AXT INC (AXTI): what the price assumes

In the published model solve dated 2026-Q2, anchored at $38.26, AXT INC (AXTI) is priced for today's economics sustained for ~38.7 years. boothcheck publishes no house fair value, target price, or buy/sell rating; individual model outputs and user-controlled scenarios are analytical inputs, not Boothcheck targets. Narrative composed 2026-07-25.

Generated: 2026-07-29 · Exported: 2026-08-01 · Source: https://boothcheck.com/report/AXTI

Headline

FieldValue
TickerAXTI
CompanyAXT INC
Sector / IndustryTechnology
Current price$38.26/sh
CompositionSubstrates 67% / Raw materials and others 33%

What The Price Assumes (Inversion)

The assumption today's price embeds, recovered by inverting the valuation.

FieldValue
Inversion basiswhole-company
Trailing margin (depressed year)-13.9%
Must persist for38.7y
Multiple paid483x mid-cycle operating income

Solve inputs: computed at a 12.5% cost of capital; growth searched up to the 25% self-funding ceiling; each 1pp moves the implied horizon ~4.3 years.

Reconcile: at the x-ray's 9.3% required return this reads ~26.7 years; the models below use their own rates.

How unusual the bet is: elevated

ReferenceValue
vs own history+0.57σ
sustained it ~10 years at this level15%
implied end-window share0%

Valuation X-Ray

Every valuation family lands below the price. The price therefore sits beyond what those standard frames encode.

How the valuation models price the stock relative to the market price. Price/FV above 1.0 means the market pays more than that lens defends (expensive); at or below 1.0 the lens can defend the price.

FamilyMedian price/FVModelsReads
Asset9.62x2expensive
Earnings—0—
Relative5.22x2expensive
Growth8.70x2expensive

Families that call it expensive: Asset, Relative, Growth

The models below discount at their own flat-beta convention rates (cost of equity 9.3%, WACC 9.1%); the inversion above states its own rate.

Per-Model Detail (n=6)

ModelFamilyFVPrice/FVApplicableMethodology
DCF Perpetual GrowthGrowth$0.00—noNegative/zero FCF — equity value floored at $0
DCF Exit MultipleGrowth$0.00—noNegative/zero FCF or EBITDA — equity value floored at $0
Relative ValuationRelative$7.335.22xyesP/S fallback (negative EPS): Sector P/S 5.0x × TTM revenue — excluded from consensus
Simple DDMGrowth——no—
Two-Stage DDMGrowth——no—
Simple Excess ReturnAsset$4.209.11xyesBook value floor: BV/sh $4.20, ROE negative
Two-Stage Excess ReturnAsset$3.7810.12xyesBook value with convergence: BV/sh $4.20, ROE converges to ke
Discounted Future Market CapGrowth$10.803.54xyesRev $0.1B, growth 3% (input: historical growth; tapered), Terminal P/S: 10.5x / 15.0x / 19.5x (bear / base = today's held flat / bull, cap 15x)
Peter Lynch Fair ValueRelative$0.00—noNegative/zero EPS — earnings-based value floored at $0
Margin TrajectoryGrowth$2.7613.86xyesMargin ramp: -15% → 25% over 7yr, rev growth 3% (input: historical growth; tapered)
Earnings Power ValueEarnings——no—
Residual IncomeAsset——no—
Graham NumberAsset——no—
EV/EBITDA RelativeRelative——no—
FCF YieldEarnings——no—
SBC-Adj FCF YieldEarnings——no—
Ben Graham FormulaEarnings——no—
ROIC-Justified P/BAsset——no—
P/Sales SectorRelative$7.335.22xyesRevenue $0.10B × sector P/S 5.0x
PEG Fair ValueRelative——no—
Earnings YieldEarnings——no—
Funds From Operations MultipleRelative——no—
Clinical Phase NPVGrowth——no—
MertonAsset——no—
V5 Mechanical———no—

Solvency

FieldValue
Net cash$49.7m
Net debt / NOPAT (after-tax)-12.07x (net cash)
Net debt / operating income (pre-tax)-9.53x (net cash)
Share count CAGR (dilution)5.7%
Burning cashyes

Leverage and coverage are computed on normalized mid-cycle operating income (mid-cycle margin 5.4%); the trailing year was depressed.

Interest expense is not separately reported in the latest filings, so interest coverage cannot be computed.

