US$11.5 Billion to Build: Who Gets Paid While ASE's New Floors Catch Up

ASE had the best quarter in its history, and it is spending faster than it earns. On 29 October the world’s largest chip packaging and test house said third-quarter sales rose 45% to NT$244 billion and its packaging and test margin passed 30%. On the same day it raised its 2026 spending plan again, to US$11.5 billion, after borrowing about NT$101 billion in three months to keep building. The money is not the problem, and neither is demand. The stuck step is the new floor: the buildings and tool installs that turn spending into working lines.

TAIWAN PULSE · SEMICONDUCTORS · Data as at 29 Oct 2026 · Next update: ASE monthly sales, early December
Affects: Semiconductors · Hyperscalers — Markets: Taiwan · Korea · US

Education only, not advice to buy or sell any security. Figures are as published by ASE in its 29 October release, or as said on its call and reported by Central News Agency and Commercial Times, unless another source is named. Our own sums are labelled.

The numbers that matter this week

Figures from ASE’s unaudited Q3 2026 release (29 Oct). Capex plan from the 29 Oct call via CNA; building count from the July call via BusinessNext. y/y = from a year earlier; q/q = from the quarter before. Status chips are Second Order’s read.

On this page: Why it matters · The quarter · Why the margin rose · US$11.5 billion · Who pays for it · Testers, not wirebonders · LEAP and next year · Our math · Who gets paid · What is stuck? · Dates next · What would change this · FAQ

Why does ASE’s quarter matter to who gets paid?

Because every AI chip has to be packaged and tested before it can go into a server, and ASE does more of that work than anyone else. A finished chip die is useless until it is joined to its memory and its base, wired up, sealed and checked. TSMC does the most advanced joining for the biggest AI chips itself (TSMC’s CoWoS Pulse), but a growing share of the steps around it, and most of the testing, goes to outside houses like ASE (who tests AI chips).

So ASE’s books are a good window into the middle of the chain. If the AI build-out were slowing, you would see it here: fewer orders, emptier lines, softer prices. This quarter shows the opposite. Lines are full, prices are firm and the company is building as fast as it can. The question for this Pulse is a narrower one: when a company is this busy, who gets paid first, and what step decides how fast the rest follows?

The answer from the 29 October numbers is that the people who sell ASE its tools and build its plants are paid first. ASE itself is paid when new lines start. And the thing that sets the pace is not money or orders but floor space that is built, fitted and running.

What did ASE report on 29 October?

ASE Technology Holding has three parts. Packaging joins and seals chips. Testing checks that they work. Together these form what ASE calls ATM: assembly, testing and materials. The third part is EMS, electronics manufacturing, run through its USI unit, which builds finished modules and boards, much of it for phone and consumer brands.

Group revenue in the third quarter was NT$244.2 billion, up 44.8% from a year earlier and 27.8% from the second quarter. In July, ASE had guided for growth of 21% to 22% from the second quarter, so it beat its own forecast by a wide margin. Net profit for shareholders was NT$27.0 billion, up 148% from NT$10.9 billion a year earlier. Earnings per share were NT$6.13.

Packaging and test (ATM) revenue was NT$151.5 billion, up 51% from a year earlier and 20.1% from the second quarter. ASE had guided that arm to grow only 11% to 13% from the second quarter. Within the group, packaging was about 50% of revenue, testing 11%, EMS 38% and other items 1%.

EMS revenue was NT$93.5 billion, up 35.4% from a year earlier and 42.1% from the second quarter. EMS swings with the yearly consumer device cycle. But EMS is a thin business: its gross margin was 8.3% and its operating margin 3.3%. Of the group’s NT$33.1 billion of operating profit, about 91% came from packaging and test and about 9% from EMS (our arithmetic). For understanding who gets paid in AI, the ATM arm is the one to watch.

One more line shows where the growth comes from. ASE sorts its packaging and test revenue by the kind of product the chips go into. Computing, which includes data centre chips, was 34% of ATM revenue in the third quarter, up from 30% in the second quarter and 25% a year earlier. On those shares, ATM revenue from computing roughly doubled in a year, from about NT$25 billion to about NT$52 billion (our arithmetic). Communication, mostly phone chips, fell from 45% to 38%.

Why did the packaging and test margin pass 30%?

