Earnings Call Highlights: AXT (AXTI) | FY2026 Q3 — Revenue Up 162%, 56% Gross Margin, and a $500M+ Backlog: InP Substrates Move from Capacity Grab to Allocation
1. Executive Takeaway: The Three Signals That Matter
First, InP is no longer just short; it is being rationed. Last quarter, in our AXT FY2026 Q2 earnings highlights, we wrote that InP had become a capacity-grab asset. This quarter management went further: every customer is on allocation and the company is significantly overbooked. Even customers with signed long-term agreements are coming back to ask for 2–3x more volume.
Second, the backlog grew roughly 4–5x in 90 days, and long-term agreements are not what inflated it. The $500M+ backlog combines multi-year agreements, contracts of 12 months or less and spot orders. The VP of business development said plainly that LTAs are not a big part of it; the increase is run-rate business. That points to breadth of demand, not a few large customers locking up volume.
Third, the gross-margin engine is shifting from volume to quality. Volume and mix are still the foundation, but three new variables showed up this quarter: productivity from a new generation of crystal-growth furnaces, the migration from 3-inch to 4-inch (and 6-inch) wafers, and the price premium for low etch pit density (EPD) in high-power lasers. None of these are easy for competitors to copy in the near term.
Management's message, paraphrased: we always believed InP would be a big opportunity, but honestly we did not expect it to be this big.
2. Revenue and Financials: Half of Last Year in a Single Quarter
Q3 results (GAAP, consistent with the STT earnings card):
Item | Q3 2026 | Q2 2026 | Q3 2025 | Change |
Revenue | $73.4M | $47.6M | $28.0M | QoQ +54.3%, YoY +162.6% |
Gross margin | 56.0% | 44.9% | 22.3% | YoY +33.7 pp |
Operating income (margin) | $28.4M (38.7%) | $10.4M | -$1.1M | Swung to profit |
Net income (attributable to AXT) | $25.7M | $11.1M | -$1.9M | Swung to profit |
Diluted EPS | $0.39 | $0.17 | -$0.04 | Swung to profit |
On a non-GAAP basis (excluding stock compensation): gross margin 56.1%, operating income $29.5M (40.2% margin), net income $26.8M and EPS $0.41.
Revenue by product line:
Indium phosphide (InP): $54.3M, about 74% of revenue, mainly from data center applications.
Consolidated raw-material companies: $9.7M.
Gallium arsenide (GaAs): $9.1M.
Germanium (Ge): $344K, now negligible.
A few numbers worth a second look:
Incremental gross margin of roughly 76%: revenue rose $25.8M and gross profit rose $19.7M. Scale alone does not explain that; price and mix are both moving up.
Cash and investments of $825.9M, up $77M in the quarter. A large part came from a $43.5M wafer prepayment under the Lumentum agreement, booked as a liability and recognized as revenue as wafers ship.
Receivables rose only $3M and DSO fell to 49 days thanks to multiple customer prepayments. Customers paying up front is the most honest gauge of supply and demand, the same playbook we saw in Corning FY2026 Q3, where customers prepay to fund capacity.
Customer concentration: the top five customers were about 35% of revenue and only one exceeded 10%. For an upstream materials supplier in a demand surge, that is a healthy spread.
▲ AXT Q3 2026 key financials: revenue $73.4M (+162.6% YoY, +54.3% QoQ), GAAP gross margin 56.0%, operating margin 38.7%, net income $25.7M, EPS $0.39, Q4 realizable revenue of about $90M; InP segment revenue of $54.3M leads the mix (Image source: Simple Tech Trend original; data from SEC 8-K earnings release, 2026-10-29, and the earnings call)
3. Business Highlights: Capacity, Furnaces and Wafer Size Moving Together
Capacity: Promised to Double, Done by August
AXT committed at the start of the year to double InP capacity, beat that goal in August and kept going. Taking end-2025 as 1, end-2026 capacity should land at 2.75–3x; the 2027 goal is to double again, and the company is already designing the conversion of a newly acquired adjacent building, with work expected to begin within months.
One remark from the VP of business development is worth remembering: even with capacity doubling again in 2027, he is not sure AXT will get its head above water. In other words, capacity is still growing more slowly than demand.
New Crystal-Growth Furnaces: The Hidden Margin Variable
AXT developed a proprietary new generation of crystal-growth furnaces, originally for 6-inch InP, that also makes the 3-inch to 4-inch transition more efficient. New furnaces will use this design going forward. This is the most underrated message of the quarter: expansion is not just buying more furnaces; each furnace is also producing more.
Wafer Size: 2027 Is the Year of 4-Inch
2026: lasers mostly on 3-inch, photodetectors mostly on 4-inch.
2027: laser material moves to 4-inch in volume; by year-end, by wafer count, 3-inch and 4-inch could be roughly equal.
6-inch: now at pilot-production-line status, still a single-digit share in 2027 but margin-accretive.
