026 it received the global number-one ranking and high ratings in multiple Top 10 large-customer service categories. In equipment, the fact that customers are willing to buy again and give high satisfaction is more important than advertising brand.

Then consider switching costs. Semiconductor test is not as simple as replacing one machine. It involves test programs, interfaces, yield ramp, production-line stability, customer engineering collaboration, and after-sales response. Advantest's product line has expanded into test systems, handlers, device interfaces, system-level test, cloud software, and the SiConic verification environment. The service segment also grows as the installed base expands. It does not have the typical network effects of an internet company, but it has a very solid engineering-niche moat.

On cost advantage, I would not define it as a "low-cost leader." The official risk factors instead emphasize that the company must keep optimizing business processes, reduce product costs, and help customers reduce total test cost, which shows that price pressure objectively exists. Put differently, Advantest's advantage is more like "high-end capability plus engineering efficiency," rather than simply relying on the lowest price.

On channels and global support, the company had 7,241 employees as of the end of March 2026, including 2,209 in Japan and 5,032 overseas. Its global depth supports customers' overseas expansion and local service needs. For large customers, this global engineering/support network is itself a barrier.

On patents, regulatory barriers, and data advantages, currently available public materials do not provide sufficiently quantified patent-pool or patent-income data, so I do not treat this as the main moat. Cloud test and data infrastructure, such as ACS RTDI, ACS Gemini, and SiConic, may strengthen data and process advantages over the medium to long term, but today they are not enough to call a decisive barrier.

Overall judgment: a moat exists, and it appears to be widening in the high-end SoC/AI test market; but this moat is not unbreakable. It must be maintained through continuous R&D and engineering execution. Moat strength score: 4/5.

Management and Capital Allocation

In governance, after June 2025 Advantest's board consists of 2 executive directors, 2 non-executive internal directors, and 5 outside directors. Two directors are non-Japanese and two are women, making independence and diversity fairly good among large Japanese manufacturers. At the management level, Douglas Lefever became Group CEO in 2024. The company has a formal succession-planning mechanism, and the risk page specifically states that the Nomination and Compensation Committee and the board continuously review succession for the CEO and key positions.

On incentives, the company uses a combination of fixed compensation, performance bonuses, and stock compensation, including RS/PSU, and sets shareholding guidelines for executives. This indicates that the incentive structure is generally long-term oriented, rather than focused only on annual profit. The issue is that the public webpages do not directly provide the complete table I would like to see for the core management team's real personal shareholding ratio. Therefore, on the question of whether management is strongly aligned with shareholders, I can only give a moderately positive but not fully evidenced assessment.

On capital allocation, the company has a clear shareholder-return framework: under its third mid-term plan, the three-year cumulative total return ratio target is at least 50%, with a commitment to stable and continuous annual dividends of at least JPY 30 per share for the full year. In April 2025, the company also announced a buyback of up to JPY 70.0 billion, with larger buyback plans subsequently advancing. The positive side is that the company is indeed returning capital to shareholders under strong cash flow. The caveat is that these buybacks occurred when the company's valuation was not cheap, so this looks more like "executing the return policy" than an extreme value discipline of "buying back heavily only when clearly undervalued."

Management has another positive point on capital discipline: the company explicitly says capital expenditure decisions will be made after assessing returns while considering the cost of capital. This statement does not guarantee every decision will be right, but at least the direction is correct.

Overall judgment: management is generally credible, the strategy is clear, and capital return is disciplined; but by a strict Buffett-style capital-allocation standard, the buyback timing is not ideal, and core executive ownership data still requires additional verification. Management and capital allocation score: 3.5/5.

Financial Quality and Owner Earnings

First, look at the most important high-confidence financial data for the past three years. The table below is organized from the company's official IR page, official earnings presentations, and Reuters/LSEG data, in billions of yen.

Metric	FY2023	FY2024	FY2025
Revenue	486.5	779.7	1,128.6
Operating profit	81.6	228.2	499.1
Net profit	62.3	161.2	375.4
Operating cash flow	32.7	286.0	335.2
Investing cash flow	-27.9	-42.2	-34.6
Free cash flow	4.8	243.8	300.6
R&D expense	65.5	71.4	78.1
Capital expenditure	20.8	21.0	34.3
Depreciation and amortization	26.1	27.1	25.6
Total assets	671.2	854.2	1,171.8
Total debt	94.4	93.5	20.2
Cash and cash equivalents	Additional data needed	262.5	340.0
ROE	Additional data needed	34.4%	57.6%

In the table, FY2023 revenue, net profit, total assets, and total debt come from Reuters/LSEG. FY2024 to FY2025 revenue, operating profit, net profit, total assets, cash, operating cash flow, investing cash flow, free cash flow, R&D, capital expenditure, and ROE come from Advantest's official IR and earnings presentations.

Looking only at these three years, financial quality is quite strong: FY2025 operating margin reached 44.2%, versus 29.3% in FY2024. FY2025 free cash flow was JPY 300.6 billion. Although this was below net profit of JPY 375.4 billion, it was still extremely high. More importantly, cumulative free cash flow over the past three years was about JPY 549.2 billion, while cumulative net profit was about JPY 598.9 billion. Three-year cumulative free cash flow was roughly 92% of net profit. This shows that the company's profits are not "pure accounting earnings." Over the long run, cash conversion passes the test, even though individual years can be heavily disturbed by the cycle and working-capital swings.