Bullet Takeaways

Bull Case

The odd thing about this company is where the bottleneck sits. Its product feeds one of the least price-sensitive construction booms in industrial history, and yet what governs how much of it reaches customers is a licensing desk in Beijing. The 10-K records that AXT received its initial export permits from the Ministry of Commerce of the People's Republic of China to resume shipping indium phosphide substrates to certain customers in Europe and Japan, while adding that it cannot predict when a permit application will be reviewed. A bull reads that as the good kind of problem. Administrative bottlenecks clear. Demand bottlenecks do not.

What the permits gate is a genuinely difficult material. Indium phosphide is the crystal underneath the lasers and photodetectors that push data through fibre inside a data centre, and the company's description of its own position is unusually blunt: InP substrates have the lowest defect densities and lowest stress and slip lines on the market and are best in class. The stickiness comes one layer up. A majority of our customers specialize in epitaxial growth, a complex series of chemical layers grown on top of our wafers. Our wafers cannot be used to make chips until the epitaxial layers are grown, which means switching substrate suppliers is not a purchasing decision but a re-qualification project. Slow to win, slow to lose.

The income statement has begun to show what volume does to a business built out of furnaces. Gross margin, the 10-K records, ran 6.4% but improved to 8.0% in the second quarter of 2025; in the third quarter of 2025, our gross margin has further improved to 22.3%. Crystal-growing capacity costs roughly the same whether it runs full or half empty, so incremental wafers arrive with most of their price attached. Through the cycle, on its own normalized margins near 5.4%, the operation earns rather than loses. The trough is a volume condition, not a structural one.

Management has responded by building well ahead of the order book. The April 2026 equity offering, roughly $632.5 million gross with the overallotment exercised, is earmarked chiefly for the Beijing subsidiary to expand indium phosphide output for export worldwide, and the stated plan doubles that capacity in 2026 and doubles it again in 2027. A company that spent the last downturn rationing capital now has the means to be early instead of late.

That is the whole bull case, and it is not subtle. The qualification work is done, the material is hard to make, the customers cannot swap it out quickly, and the end market is being built by buyers who care more about schedule than about wafer cost. What is missing is throughput, and throughput is the one thing money can buy.

Bear Case

Begin with who is paying for the growth. It is not the business. AXT's share count has climbed about 5.7% a year over the four years to March 2026, and in April 2026 the company sold 8,560,311 more shares on top of that base. The balance sheet carried in this record predates the sale: as of the March 2026 quarter it showed net cash of $49.7M measured on funded borrowings, with liquid assets of $123.7M against gross debt of $74.0M. Count lease obligations as well and the same balance sheet tips slightly the other way, into a small net liability. Either measure was a thin cushion for the plan being described, which is exactly why the equity was sold. Issuing shares is the one funding source whose cost never appears on the income statement, and it is the most expensive money on offer here.

The operating line explains the need. Over the trailing year the operating line lost 19.7 million dollars, and the business consumed cash rather than producing it. That is not unusual for a substrate maker between cycles. It does mean the expansion, the working capital behind it, and the losses in the meantime all draw on the same pool, and that pool is refilled by selling equity.

The assets themselves sit inside a jurisdiction that has already shown it will treat them as leverage. Export permits govern which indium phosphide leaves China and for whom, and the company can offer no timetable. Raw materials carry the same exposure from the other direction: the 10-K describes lower production and shipments of gallium as a result of reduced services provided by the aluminum plant, which forces sourcing from third parties at higher costs and reduced gross margin. A supply chain that runs through state permitting is one a large customer eventually designs around, and it is one no shareholder can forecast.

There is also an ownership question that has been open for four and a half years. Tongmei submitted its IPO application to the Shanghai Stock Exchange in December 2021 and it was formally accepted for review on January 10, 2022. The Shanghai Stock Exchange approved the IPO application on July 12, 2022, and the listing remains subject to review. Minority investors already sit inside the subsidiary that holds the crystal-growing assets. A holder of the parent owns a claim that is shared, on operations governed by a process nobody here controls.

