AXT Inc (NASDAQ: AXTI), initiation of coverage, Nov 10, 2026
AXT’s price assumes a 54% margin for good; we expect a normal 40%
AXT makes the indium phosphide wafers that sit under AI optical lasers, and demand now exceeds supply. Our earnings forecasts match analysts’, but the price pays for a 53.8% normal-year margin where we model 40%.
- Our rating
- Sell
- 12-month target
- $33.33
- Price, Nov 6, 2026
- $38.88
- Expected return, with dividend
- −14%
By Penmark Team
What AXT is now
A wafer maker whose margins jumped as AI lasers ran short of its crystals
AXT grows and polishes crystal wafers in China: indium phosphide (InP) for optical lasers and gallium arsenide for radio chips. InP brought in $54.3m of $73.4m third-quarter revenue, or 74%.
China’s export controls and AI demand created a shortage. Order backlog passed $500m, and gross margin went from 21.5% to 56.1% in three quarters. The best full year since 2009 was 43.0%.
Source: company calls, Penmark Research.
Non-GAAP gross margin for reported quarters; 4Q26 is our estimate. Source: company releases, Penmark Research estimates.
What you are really buying
The price sits about $8 above our value, almost all of it on three bets
Our base value is $31.05 a share, so the $38.88 price is $7.83 higher. Almost all of that gap rests on three bets the market makes.
It assumes a 54% gross margin in a normal year, wafer prices holding at 0.88 of today’s level, and a certain listing for Tongmei, AXT’s main subsidiary. Together those are worth $7.78 a share.
Each step sets one driver to the market view in our discounted cash flow model. Source: Penmark Research estimates.
Where we disagree: margin
A normal year earns about 40% gross margin, not the 54% the price needs
AXT earned between 12.7% and 43.0% of sales as gross profit from 2009 to 2025, with a median of 29.8%. Peers Photronics and Entegris had a median of 40.1%.
We let the shortage run to 2028, at 60% and 56%, then fade to 40% from 2031. The market view of 46% would add $3.41 a share to our $31.05.
A normal year is the mid-cycle gross margin in our model. Source: company filings, Penmark Research estimates.
Where we disagree: wafer prices
Rivals adding 3 to 10 times capacity will pull wafer prices down from 2029
AXT and Sumitomo supply about 80% of InP today, so scarcity sets the price. JX plans roughly 7 to 10 times more capacity and Sumitomo 3.1 times its 2024 level by 2028.
We cut AXT’s price index from 1.05 to 0.73 by 2031, while the market holds it at 0.88. That gap is worth $3.25 a share, and a further 0.10 drop would cost $2.60.
Price index relative to the 4Q26 price and mix. Source: Penmark Research estimates.
Where we disagree: Tongmei
Tongmei’s funding round costs AXT shareholders about $1 a share
AXT is putting $400m into Tongmei, its main operating subsidiary, alongside up to $545m from outside investors. That cuts AXT’s stake from 85.5% to 73.4%.
If Tongmei does not list within three years, the new investors can redeem at cost plus 8% a year. A past listing attempt was withdrawn, so we put a 40% chance on no listing, worth $1.12 a share.
Other holders are existing minorities and the new investors. Source: company 8-K, Penmark Research.
The odds
Only the good case beats today’s price, and it needs several things at once
Our middle case, 50% likely, doubles capacity in 2027 and fades margin to 40%, for $34.74 in 12 months. The good case, 20% likely, keeps the shortage into 2029 and reaches $61.85.
The bad case, 30% likely, has export permits squeezed and prices falling hard, for $11.98. Weighted together, they give our $33.33 target, 14.3% below the $38.88 price.
- Good, 20%InP stays short into 2029, capacity reaches 2.2 times, US permits are granted, Tongmei lists and margin settles at 46%: $61.85.
- Middle, 50%Capacity doubles in 2027, utilisation eases from 97% to 82%, the price index falls to 0.73 and margin fades to 40%: $34.74.
- Bad, 30%Export permits are squeezed, 2027 utilisation drops to 75%, the price index hits 0.55 and margin is 29.8%: $11.98.
Circle size shows how likely we think each case is. The target weights all three and adds 12 months of expected return. Source: Penmark Research.
What would change our mind
A lower cost of equity is the one change that would lift us to Hold
Only one of our tests changes the rating: a cost of equity 1 point lower gives a $35.84 target, 7.8% below the price, a Hold. The Sell also ends if the price falls below $37.04.
Near term, watch China’s export permits in November and fourth-quarter results in February. Over time, margins still above 50% in 2029 would break our view.
A squeeze on permits would cut 2027 utilisation from 97% to 75%, costing $0.52 a share.
We expect $90.0m of revenue and $0.58 of earnings per share; analysts expect $91.5m and $0.57.
Our base case doubles capacity. Reaching 2.2 times would add $2.27 a share.
JX and Sumitomo start dates. Lead times falling from more than nine months would signal the turn.
A filed Hong Kong prospectus before the deadline would remove the 8% redemption floor.
Inside the full report
What each page covers after the summary on pages 1 and 2. Each opens the full report at that page.
See all 21 pages
- 3Returns turn positive in FY26E and we are within 2% of consensus on EPS; the debate is margin, not volume
- 4Next 12 months hinge on permits and capacity; AXTI is not cheap on history, and only fairly priced against peers
- 5The stock follows gross margin, and the price already pays for a 54% mid-cycle level
- 6Three debates decide the stock, and the price sits on the market side of all three
- 7Quarterly sales leapt from $28m to $73m in a year, and the price assumes the shortage never ends
- 8InP now brings in 74% of revenue and 56 cents of every dollar becomes gross profit
- 9Revenue is capacity times utilisation times price, and price is the input we expect to fall
- 10Two suppliers hold about 80% of InP today, and three rivals are adding 3x to 10x capacity
- 11Management has beaten its last two revenue guides by 11% or more, but insiders sold $29m in June to August
- 12The stock trades on InP margins and optics sentiment, and the market has priced both above our value
- 13Each driver is worth $2 to $4 a share to us; none supports the price on its own
- 14Chapter 1: the price needs a 54% mid-cycle gross margin; we model 40%
- 15Chapter 2: InP pricing falls from FY29, worth $3.25 a share against the market view
- 16Chapter 3: the Tongmei round dilutes AXT to 73.4% and carries an 8% redemption floor
- 17AXT lost money in three of five years; GAAP and non-GAAP EPS differ by only $0.07
- 18Our forecasts match consensus on revenue and EPS and differ on how long the margin lasts
- 19Both valuation methods land at $28 to $31 a share, below the $38.88 price
- 20The $33.33 target implies −14.3%, and only a lower cost of equity moves the rating
- 21Index funds and trading firms hold the stock, and short interest is up 59% since April
- 22The model takes revenue from $88.3m to $717.0m and EBITDA to $360.3m by FY28E
- 23Reference: GAAP to non-GAAP, definitions, peer set and method
Penmark Research. This summary draws on our initiation report on AXT Inc, published Nov 10, 2026. It is not investment advice. Analyst certification, rating definitions and important disclosures are in the full report.