Looking further back at cash flow and investment intensity shows both the "good" and the "difficult" parts of this business more clearly:

Metric	FY2020	FY2021	FY2022	FY2023	FY2024	FY2025
R&D expense	42.7	48.4	60.1	65.5	71.4	78.1
Capital expenditure	13.7	18.0	25.0	20.8	21.0	34.3
Depreciation and amortization	11.8	15.0	21.4	26.1	27.1	25.6
Operating cash flow	67.8	78.9	70.2	32.7	286.0	335.2
Investing cash flow	-16.8	-46.9	-26.7	-27.9	-42.2	-34.6
Free cash flow	51.0	32.0	43.5	4.8	243.8	300.6

All of these figures come from the company's official annual earnings presentations. They reveal an important fact: Advantest is a company with light capital expenditure, steadily rising R&D investment, and cash flow that is highly sensitive to the cycle. In good years, free cash flow can gush out. In bad years, free cash flow can quickly approach dryness. For long-term investors, this means the company is not a model that "needs more and more money as it grows," but it is indeed a model where "the closer it gets to the cyclical trough, the worse cash flow looks."

On working capital, FY2025 total assets increased by JPY 317.6 billion, mainly due to an increase of JPY 115.7 billion in receivables, JPY 77.4 billion in cash, JPY 41.8 billion in other financial assets, and JPY 22.0 billion in inventory. On the liability side, payables increased by JPY 35.0 billion and taxes payable increased by JPY 39.4 billion, while short-term borrowings decreased by JPY 75.0 billion. In other words, FY2025 did consume working capital under rapid growth, but because operating profit was so strong, operating cash flow still reached a record. This structure is healthy, unlike forcibly beautifying cash flow by squeezing inventory and payables.

Financial safety is a clear strength. At the end of March 2026, the company had JPY 340.0 billion in cash and about JPY 20.19 billion in total debt, implying net cash of about JPY 319.8 billion. The equity ratio was 67.9%, and the R&I rating was A+. This gives the company strong survival and reinvestment capacity even if the industry turns down. From the perspective of "permanent capital loss," this is very important.

On accounting quality, I currently do not see direct signs of obvious financial fraud or aggressive accounting, but two points should be recognized. First, FY2025 Q4 recognized about JPY 17.3 billion in financial income from fair-value measurement after exercising subscription rights related to a strategic investment. Second, FY2024 operating profit included about JPY 21.4 billion in impairment of goodwill and intangible assets. In other words, annual profit contains some non-operating noise. Valuation should not focus only on net profit, but should return to operating cash flow and owner earnings.

Owner Earnings Analysis

Using a Buffett-style owner earnings approach, I would start with the most conservative version and would not casually classify any capital expenditure as "pure growth capex."

FY2025 net profit was JPY 375.4 billion; depreciation and amortization was about JPY 25.6 billion; capital expenditure was JPY 34.3 billion. If we make only a simplified "income-statement perspective" adjustment, the cash approximation of after-tax operating profit is still high. But because operating cash flow already incorporates working-capital changes, the more prudent calculation is: operating cash flow of JPY 335.2 billion minus total capital expenditure of JPY 34.3 billion equals JPY 300.6 billion. This is in fact FY2025 free cash flow, and can also be viewed as my conservative owner earnings.

Why is this conservative? Because treating all capital expenditure as maintenance capex assumes all capacity expansion, IT, and facility investments are only "baseline needs." In reality, some of it must be growth-oriented. Conversely, FY2025 receivables and inventory growth also consumed some cash, so in a more neutral year, owner earnings may be slightly higher than that year's free cash flow. For conservatism, I do not mark this up in valuation. Instead, I treat JPY 280.0 billion to JPY 300.0 billion as the current relatively credible conservative owner-earnings range.

Based on the June 3, 2026 close of JPY 27,660 and roughly 732 million shares outstanding, the equity value is about JPY 20.2 trillion. This corresponds to a conservative Owner Earnings multiple of about 67 to 72x. For an excellent but clearly cyclical semiconductor test equipment company, this multiple is very high within a value-investing framework.

Valuation and Margin of Safety

Owner Earnings Discount Method

Here I do not use the "most optimistic FY2026 guided profit" directly as the base. Instead, I use a more conservative owner-earnings range for modeling. The reason is simple: you are not facing a company that has not yet delivered growth, but a company that has already translated AI strength into its income statement. The model assumptions are below. The discounted values are my manual estimate ranges, intended to judge margin of safety, not decimal-point precision:

Scenario	Starting owner earnings	First 10 years growth	Discount rate	Perpetual growth	Derived intrinsic value per share
Conservative	JPY 220.0 billion to JPY 250.0 billion	3% to 5%	10%	2%	About JPY 4,500 to JPY 6,500
Base	JPY 280.0 billion to JPY 320.0 billion	6% to 8%	9%	2.5%	About JPY 8,000 to JPY 12,000
Optimistic	JPY 350.0 billion to JPY 380.0 billion	10% to 12%	7% to 8%	3%	About JPY 15,000 to JPY 22,000

These assumptions are not invented casually: starting earnings are based on FY2025 free cash flow of JPY 300.6 billion and FY2026 high-cycle guidance; the discount rates refer to the current Japanese 10-year government bond yield of about 2.651% and the U.S. 10-year Treasury yield of about 4.481%, while also accounting for the fact that this is a high-beta, clearly cyclical equipment company; growth assumptions consider the company's medium-term targets, AI-related demand, and share advantage. Even so, in my view, the current share price of JPY 27,660 is still above my value range under the "base" case and most "optimistic" scenarios.

Therefore, from an Owner Earnings DCF perspective, my ranges are:

Conservative intrinsic value range: JPY 4,500 to JPY 6,500
Reasonable intrinsic value range: JPY 8,000 to JPY 12,000
Optimistic intrinsic value range: JPY 15,000 to JPY 22,000

Compared with the current JPY 27,660, the market price carries a clear premium to the reasonable range. Even against the optimistic range, it is broadly at a high level or beyond the upper end. For a balanced but conservative long-term investor, this does not meet the standard of a "sufficient margin of safety."

Relative Valuation Method

On relative valuation alone, Advantest is not necessarily the most expensive company in the sector. Current market data broadly shows: Advantest trailing P/E of about 50 to 51x and EV/EBITDA of about 34.4x; Teradyne market cap of about USD 57.84 billion, trailing P/E of about 68.6x, and EV/EBITDA of about 54.9x; Cohu currently has almost no meaningful trailing P/E, forward P/E is close to 96x, and market cap is about USD 2.49 billion to USD 2.72 billion. From this, Advantest is not the most extremely expensive relative to Teradyne, and in some sense is even cheaper.

