A record quarter, a third capex hike, and a quiet admission that the old profitability model no longer applies.
ASX · Earnings Call · 2026-10-29
The ceiling became the floor
ASE spent years teaching investors to anchor ATM gross margin to a "structural range" in the mid-to-high 20s. In Q3 2026, ATM gross margin hit 30.6%, up 8 points year-over-year, and CFO Joseph Tung did something unusually direct: he retired the framework.
I think the so-called structural margin no longer applies because of the significant and rapidly changing business structures that we are having today... the so-called structural margin ceiling has now become the base for us.
That is a bigger statement than the headline beat. Consolidated revenue was a record TWD 244.2 billion, up 28% sequentially and 45% year-over-year; operating margin reached 13.6%, up 5.8 points annually; net income rose 148% year-over-year to TWD 27 billion. But the real change is conceptual. Management is arguing that mix itself — the LEAP business — has structurally lifted the earning base, and that sequential improvement should continue "from the base of 30%." The prior calls show how deliberate the old anchor was. In October 2025 Tung said they were “very confident about coming back to our structural margin range in 2026” — Joseph Tung, Chief Financial Officer · 2025-10-30, and in July 2025 he was still “keeping our structural margin range as it is for now” — Joseph Tung, CFO · 2025-07-31. Now the structural margin ceiling is explicitly the floor. Tung also described a pricing environment that is "very pricing-friendly" — a noticeably warmer phrase than the "resilient" pricing language of a year ago, even as he declined to quantify it.
Four billion, then double — and a new CPO number
The demand side explains the confidence. Tung raised 2026 LEAP revenue to more than USD 4 billion from USD 3.7 billion, and “for 2027, we're expecting to remain as we last projected to double our LEAP revenue for the year” — Joseph Tung, Chief Financial Officer · 2026-10-29. Full-process packaging — ASE's CoWoS-like work — is on track for USD 300 million this year and more than USD 1 billion in 2027. Then came the genuinely new disclosure: co-packaged optics. “We're expecting to generate more than USD 0.5 billion worth of revenue coming from CPO for next year” — Joseph Tung, Chief Financial Officer · 2026-10-29, with mass production already starting in Q4. That is the first hard number ASE has put on co-packaged optics, a theme it has name-dropped since 2024 without sizing. It matters because the global tape has already cooled on exactly that cluster: co-packaged optics was a top-360-day advancer but a 90-day decliner, with every tracked name down, and AI data center shows the same shape — strongly positive over a year, negative over the last 90 days. ASE is accelerating into a de-rating tape. The binding constraint is not demand. “The biggest bottleneck today is really still the space” — Joseph Tung, Chief Financial Officer · 2026-10-29, Tung said, with 22 building projects running in Taiwan and suitable fabs hard to buy because "competition is pretty keen."
The capex question nobody can dodge
ASE raised 2026 capex to USD 11.5 billion from USD 10.5 billion — its third upward revision this year — and guided 2027 higher again, with negative free cash flow in the interim. “For next year, we will continue to see negative cash flow” — Joseph Tung, Chief Financial Officer · 2026-10-29, Tung said, though he expects it to "be balanced off in due course" via multiple funding sources including customer trade financing. The balance sheet already shows the strain: total interest-bearing debt rose TWD 100.2 billion to TWD 406.5 billion, and net debt-to-equity hit 66%, which management called the near-term peak. This is the same company that told investors in 2025 it would "not be shy" about investing; the difference is the scale and the repeated upward revisions. Meanwhile the AI build-out is cutting both ways. In EMS, shortages driven by that same build-out are inflating pass-through costs, lifting reported revenue while compressing margin — EMS gross margin slipped 0.6 points sequentially to 8.3% and operating margin fell 0.4 points annually to 3.3%. The component cost dynamic is a real, if modest, drag on the consolidated picture, and Q4 guidance has EMS flattish with operating margin around 2%.
What got quieter: the test narrative and the non-AI recovery
For several years, test was ASE's favored growth story — the company repeatedly said test would outgrow packaging. This quarter the emphasis shifted. Assembly is now growing faster than expected, and Tung said capital allocation will favor it: “The space that we use for assembly is more efficient than test at this point. So given the constraint, we will allocate... our priority on assembly and for test will be for wafer sort.” — Joseph Tung, Chief Financial Officer · 2026-10-29The wafer sort priority is a rational response to a facility bottleneck, but it is a quieter test narrative than the one investors heard in 2025. The other underappreciated thread is non-AI. Tung expects “more than 20% growth in our general market business as well” — Joseph Tung, Chief Financial Officer · 2026-10-29, with automotive and industrial recovery plus AI-adjacent chips more than compensating for soft PCs and phones. The general market is doing real work in the model again — which makes ASE less of a pure AI derivative than its LEAP headline suggests. Put together, the story is a company voluntarily demolishing its own margin ceiling, funding the next leg with debt and negative free cash flow, and doing so while the tape that should reward it has spent the last 90 days fading. That is the tension worth watching into 2027.