Earnings Call Highlights: AXT (AXTI) | FY2026 Q2 — Record $47.6M Revenue, Gross Margin Back to 45%, and InP Becomes the AI Data Center's Most Contested Capacity

2 days ago

7 min read

This quarter, AXT didn't just "grow revenue" — it "shifted gears." Quarterly revenue hit a company record of $47.6M, up 77% QoQ and 164% YoY. Gross margin climbed from 8% a year ago to 45%, swinging from a loss to a quarterly profit of $11.1M (GAAP). There is really only one engine: indium phosphide (InP). InP contributed $30.7M, also an all-time high, all of it driven by optical communications demand from AI data centers. Management raised this year's InP capacity target from "doubling" to "more than tripling," backlog surged past $100M, and within a single week the company signed long-term agreements with Coherent, Lumentum, and another customer. Next quarter (Q3) is guided to "at least $66M." This piece skips valuation and simply breaks down the hard numbers and signals. This article is for technology and industry trend analysis only and does not constitute investment advice.

1. Three Key Takeaways

First, AXT has reached a genuine profitability inflection — and it was pulled there by a single product, InP. A year ago the company was losing $6–7M a quarter with an 8% gross margin; this quarter it posted a 45% gross margin and $11.1M in net income. The difference isn't cost control but product mix — InP's margin structure is inherently far better than gallium arsenide (GaAs) and the raw material JVs, so when all incremental growth lands on InP, margin expansion is "arithmetically inevitable." Management said this plainly in the Q&A. Second, supply, not demand, is now the company's only ceiling. Backlog is already "well over $100M" and still growing even as shipments continue. Management even said "we're actually not taking orders" — capacity is fully booked, so taking orders it can't deliver would only distort the backlog figure. It's an unusual kind of "good problem." Third, this quarter is only the prologue; the real ramp comes later. The Lumentum LTSA announced this week, along with the earlier Coherent agreement and another customer's deal, contributed almost nothing to Q2 and will be recognized gradually from Q3 and Q4 into 2027. Management raised the "double InP capacity this year" goal to "more than triple," and plans to double again in 2027. In last quarter's Earnings Call Highlights: AXT (AXTI) | FY26 Q1 we flagged the starting point of this capacity ramp; this quarter is evidence that it's accelerating.

2. Revenue and Financials

First, the headline numbers (GAAP basis; non-GAAP is nearly identical):

• Total revenue of $47.6M, up +77% QoQ (Q1: $26.9M) and +164% YoY (prior-year quarter: $18.0M). A company record.

• Gross margin of 44.9% (GAAP) / 45.0% (non-GAAP), up about +36.9 percentage points from 8.0% a year ago, and another big step up from 29.6% last quarter.

• Operating income of $10.4M (GAAP) / $11.2M (non-GAAP), an operating margin of about 21.8% — versus a $6.7M operating loss in the prior-year quarter.

• Net income of $11.1M (GAAP) / $11.9M (non-GAAP), versus a $7.0M net loss a year ago — officially back in the black.

• Diluted EPS of $0.17 (GAAP) / $0.19 (non-GAAP), versus a loss per share a year ago.

The product-line breakdown is the key this quarter, because it explains why gross margin jumped the way it did:

• Indium phosphide (InP): $30.7M (about 65% of total revenue), a company record, almost entirely from data center applications.

• Raw Material JV: $10.0M, also a quarterly record.

• Gallium arsenide (GaAs): $6.6M, up QoQ on industrial, robotics, and data center laser applications.

• Germanium: $0.272M.

A note on the balance sheet: the secondary offering completed on April 22 brought in about $632M, lifting cash and investments at quarter-end to $748.8M (vs. just $123M at the end of last quarter). The purpose of this money is clear — it's being bet on capacity expansion.