But relative valuation can only show that "it is not much more expensive than the most expensive peer." It cannot show that "it is cheap." The issue is that Teradyne itself is also in an AI-driven high-valuation state, while Cohu's multiples are weak because the profit base is weak and expected elasticity is high. So the best conclusion relative valuation can offer here is: Advantest is "less expensive among peers," not "absolutely cheap." For value investing, these two propositions are very different.

Asset Value or Liquidation Value Method

From an asset-value perspective, the conclusion is even colder. At the end of March 2026, the company had JPY 340.0 billion in cash, about JPY 20.19 billion in total debt, and JPY 795.7 billion in total equity. In other words, it is certainly safe and has a net-cash advantage; but compared with the current rough equity market value of more than JPY 20 trillion, this net cash and book equity provide almost no valuation cushion. If your investment logic requires an "asset floor" as protection, Advantest does not satisfy it. The investment logic here can only come from future cash flow, not existing assets.

Margin of Safety Judgment

Putting DCF, relative valuation, and the asset method together, my conclusion is clear: This is a "good company, but most likely a bad price."

The most fragile assumption in the current valuation is that the market assumes the testing boom from AI/HBM will last long enough, and that Advantest can retain current high margins near FY2025 levels for a fairly long period. If growth is below expectations, or if operating margin falls from FY2025's 44.2% back to the company's revised MTP3 target range of 33% to 36%, returns at the current share price would deteriorate materially. If valuation multiple compression is added, a large degree of permanent capital loss is entirely possible.

Therefore my price judgment is:

Ideal buy price range: JPY 7,000 to JPY 10,000
Acceptable holding price range: JPY 10,000 to JPY 14,000
Clearly overvalued price range: above JPY 18,000 to JPY 20,000

These ranges are conclusions under a strict Owner Earnings framework, not under a trend-trading framework. If you are willing to accept a lower discount rate, higher sustained growth, and a high premium for scarce AI assets, the ranges can move up. But that would no longer fit the "balanced but conservative" margin-of-safety standard.

Risks, Comparisons, and Final Judgment

Risks and the Strongest Bear Case

The core risk is not short-term share-price volatility, but buying an excellent yet highly cyclical company at an extremely high price near a cyclical peak. The official risk map classifies industry, demand sensitivity, and currency as "high-probability plus high-impact." The company also clearly emphasizes that major customer capex changes, customer product failure, international operating risks, regulatory changes, and intensifying competition could all hurt performance. Especially given the very high concentration in Asia, geopolitics, export restrictions, and currency fluctuations can be amplified.

The second major risk is technology and competition. Advantest is strong, but it has rivals. The company itself lists Teradyne, Cohu, and multiple Chinese and Korean peers. It also specifically notes that if competitors obtain core technologies, customers develop internal test solutions, or key PCB design and manufacturing technologies leak, the company's product-performance advantage and pricing power could be damaged. For a high-margin equipment maker, this is a real moat risk.

The third major risk is "margin illusion." FY2025's high margin is clearly related to a higher high-end SoC mix, AI strength, and improved supply capacity. But the company also clearly wrote that mature fields such as automotive and industrial remain soft, while operating-cost businesses such as interfaces face long-term price-reduction pressure. If the market treats FY2025's 44.2% operating margin as the long-term steady state, it can easily overestimate the company's true long-term earnings power.

The fourth risk is "high-valuation buybacks" in capital allocation. The company's shareholder-return framework is clear, cash is abundant, and buybacks are not inherently bad. The issue is that high-multiple buybacks do not necessarily increase intrinsic value per share. For value investors, "the company is buying back shares" does not automatically mean "shareholders benefit." The key is the repurchase price.

The strongest opposing view can be summarized in one sentence: Advantest is indeed great, but this AI dividend has pulled too many years of future good news into the current price. If AI test demand over the next three to five years merely "slows" rather than "collapses," the company may still operate excellently, but shareholders may still receive poor returns because of valuation decline. That is the most dangerous "good company, bad price" situation in long-term investing.

The facts that could overturn my current cautious view, meaning prove me wrong, mainly fall into several categories: first, Advantest continues to significantly expand its share in high-end SoC and AI accelerator customers over the next two to three years and turns FY2026 guidance into an even higher cash-flow platform; second, service, consumables, and software revenue continue to rise as a share of the total, clearly increasing business recurrence; third, the high margin is proven to be a new normal under a new technology generation, rather than a cyclical peak. Conversely, if SoC share falls materially, service revenue no longer grows with the installed base, free cash flow remains below net profit for a long period, or medium-term ROIC targets cannot be delivered, I would more quickly admit that my view was too optimistic.

The largest permanent capital loss scenario, in my view, is: you buy at about 50x P/E and about 67x conservative owner earnings; then the industry cycle turns down, operating margin returns to a more normal level around 30%, and the market simultaneously compresses the valuation to around 25x. In that case, even if the company remains the industry leader, the share price could fall 50% to 70% from the current level. This is an inference based on historical cash-flow volatility, the current high valuation, and the characteristics of cyclical industries, not a short-term share-price forecast.

Comparison with Other Opportunities

Compared with Teradyne, Advantest may have a stronger position in high-end SoC test, and its current relative multiple is slightly lower. But Teradyne itself is also expensive, so this comparison cannot conclude that "Advantest is cheap." More precisely, it is simply a less outrageous member of the expensive AI test sector.

Compared with a broad market index, I do not think Advantest at the current price is clearly superior to the index. Its business is concentrated in a single high-volatility track, while the current trailing earnings yield is only about 2.0%, and the conservative Owner Earnings yield is only around 1.5%. This means your expected return mainly comes from future growth, not from a cheap entry point today. By contrast, a broad market index lacks this sharp AI elasticity, but also lacks this level of single-theme risk. S&P Global defines the S&P 500 as 500 large companies covering about 80% of U.S. market capitalization, making it a diversified asset by design.