This lines up exactly with what we saw in Luxnet's 2026 Q2 earnings highlights, where InP moves from 2-inch to 4-inch: substrate makers and laser makers are switching wafer sizes in step, and 4-inch is the common language of the InP supply chain in 2027.
Low EPD: High-Power CW Lasers Raise the Bar
When AXT first qualified at a major US laser maker, the EPD requirement was below 500; AXT now delivers less than one-tenth of that. Management said a major customer is asking for even tighter EPD for high-power lasers, and that the higher the bar, the fewer the competitors and the bigger the price premium.
This is the other side of what we covered in LandMark Optoelectronics' 2026 Q3 earnings highlights, where the same wafer yields 30–50% fewer lasers: high-power CW laser dies for scale-up are larger and their yield depends heavily on substrate defect density, so upstream substrate quality directly sets how many good lasers come off each wafer.
GaAs and Raw Materials: The Supporting Cast
GaAs: demand from wireless, industrial robotics and data center lasers grew sequentially, with a seasonal dip expected in Q4. Notably, customers are showcasing slow-and-wide VCSELs on 6-inch and 8-inch GaAs, where AXT sees itself as the leader. For context, see Lumentum's Slow and Wide VCSEL talk at OCP Global Summit 2026.
Raw materials: subsidiary Jinmei is now refining high-purity indium, pulling another critical input in-house, and AXT is investing to expand its capacity.
Tongmei Private Placement and Hong Kong Listing
Majority-owned subsidiary Tongmei signed a definitive private equity financing agreement, selling its own shares (not AXT shares) to bring RMB directly into its China operations. At the maximum $545M, AXT would keep control with roughly 73.4% ownership. Tongmei continues to work toward a listing on the Hong Kong Stock Exchange.
4. Outlook: Q4 Guidance and How Management Worded It
Q4 guidance:
Revenue: about $90M, explicitly defined as revenue that already has an export permit or does not need one, implying roughly 23% QoQ growth.
Operating expenses: about $13M non-GAAP and $14M GAAP.
EPS: GAAP $0.56–0.58 and non-GAAP $0.57–0.59 on about 66.1M shares.
All of the increase is InP: the VP of business development said the Q4 growth should all be added to InP.
Reading the language:
On revenue, management cited a high degree of confidence and tied it to permits in hand, which is confirmation-level language rather than expectation. Put the other way, $90M is what AXT can ship, not what customers want.
On gross margin, the CFO first said he does not think it has peaked, then that it could begin with a six next quarter, though not a high six. That is hedged expectation-level language; when pressed on whether 60% is the new baseline, he pulled back, saying things are moving too fast to predict.
On overcapacity, the CFO said 2027 and 2028 capex will be less than needed rather than too much, with 2029 to be reviewed later. That is target-level language with a clear timeline.
One detail worth noting: AXT is deliberately not signing many long-term agreements. Management wants to protect pricing and keep flexibility. When supply is short, that is the right call; it also means AXT has a thinner pricing cushion than peers if the cycle turns.
5. Supply Chain and Customer Clues
Lumentum agreement: signed in Q3 with a $43.5M prepayment; management clarified it is the previously announced deal, not a second one. Together with the Coherent–AXT three-year InP supply agreement we analyzed earlier, both leading US laser makers have now locked up AXT wafers with long-term agreements.
US hyperscale data centers: demand for EML and silicon photonics transceivers and high-speed photodetectors is strong, CW laser demand for scale-up is taking shape, and management believes its material is already deployed across multiple US hyperscalers.
China: InP laser revenue rose sharply again, every Chinese customer is on allocation and lead times exceed nine months. Shipments within China need no export license, a structural advantage for AXT.
Competition: 6-inch InP is a two-player race, with AXT roughly level with its main rival and a distant third. New Chinese entrants want into InP substrates, but management sees no threat for two years and says three years out is hard to call.
The ceiling on the demand story: a major customer told AXT optical interconnect demand will grow not 20x but 1,000x, with scale-across, scale-out, scale-in and scale-up each contributing close to 10x. That is a customer claim and management admitted it did not know how the math works, so it should not be read as a forecast.
6. Summary
Three metrics to track:
Whether Q4 gross margin clears 60%: the first test of whether the new furnaces, the 4-inch migration and the low-EPD premium deliver at the same time. If it stalls at 56–58%, raw-material costs and new-line ramp drag are bigger than management suggested.
InP share of revenue and 4-inch mix: Q4 growth is all InP; in 2027, watch whether 4-inch wafer counts catch up with 3-inch. It is also a leading indicator for expansion at downstream Taiwanese epi and laser makers.
The gap between export permits and backlog: $90M is what AXT can ship, $500M+ is what customers want. If that gap keeps widening, the bottleneck is no longer the furnaces but the permits.
Bottom line: this quarter shows InP substrates have moved from whether supply exists to who gets it first, and pricing power is flowing back to substrate makers. What caps the upside now is not demand but the pace of capacity ramp, export permits and the supply-demand balance after 2029.
This article is for technology and industry trend analysis only and does not constitute investment advice.
Comments