Then the price. Every standard family of method lands below today's quote, and not by a little: the asset-value lens sits near a tenth of it, and a sector revenue multiple, the most generous frame available when there are no earnings, still reaches under a fifth. Run the price backwards into a required operating-profit path and it does not resolve at all; push the required pace to the ceiling a company can fund out of its own earnings, hold it out four decades, and the arithmetic still falls short. Roughly a third of the fast-growing companies historically priced at this level sustained the pace even five years. The bull answer is that substrate supply for AI optics is genuinely scarce and this is one of very few qualified sources, which is true. The bear point is narrower: a business this small, this cyclical, and this exposed to one government's export desk has to execute close to perfectly for a long time to grow into what is already being paid, and it has to do it while issuing shares.

Valuation

Start with the size mismatch, because everything else follows from it. Roughly $100M of trailing revenue carries a market value of about $2.5B. The operating line lost 19.7 million dollars across that same trailing year. Neither fact is by itself an indictment; plenty of businesses have gone on to earn a ratio like that. What both facts mean together is that nothing in the trailing accounts reaches what the shares now fetch.

Inverting the price into a required assumption does not resolve here, and saying so is more useful than manufacturing a number. Push the required operating-profit pace up to the ceiling a business can fund out of its own earnings, hold it out four decades, and the price is still not reached. What comes back is a boundary rather than a measurement: the price sits outside the range an operating-profit path can describe, so whatever is being paid for is being paid on grounds this arithmetic does not encode.

The families of method agree with each other, which is itself unusual. The asset-value lens, which is book value of $5.16 a share adjusted for returns, sits near a tenth of the quote. Peer multiples, applied to revenue because there are no earnings to apply them to, travel furthest and still land under a fifth of it. The forward-growth methods, the ones built specifically to credit a company for what it has not yet done, come out no better. When no family reaches the price, the operative conclusion is not that the stock is expensive by a measurable quantity. It is that the standard frames have stopped being the ones setting the price, and a demand narrative is doing that work instead.

The comparison group is the broader compound-semiconductor and analog set the filer sits beside rather than a clean pure-play match, and it is useful on one axis. LSCC turned over $574.0M of revenue at a 5.3% operating margin. POWI did $446.3M at 1.1%. MXL ran $508.9M with an operating margin of -19.3%, and SITM $379.9M at -13.5%. Losing money at this point in the cycle is ordinary in this group. Being a fraction of their size while carrying this valuation is not.

The balance sheet deserves its date stamp. As of the March 2026 quarter it held net cash of $49.7M on a funded-borrowings basis, liquid assets of $123.7M and gross debt of $74.0M; on a lease-inclusive basis the same quarter reads as a small net liability instead. Then in April the company sold 8,560,311 shares for roughly $632.5 million gross including the overallotment. Any liquidity read that stops at the March quarter understates the cash on hand and understates the dilution by the same event.

Which makes the share count the number worth following. It was already climbing about 5.7% a year through March 2026 before the April sale, and the published plan doubles capacity twice more. The funding for that is coming from issuance rather than from operations, and the per-share claim on whatever the substrate market delivers narrows each time it does.

Catalysts

Second-quarter results are scheduled for after the close on July 30, 2026. Management guided second-quarter GAAP earnings per share to a range of $0.05 to $0.07. Set against a trailing operating loss, a quarter printed in the black would read as an inflection rather than a beat, and the market has already been told to expect one.

Capacity is the medium-term file. Proceeds from the April offering are earmarked chiefly for the Beijing subsidiary to raise indium phosphide output for export worldwide, and the published plan doubles that capacity during 2026 and doubles it again during 2027. Capacity added into a market that stays tight becomes revenue. Capacity added into one that loosens becomes depreciation, and the decision has already been made.

Two items carry no schedule at all. Export approvals from China's Ministry of Commerce determine which indium phosphide leaves the country and for which customers, and the company has said it cannot forecast the review timetable. Separately, the Shanghai listing of the operating subsidiary has sat under review since 2022. Neither has a date attached, and either one landing would matter more to the shares than a single quarter's print.

Peer Cohorts (Per Segment, With Filing Citations)

AXT Inc (consolidated) (reported)

Methodology Note

Fundamentals sourced from SEC EDGAR filings. Current price from Databento. The priced-in inversion and valuation x-ray are computed by the boothcheck engine; narrative composed by AI from the structured data.

Sources

AXT public offering pricing announcement, April 2026; AXT earnings advisory, July 2, 2026 · AXT public offering announcement and first-quarter 2026 expansion plan, April 2026 · AXT public offering pricing announcement, April 2026 · AXT earnings advisory, July 2, 2026 · AXT first-quarter 2026 results release, April 2026

View the full interactive AXTI report on boothcheck