Compared with the risk-free rate, buying Advantest at the current price also does not have an attractive starting yield. Reuters/LSEG pages show the Japanese 10-year government bond yield at about 2.651% and the U.S. 10-year Treasury yield at about 4.481%, while Advantest's current trailing earnings yield is about 2.0%, and its conservative owner-earnings yield is even lower. In other words, buying it does not win through "current yield," but by betting that future high cash-flow growth will continue to materialize. For conservative investors, this is uncomfortable.

If I could hold only five assets, I would not put it in the portfolio at the current price. I am willing to track it long term, and I acknowledge it may continue rising, but from the perspective of capital-use efficiency and margin of safety, this is not where I most want to place a bet.

Investment Checklist

Check item	Conclusion
Can I understand this business?	Pass
Does it have long-term stable demand?	Pass
Does it have a durable moat?	Pass
Does it have pricing power?	Uncertain
Can it generate stable free cash flow?	Uncertain
Are its returns on capital excellent?	Pass
Is management trustworthy?	Pass
Is capital allocation rational?	Uncertain
Is the balance sheet robust?	Pass
Is valuation below intrinsic value?	Fail
Is the margin of safety sufficient?	Fail
Would long-term holding make me comfortable?	Uncertain
What key facts would make me sell?	Pass
Do I want to buy only because the share price has risen or market sentiment is strong?	Fail

The basis for these conclusions is that the company's business model and industry trend are clear, the moat is solid, and the financials and balance sheet are excellent, but the current valuation is materially above the reasonable value range I derive under the Owner Earnings framework.

Final Investment Conclusion

【Final Rating】 Watch

【One-sentence investment thesis】 Advantest is a high-quality semiconductor test leader with strong competitive advantages and the ability to earn money over the long term, but at the current price, investors are effectively paying upfront for perfect execution over the next decade.

【Core bullish reasons】 First, the company sits deep in the beneficiary chain of rising semiconductor complexity and AI/HPC/HBM expansion, so the long-term demand direction is right. Second, its estimated share of the high-end SoC test market has reached 66%, indicating a strong competitive position. Third, service and consumables revenue grows with the installed base, giving it better recurrence than a pure equipment-sale model. Fourth, the capital structure is safe, net cash is significant, the equity ratio is 67.9%, and the company can endure the cycle. Fifth, customer satisfaction and technology leadership have been externally validated over the long term.

【Core bearish reasons】 First, the current valuation is too high, with trailing P/E of about 50 to 51x, EV/EBITDA of about 34.4x, and a conservative owner-earnings multiple of about 67 to 72x. Second, FY2025 margins very likely contain a meaningful cyclical and product-mix benefit. Third, customer capex, the pace of AI investment, and exchange rates will make performance highly volatile. Fourth, competition is not easy, with Teradyne and Chinese and Korean vendors all catching up. Fifth, although the company is active in buybacks, they did not occur during a period of clear undervaluation.

【Key assumptions】 For the investment case to work, at least four conditions must hold: AI and high-performance memory test demand does not fall off a cliff in the medium term; Advantest maintains its technology and share advantage in high-end SoC; service/consumables/software revenue continues to rise with the installed base; and even if margins decline, they can still stabilize near the company's revised MTP3 target range.

【Fair Buy Price】 Under a strict Owner Earnings framework, the more attractive buy range is JPY 7,000 to JPY 10,000. If you already hold the stock and tax or portfolio reasons make selling difficult, I think JPY 10,000 to JPY 14,000 is the price band where holding can be relatively comfortable. Above JPY 18,000 to JPY 20,000, the valuation has clearly priced in too much. This conclusion is based on conservative cash-flow assumptions, not market momentum or a scarcity premium for the theme.

【Target holding period】 If bought at a reasonable valuation, this is a business that could be considered for five to ten years or more. But at the current price, I value waiting more than rushing to own it.

【Expected annualized return】 If bought at the current price, my 10-year expected return range is: conservative scenario -10% to -3%, base scenario -2% to +4%, optimistic scenario +5% to +9%. The optimistic scenario requires an FY2026-like high-cycle profit platform to keep expanding for many years, with only limited valuation compression.

【Maximum loss risk】 In the worst case, I believe there is a 50% to 70% risk of permanent capital loss. The reason is not that the company will fail, but that industry strength could fade while the valuation multiple normalizes at the same time. For buyers at high multiples, this "the company remains excellent, but shareholders lose a lot" outcome is real. This judgment is based on the current high valuation and cyclical characteristics.

【Tracking indicators】 I will focus on tracking: high-end SoC tester market share; AI/HBM-related order and shipment rhythm; growth in service and consumables revenue; whether operating margin can remain above 33% to 36%; the three-year cumulative match between free cash flow and net profit; receivables and inventory turnover; changes in the revenue share from Taiwan/China/Korea; major customer capex outlook; buyback price and buyback scale; and whether net cash is maintained.

【Signals that trigger reassessment】 If the following occur, I would immediately review the thesis: high-end SoC market share declines materially; AI-related demand is weaker than expected for two consecutive quarters; the service segment no longer grows with the installed base; free cash flow lags net profit for a long period; the company shifts from net cash to meaningful net debt; management continues large-scale buybacks at extremely high valuations; or regulatory/geopolitical events materially affect Asian customers and the supply chain.

【Final recommendation】 Calmly stated, Advantest deserves respect and long-term tracking. But if your framework is "buy stocks as if buying a business," you should not give up price discipline just to own a good company. For balanced but conservative investors with a horizon of more than 10 years, my recommendation is not to chase the stock higher, but to put it on a high-priority watchlist and wait for a real margin of safety created by any future industry correction, customer destocking, cooling AI expectations, or valuation compression.