3. Technology and Business Highlights

InP demand is driven by the optical module migration from 800G to 1.6T. Management was explicit: the industry is jumping to the 800G and 1.6T transceiver generations, and for the lasers and detectors in these generations InP is a "required material," not an "option." The higher the speed, the fewer the substitutes. In 800G vs. 1.6T Optical Modules: A BOM Teardown we showed how this generational jump pushes cost and bottlenecks upstream — and AXT is the first link to benefit once things are "pushed upstream." The longer-term demand driver is CPO/NPO (co-packaged / near-packaged optics). Management singled out near-packaged and co-packaged optics as something that will "keep pulling up" InP demand. One often-overlooked point: the most fragile link in CPO, and the one most holding back volume production, is the laser light source; we explained why in The Most Fragile Link in CPO Is the Laser. And whether the path is an external laser source or a comb laser, InP is the core material — for details on the light-source debate, see Will Comb Lasers Replace DFB Arrays?. Two concrete technical milestones: First, clear progress in volume production capability for 6-inch InP substrates — 6-inch is "exponentially harder" than 3- or 4-inch, and this is a moat that widens the competitive gap. Second, vertical integration on the raw material side: subsidiary Jinmei has begun refining high-purity indium, effectively putting InP's most upstream critical raw material in AXT's own hands. There's also a quieter storyline on the GaAs side: VCSELs. In the Q&A, management confirmed that AXT is already a GaAs VCSEL substrate supplier to two data center / laser companies, and that a U.S. customer is also in talks about a GaAs-based VCSEL scale-up solution. Management's own view, however, is that speed decides the winner: above 200G, InP lasers and detectors have the edge, and VCSELs and InP will "coexist" rather than replace each other.

4. Management Outlook

Q3 guidance came with a floor: revenue of "at least $66M." This $66M consists of orders that "already have export licenses or don't require one," which management said it has "high confidence" in recognizing, with significant upside if more licenses are granted. That implies QoQ growth of about +39%.

• Q3 non-GAAP EPS guidance of $0.30–$0.32; GAAP $0.29–$0.31.

• Q3 operating expenses of about $10.5M (non-GAAP) / $11M (GAAP).

• Estimated Q3 diluted share count of about 66.5M shares.

Looking further out, management offered two big numbers. First, this year's InP capacity plan was raised from "doubling" to "more than tripling," with a year-end InP quarterly revenue target in the ~$60M range. Second, it plans to double again in 2027, with total quarterly revenue exiting 2027 targeted at about $130M. Management repeatedly stressed that "this is a moving target, because demand is running faster than we can expand." A nuance in wording worth flagging: on gross margin, the CFO explicitly "targeted the 40s" and hinted at "working toward the 50s (50%+)," but the tone was conservative, repeating "let us get there first, don't push us." This is a "target," not "guidance," so its certainty should be discounted.

5. Supply Chain and Customer Clues

The most important signal this quarter is hidden in three long-term agreements. In Q2, AXT signed long-term supply agreements (LTSAs) with Coherent and with a customer referred to as Casela, receiving prepayments of $25.4M and $22.3M respectively; this week it announced another agreement with Lumentum. Prepayments are booked as liabilities first and converted to revenue as shipments occur. A willingness to "pay up front and lock in capacity" is itself the strongest signal that the InP shortage has escalated from "fighting for goods" to "fighting for capacity." We called this turning point early in Locking Up the Upstream Wafer: Decoding the 3-Year Coherent–AXT InP Supply Agreement, and this quarter's second and third LTSAs extend that trend. A key shift in geographic mix: China accounted for more than 50% of revenue this quarter. China's InP laser-related revenue more than doubled QoQ. Management's read is that China is building out its own AI supply chain at full speed, and since shipments within China "don't require export licenses," this actually works in AXT's favor. Going forward, management expects the China / non-China revenue split to fall in the 40%–60% range. This also explains why the backlog is so large yet "not all in hand" — many overseas orders are held up by export licenses, which are "becoming more regular and faster" but remain unpredictable. A pivot in 北京同美's listing plan. Subsidiary 北京同美 withdrew its IPO application on China's STAR Market at the end of June and will instead pursue a listing on the Hong Kong Stock Exchange, expected to take about a year. This triggered redemption rights for the PE funds that invested $49M in 2021, but management said those PE investors currently want to stay in rather than redeem, and the company has ample cash to cover any redemptions.

6. Conclusion

This quarter, AXT turned an abstract industry thesis — "the AI optical communications shortage will extend all the way to the most upstream substrates" — into hard numbers on the income statement: revenue more than doubled, gross margin grew more than fivefold, and the company swung to profit. Its position is now simple: demand is not in question, licensing is gradually easing, and the only variable is how fast it can expand its own capacity. For next quarter (Q3 FY2026), we suggest tracking three indicators. First, whether actual revenue clearly beats the $66M floor — the size of the beat reflects the real pace of export license easing. Second, whether gross margin holds at 45% or even moves toward the upper 40s — a thermometer for InP's rising share of the mix. Third, whether year-end InP quarterly revenue reaches the ~$60M range — the checkpoint for the "more than triple" commitment. In one sentence: AXT is no longer a materials story "waiting for CPO to land"; it is already the upstream player in the AI optical supply chain that has been first to turn demand into cash.