Open Questions and Data Limitations

Three points need to be clearly marked. First, for some earlier years, the complete consolidated revenue, operating profit, ROIC/ROA, and dividend sequences were not as complete in webpage parsing as the most recent three years, so this report focuses on high-confidence data from the past three years, supplemented by longer-cycle cash-flow and R&D data. Second, the precise shareholding ratio of core executives is not fully laid out in currently available webpages, so the assessment of "management-shareholder alignment" can only be made with medium confidence. Third, different data sources have delays and methodology differences for near-term share price/market value, so this report builds its current valuation analysis more on multiples and cash-flow capacity than on over-precise treatment of one point-in-time price.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.

Mentioned Tickers

TER COHU

Baillie Framework · Ten Questions for Growth Investing

Hunting ten-year five-baggers among great growth stocks — pressing the upside question: "Can it get much bigger?"

How high is its market ceiling? Is it expanding an existing pie, or creating an entirely new market?5/10

Bottom line: this looks more like “expanding an existing and growing semiconductor test pie” than “creating a new market.” The ceiling is real, but growth is moderate. On top of that, Advantest already holds about 66% share in SoC testers, so a simple TAM extrapolation is unlikely to support the “fivefold in ten years” that Baillie Gifford looks for.

Start with the pie itself. Global semiconductor ATE/test equipment is a mature installed-base market: narrow-scope ATE is about USD 7–9bn, with CAGR of about 4–5% before 2030, while the broader “test equipment” market is about USD 15bn (2025) → 21.6bn (2031), with CAGR of about 6.1%. This is an existing market growing in the single digits, not a new track emerging from zero.

The real upside comes from rising “test intensity” in both volume and price: 2nm, HBM (6.4 Gb/s/pin), advanced packaging/chiplets all push up test time per chip, pin count, and equipment ASP at the same time, while system-level test (SLT) is expanding from about 15% share at a double-digit growth rate. Advantest is precisely the company monetizing this segment. The report states its SoC tester share is about 66% (about +10pct YoY), FY2025 revenue was ¥1.1286 trillion, and SoC tester revenue rose +74.3% to ¥767.4 billion, increasing to 68% of revenue. Its AI/HPC test revenue CAGR has been about 58% over the past two years, far above the industry.

But judged honestly through the Baillie Gifford lens: by nature, this is “making the existing pie larger and deeper,” not “creating a new pie.” The ceiling is capped by semiconductor testing, a moderately growing installed-base market, while 66% share means most of the share upside has already been captured and the room for further gains is limited. Future incremental growth mainly depends on an upcycle in test intensity, which is highly cyclical. That is exactly why the report concludes “strong cyclicality + no valuation margin of safety (P/E about 50x).” Conclusion: the ceiling is good enough, but not “limitless”; it is difficult for it to naturally support a fivefold gain over ten years. This is not a typical LTGG-style stock.

Jun 4, 2026
Can its revenue at least double over the next five years? Will growth mainly be driven by volume, price, or new businesses?4/10
A
Bottom line: within five years (to around FY2030), revenue doubling from ¥1.1286 trillion to more than ¥2.26 trillion only works under an optimistic scenario in which the AI/HBM test boom stays strong for a long time. It is not the base case. More honestly, under the company’s own medium-term planning framework, normalized revenue is actually below the current level.

First, break down the drivers (almost all current incremental growth comes from “volume + price”; new businesses are still supporting actors):
- Volume is the main engine. Demand for testing AI/HPC high-performance SoCs and HBM/high-performance DRAM is driving shipments. The company has already expanded SoC tester capacity from 3,000 units in July 2025 to 5,000 units in 2027, and SK hynix expects HBM demand to grow at about 50% CAGR from 2024–2028.
- Price/mix is the source of profit leverage. The rising share of high-end SoC testers has pushed operating margin to 44.2%, but that includes cyclical and product-mix tailwinds.
- New businesses (Services and Others ¥109.2 billion, SLT, cloud test software) are growing moderately with the installed base and becoming more recurring, but their share remains small and they are unlikely to independently support a doubling within five years.
But the strong cyclicality must be faced directly: revenue went from FY2023 ¥486.5bn → FY2024 ¥779.7bn → FY2025 ¥1,128.6bn, which itself is a near-doubling type of violent fluctuation, and it can also fall sharply in reverse. The report repeatedly stresses that “short-term revenue rises and falls sharply with customer capital expenditure.” The strongest counterevidence is the company’s own framework: after the October 2025 upgrade, the MTP3 medium-term sales target is only ¥835.0–930.0 billion, with operating margin of 33–36%, which is below the current run-rate. This shows management does not treat ¥1.1–1.42 trillion as a sustainable platform, but as a cyclical high point in the AI upcycle.

Honest judgment: FY2026 guidance has already jumped to ¥1.42 trillion, so the starting point is high, and looking at one year alone, the growth rate is not modest. But to double again from here and hold that level requires AI accelerator and HBM capex to stay strong for many consecutive years, no share loss, and no margin normalization. A five-year doubling is an optimistic scenario, not the base case; the base case is more likely “high-level volatility with one clear pullback.”
Jun 4, 2026
After five years, what will take over as the next growth engine? Does this “second curve” exist today?3/10

Bottom line: Advantest does not have a “landed and independently capable of taking over” second curve today. The closest source of recurring revenue, “services/consumables/software,” is real, but small and growing far more slowly than the core business; the truly exciting cloud data business is still early-stage investment rather than confirmed revenue; five years from now, the likely successor is still the “same” high-end test main curve widened by HBM/advanced packaging. That concentration is exactly what deserves the most caution under Baillie Gifford’s “firepower in years 3-10” standard.

Rank the candidates honestly by “scale × certainty”:

① Services/consumables/software (Services and Others): the only candidate already generating scaled revenue. FY2025 revenue was ¥109.2 billion, up only +12.7% YoY, far below Test System’s +49.3%, and less than 10% of total revenue; the segment only just turned profitable (segment profit ¥8.8 billion, including about ¥2.5 billion of disposal gains). It does become more recurring as the installed base expands, but today it is a “stabilizer,” not an engine capable of carrying a fivefold gain over ten years.

② System-level test SLT and device interfaces: a growth direction in the narrative, but small in scale, and interface/handler sales move in the same direction as testers themselves, so this is not a new curve independent of the main cycle.

③ Cloud test software/data infrastructure (ACS RTDI, ACS Gemini, SiConic): the most imaginative area, but ACS Gemini was only launched in December 2024 as a developer/digital twin platform, and the company has not disclosed standalone revenue to date. This is early-stage positioning/concept, not a landed second curve.

④ High-performance memory/HBM testing: the most certain growth area (HBM4 mass production is approaching in 2026), but it is endogenous to the Test System core business and is an extension of the first curve, not a “second.”

Honest judgment from the Baillie Gifford perspective: firepower in years 3-10 is highly dependent on the same SoC/HBM test main curve staying strong, and there is no already-realized, low-correlation relay baton. This is a good company, but the “second curve” question should receive a relatively weak score, consistent with this report’s Watch rating, rather than being marked up for the growth story.

Jun 4, 2026
What is its core competitive advantage? Will this moat widen or narrow over the next three to five years?6/10

Bottom line: the core competitive advantage is a “compound moat” of technology leadership / scale share / high switching costs / global service network (report score 4/5). Over the next three to five years, it will probably “widen at the high end, but remain asymmetric overall”: rising AI/HBM test complexity makes the fattest profit pool deeper, but the moat depends on continuous R&D engineering rather than an unbreakable structural barrier. The probability of sustaining it for ten years is “fairly high but not certain,” and it is not enough to unconditionally justify the moat premium implied by about 50x P/E.

Core advantage breakdown. This is not brand or network effects, but an engineering niche: ① scale share: SoC tester share rose to about 66% (YoY +10pct, from 56% to 66%), and together with Teradyne it forms an approximately 80% duopoly structure; ② switching costs: test programs, interface boards, yield ramp, production-line stability, and customer engineering collaboration create deep lock-in, so replacing a tester is not simply “changing one piece of equipment”; ③ customer validation: TechInsights’ 2026 survey ranked it No. 1 globally for the seventh consecutive year and first among assembly/test equipment suppliers; ④ an engineering support network of 7,241 employees worldwide. On cost, it is a “high-end capability + engineering efficiency” player rather than a low-price leader.

Forces widening the moat. AI accelerators, custom ASICs, and HBM make testing harder and more expensive. HBM has become the leading cause of data-center GPU failures, forcing testing to “shift left” while site count and complexity both rise. V93000 EXA Scale benefits from a large installed base and has become a de facto standard, helping the leader deepen its moat.

Forces narrowing the moat (must be stated honestly). Teradyne is in a phase of “regaining share” and counterattacking in high-end compute testing, while Chinese and Korean vendors are catching up at lower prices; interface and other operating-cost businesses face long-term price pressure; the report also lists risks including customers developing in-house test solutions and leakage of core PCB processes. More importantly, FY2025’s 44.2% margin contains cyclical and product-mix tailwinds, and is not pure evidence that the moat has widened.

Baillie Gifford conclusion. This moat is likely to be “wider in the high-end segment, pressured in mature segments” over the next three to five years. It can outperform its core market, but it must be maintained through ongoing engineering investment and can be eroded. Buying at about 50x P/E effectively prepays for a perfect script in which “the moat must keep widening for ten years.” Good company, expensive price: the moat supports a good business, but not the current price.

Jun 4, 2026
If its core business is disrupted, does it have the DNA to reinvent itself? How does it deal with mistakes and bad news?5/10

Bottom line: Advantest looks more like an engineering organization willing to iterate on itself and actively buy or build “disruptors” into its own product line, rather than a path-dependent company clinging to a single high-end ATE paradigm. Its disclosure around bad news is also relatively candid. But we must be honest: its reinvention so far remains within the core radius of “testing,” high-end ATE is still the profit lifeline, and true paradigm substitution has not yet been proven.

① Reinvention DNA: active expansion through M&A + internal development. The flagship V93000 SoC platform was not originally developed in-house, but iterated for more than twenty years after the 2011 acquisition of Verigy for about USD 1.1 billion. Facing “SLT replacing traditional ATE,” the most direct disruption path, it acquired Astronics’ SLT business in 2018 for USD 185 million, turning a rival into its own product line. More importantly, the SiConic validation environment launched in 2025 explicitly “offloads” bring-up/debug from ATE to the bench, combined with ACS cloud subscription. In effect, Advantest is personally building the things that could cannibalize the “one-shot machine sale” model. This willingness to “disrupt itself” is exactly the adaptability signal Baillie Gifford values; the product line has expanded to handler/prober/SLT/cloud software/failure analysis.

② Candor around bad news: above passing. FY2024 operating profit openly included about ¥21.4 billion of goodwill/intangible asset impairment, and the company explicitly said this was because the Essai socket business was hit by weak demand from major customers and slower-than-expected new-customer development. The report also does not shy away from the fact that the 44.2% margin contains cyclical and product-mix tailwinds, or that it may fall back to 33%–36%. Admitting an overpaid asset and flagging the margin illusion are positive evidence of a correction culture.

③ Governance: professional managers + succession mechanism, supportive of long-term transformation. Douglas Lefever became Group CEO in April 2024. He is a 25-year insider and industry veteran who led the Verigy/Astronics acquisitions, not a family controller. The board has a formal succession review process. Internal promotion of an industry veteran plus institutionalized handover lowers the risk of “paradigm lock-in.”

Honest landing point: this is a resilient, iteration-minded engineering organization, but if DFT/BIST and customer in-house development truly pull volume and price away from high-end ATE, the 4/5 moat will still come under pressure. That is the underlying reason for a “Watch” rating rather than “Buy.”

Jun 4, 2026
Does management (especially the founder) have a long-term perspective, and are its interests deeply aligned with the company? Is it willing to sacrifice current profits for five to ten years from now?4/10

Conclusion: moderately positive, but “natural long-term alignment” is absent and evidence of interest alignment is relatively weak. First, one premise must be made explicit: Advantest has no controlling founder and no family blockholder. The company was founded by Ikuo Takeda in 1954 as Takeda Riken and renamed in 1985. Today it is a typical large Japanese listed company with professional managers + institutional ownership, so the founder-style natural long-termism favored by Baillie Gifford, where the founder and company share the same fate, does not structurally exist here. The evaluation must instead shift to whether professional management is long-term oriented and aligned with shareholders.

There is meaningful positive evidence: Douglas Lefever became Group CEO in April 2024, the chair and CEO roles are separated, outside directors form a majority, the succession mechanism is formal, and governance is relatively clean among Japanese corporates; the medium-term plan (MTP3) framework is clear, with a three-year cumulative total return ratio target of ≥50% and an annual dividend per share of at least ¥30; R&D reinvestment continues (FY2025 R&D of ¥78.1 billion and rising year by year), and capital expenditure emphasizes evaluating returns after considering the cost of capital. These point to “operating for the long term rather than for single-year profit.”

But two weaknesses must be stated honestly: ① insufficient evidence of interest alignment: fixed salary + performance bonus + stock compensation (RS/PSU) plus shareholding guidelines point in the right direction, but public materials do not fully disclose the actual shareholding ratios of core executives, so the hard evidence of “net worth tied to the stock” is missing; ② buybacks at high valuation: a ¥70.0 billion buyback in April 2025, followed by another buyback of up to ¥150.0 billion in October 2025, with the latter announced right after strong earnings and a one-day stock surge, at a high valuation of about 50x P/E. This looks more like “executing a return policy” than “buying back only when materially undervalued,” and buybacks at high multiples may not increase intrinsic value per share. Overall, management’s long-term orientation and discipline are credible, but the evidence that it is “deeply in the same boat as long-term shareholders” is not yet sufficient.

Jun 4, 2026
If it disappeared tomorrow, how much would customers miss it? Is its growth method sustainable and not dependent on harming society or regulation?6/10

Bottom line: if Advantest disappeared tomorrow, global high-end chipmakers would miss it badly. It is a “chokepoint” key leader in the AI test chain, but not the only source. Its growth is based on real industrial demand and does not cross regulatory red lines, but 97.8% overseas revenue + high concentration in Asia significantly amplify geopolitical and FX risks. Overall, this remains consistent with “Watch,” not a reason to mark it up.

① Indispensability: key leader, not sole source. In the most complex and most profitable high-end SoC testing, Advantest has about 66% share (YoY +10pct, with the market expanding from USD 4.1 billion to USD 6.9 billion), making it effectively the “ASML of testing.” A Blackwell-class accelerator can spend more than 20 minutes on the same tester, so machine time/yield directly constrains customer shipments. Together with high switching costs from test programs, interfaces, yield ramp, production-line collaboration, and after-sales support, customers truly “cannot switch away” in the short term. But it is not a monopoly: Teradyne, Cohu, and Chinese/Korean vendors can still substitute, and the report also explicitly notes possible customer in-house test solutions and long-term pressure on interface businesses. Therefore the right description is key leader, high switching costs, but not exclusive. If it disappeared tomorrow, it would severely disrupt the industry’s capacity-ramp schedule, but not create a permanently unsolvable problem.

② Growth sustainability: real demand, but amplified geopolitical/FX risks. Growth is built on hard industrial demand: Advantest is expanding capacity from 3,000 units toward 5,000, with a path toward 10,000 SoC systems per year, corresponding to B300, AMD MI400, Broadcom/Marvell custom silicon, and HBM4 mass production; Teradyne also confirms the AI-driven test upcycle. This is not growth through social harm. But the red line is external: overseas revenue is 97.8%, shipments are highly concentrated in Taiwan/China/Korea, BIS tightened controls on 24 categories of semiconductor equipment in January 2026, and the export environment can change at any time; in FX, USD/JPY was near 160 on 2026-06-04, with the probability of a central-bank rate hike about 78%. A weak yen is a current tailwind, but also a reversible source of high volatility. The report lists industry/sensitivity/FX as “high probability + major impact,” consistent with this view: good business, real demand, but real fragility, so maintain Watch.

Jun 4, 2026
How are the unit economics of this business (gross margin, incremental returns)? Do they improve or deteriorate as scale increases? Where does the money it earns go?7/10

Bottom line: the “level” of unit economics is outstanding, but the “stability” is poor. This is a high-return machine amplified by the cycle, closer to “a cyclical stock in a good year” than a “steady-state compounding machine.” The money earned goes into R&D reinvestment, dividends, and buybacks. The direction is right, but the buybacks happen to occur during a high-valuation period. For the “high and sustainable incremental return on capital” that Baillie Gifford cares most about, Advantest satisfies “high,” while “sustainable” is questionable.

① Unit economics level: top-tier, asset-light, high-return. FY2025 operating margin was 44.2% (up 14.9 percentage points from 29.3% the previous year, with gross margin rising from 57.1%→64.3%), and ROE surged from 34.4% to 57.6%. The root cause of extremely high incremental returns is “R&D-heavy, capex-light”: FY2025 capex was only ¥34.3 billion (about 3% of revenue), R&D was ¥78.1 billion (6.9% of revenue), and depreciation/amortization was ¥25.6 billion. Growth is not mainly driven by piling on heavy assets, but by engineering capability. This is exactly the asset-light, high-compounding feature Baillie Gifford likes.

② As scale increases, it improves and deteriorates at the same time. The improving side: a higher share of high-end SoC directly lifts gross margin and operating margin, while share gains strengthen pricing power. The deteriorating side has two layers: first, mature-node and interface “operating-cost businesses” face long-term price pressure; second, and more fatal, FCF is extremely cycle-sensitive. FY2023 was only ¥4.8bn, FY2024 was ¥243.8bn, and FY2025 was ¥300.6bn; in a bad year it nearly dries up. Honest judgment: this looks more like “a cyclical stock amplified by the boom” than a smooth compounding machine. The positive point is through-cycle earnings quality: cumulative three-year FCF was about ¥549.2 billion ≈ 92% of cumulative net profit of ¥598.9 billion, so the profits are not just accounting paper.

③ Where the money goes: R&D reinvestment, dividends (annual dividend per share at least ¥30), buybacks (up to ¥70.0 billion in 2025-04, then up to another ¥150.0 billion in 2025-10, capped at 18.00 million shares), and net cash of about ¥319.8 billion. The issue is that buybacks are taking place at about 50x P/E, a high-valuation period, so they may not increase intrinsic value per share.

Honest landing point under the Baillie Gifford framework: incremental return on capital is unquestionably “high,” but “sustainable” deserves a question mark. The 44.2% margin is above the company’s own revised MTP3 target of 33%–36% (itself raised from the original 22%–28%), meaning management also assumes the current margin contains cyclical tailwinds and carries risk of falling back to 33%–36%. The unit economics are “truly good,” but they should be understood as “good at a cyclical high,” not linearly extrapolated into a permanent steady state.

Jun 4, 2026
What conditions must all hold for it to rise fivefold in ten years? Are these conditions realistic? What expectations does today’s share price imply?2/10

Bottom line: a fivefold gain in ten years is not the rational central scenario at the current price. It is an optimistic tail scenario that requires multiple extreme assumptions to come true at the same time. Lay out the math clearly: current market cap is about ¥19–20 trillion (stockanalysis shows about ¥19.08 trillion on June 2, with the share price about ¥28,170), so a fivefold gain means about ¥95–100 trillion, putting it among the very largest semiconductor companies globally by market cap. To get there, the following conditions must hold simultaneously:

① Profit must explode. If the market is only willing to pay a “normalized” equipment-stock multiple ten years from now (P/E 25–35x), a ¥95–100 trillion market cap implies net profit of about ¥2.7–4.0 trillion, which is about 8–10 times FY2025’s ¥375.4 billion, requiring net profit CAGR of about 23%–26%/year, continuously for ten years.

② Even if today’s ~50x high multiple is maintained, it still needs ~17.5%/year. Net profit would still need to reach about ¥1.9–2.0 trillion (5 times), and the market would have to still pay nearly 50 times earnings for a strongly cyclical equipment company ten years later. That itself is an unrealistic stacked assumption.

③ This requires all three things to go right: AI/HBM test demand stays highly prosperous for ten years (while the long-term semiconductor test market grows at only mid-single digits), operating margin stays locked at FY2025’s high 44.2% level for the long term (rather than falling back to the company’s revised MTP3 target of 33%–36%), and about 66% SoC share is not eroded by Teradyne/Chinese and Korean competitors. Each item may be achievable on its own, but the joint probability of all three hitting for ten years is very low, and it badly diverges from the industry’s own growth rate.

Today’s share price already implies very high expectations: trailing P/E about 50x, conservative owner earnings about 67–72x, and owner earnings yield only about 1.5%. The optimistic upper end of the report’s Owner Earnings DCF intrinsic value, ¥22,000, is still below the current price of ¥28,170; even in the optimistic case, ten-year annualized return is only +5%~+9%, while worst-case valuation normalization could produce a 50%–70% drawdown. The honest conclusion: buying at the current price makes “fivefold in ten years” a low-probability optimistic tail, not a central expectation suitable as a decision anchor.

Jun 4, 2026
Why has the market not recognized all of this yet? Is it too hard to understand, too easy to dismiss, or too far out? What will become the “narrative inflection point”?3/10

Bottom line: unlike Baillie Gifford’s default pattern of “the market is underestimating it,” Advantest has the opposite problem. The market is not failing to understand it or dismissing it; it understands it too clearly and believes it too fully. Advantest is neither obscure nor doubted: current trailing P/E is about 50–51x, EV/EBITDA about 34x, conservative owner earnings about 67–72x, and the analyst camp is highly aligned: among 20 analysts, 16 rate it Buy, 4 Hold, and 0 Sell, with a 12-month average target price of about ¥32,695 (high ¥39,200, low ¥23,000), still implying about 25% upside from the current price. In other words, the market has not only priced in the long AI/HBM testing boom, but continues to look higher from the current price. This is the opposite of “not looking far enough”: it is looking too far and extrapolating too fully.

Therefore, the true “narrative inflection point” is more likely to be triggered to the downside than the upside. Three fuses: ① operating margin falling from FY2025’s 44.2% toward the company’s MTP3 normalized target of 33%–36%. If the market has mistaken a cyclical peak + high-end SoC product-mix tailwind for a new normal, that gap is the starting point for valuation compression; ② AI/HBM capex peaks or slows, so the high-boom profit platform cannot settle into normality; ③ Asian customer concentration (overseas revenue share 97.8%), FX, and export-control risks materialize. If any fuse is lit, valuation could move from about 50x back toward normal equipment-stock multiples. This report judges that the worst case could mean a 50%–70% drawdown, with ten-year annualized return in the neutral case only -2%~+4%.

To fairly state the upside countercase: if AI/HBM demand proves structurally durable, FY2026 high-cycle profits truly settle into a new normal, and SoC share (about 66%) continues to expand, the market will further confirm the “scarce AI test asset” narrative, and the high valuation would have a reason. That is exactly the script the current buy-side consensus is betting on.

Honest judgment: at around ¥28,000, there is no undiscovered bargain here; the more important risk is a downward narrative inflection point.

Jun 4, 2026

Ask about this report

Members can ask about this report; once answered it appears under "Reader Q&A" on this page. You can also highlight a passage in the text to ask about it directly.