Gross margin is the share of each sale left after the direct costs of making it. ASE’s packaging and test gross margin was 30.6% in the third quarter, up from 27.3% in the second quarter. In July ASE had guided for 28% to 29%. Operating margin for the arm, after research and overheads, was 20.0%, up from 15.7%. ATM operating profit was NT$30.2 billion, up 53% from the second quarter and about 2.8 times a year earlier (our arithmetic).

ASE used to describe about 30% as the structural ceiling for its packaging and test margin. On the call, CFO Joseph Tung said the old structural ceiling had now become the base, and that margin could keep improving quarter by quarter from that 30% level as capacity is added next year. ASE told Commercial Times that 30.6% was not the limit, helped by a favourable pricing environment and more high-value products.

Three things are doing the work. First, mix: more of the work is advanced packaging and testing for AI and computing chips, which pays better than older wire-based packaging. Second, price: when capacity is short, a packaging house can pass on more of its costs and charge more for scarce slots. Third, fullness: when lines are busy, the fixed cost of each machine is spread over more chips. All three depend on demand staying ahead of ASE’s capacity. That is why the capacity plan matters so much.

For the fourth quarter, Tung guided for group revenue up 6% to 8% from the third quarter, a group gross margin of 23% to 24% and an operating margin of 14.5% to 15.5%, assuming NT$31.9 to the US dollar. Packaging and test revenue is guided up 11% to 13% with its gross margin up another one to two percentage points. EMS is guided flat with an operating margin of about 2%. On those ranges, fourth-quarter group revenue would be about NT$259 billion to NT$264 billion and ATM revenue about NT$168 billion to NT$171 billion (our arithmetic), both new records.

What is the US$11.5 billion for?

ASE’s 2026 spending plan has gone up again and again. By the July call it had reached about US$10.5 billion, after ASE added US$1 billion each to buildings and to equipment. On 29 October Tung raised it to US$11.5 billion. He said about half of the extra US$1 billion would go to facilities to keep preparing new plants and about half to equipment. Of the equipment money, about 60% is for advanced packaging and test projects and about 40% for mainstream products.

The third quarter shows the pace. ASE spent US$2.1 billion on equipment alone in the quarter: US$991 million in packaging, US$1,067 million in testing, US$41 million in EMS and US$5 million elsewhere. A year earlier, equipment spending for packaging and test plus EMS was about US$776 million (our arithmetic from its tables), so the quarterly rate has nearly tripled.

Equipment is only part of it. In cash terms, ASE paid NT$125.9 billion for property, plant and equipment in the quarter, about US$3.9 billion at the quarter’s average rate (our arithmetic). The rest of that cash went into buildings, clean rooms and the plant that holds the tools. On the balance sheet, property, plant and equipment rose from NT$517.7 billion to NT$622.0 billion in three months, up about 20% (our arithmetic).

On the July call ASE said it was running 13 new building projects plus eight purchases and conversions of existing buildings at once. It said these plants would support it to 2028, with some reaching into 2029. COO Tien Wu put the risk plainly then: demand was not what he was worried about; the uncertainty was ASE’s own ability to build plants, install tools and ramp output on time.

ASE is not alone. China’s JCET lifted its own 2026 spending plan by 65% this year (the JCET Catalyst), and the race for packaging floor has pulled Taiwan’s suppliers into new land in Chiayi and new parks in the US (Chiayi packaging land; Taiwan’s US parks).

Who is paying for the build?

Not the quarter’s profits alone. ASE’s operations brought in NT$50.5 billion of cash in the third quarter. It paid NT$125.9 billion for plant and equipment. That leaves a gap of about NT$75.4 billion for the quarter, and about NT$115.9 billion for the first nine months (our arithmetic). It also paid NT$29.0 billion of dividends in the quarter.

Borrowing filled the hole. Net new borrowing was NT$100.9 billion in the third quarter, up from NT$39.9 billion in the second. Long-term bank loans rose from NT$209.6 billion to NT$276.6 billion, and bonds from NT$15.0 billion to NT$42.2 billion. Net interest expense was NT$2.6 billion, up about 83% from a year earlier (our arithmetic). The net debt to equity ratio, a simple gauge of how much a company leans on loans, rose from 0.47 to 0.66 in one quarter.

This was flagged. In July ASE said its heavy spending would keep free cash flow negative for some time, and that its balance sheet was healthy with several ways to fund the build. It still had NT$457.3 billion of unused credit lines at the end of September, and cash of NT$88.4 billion. Nobody is short of money here. The point is the order of payment: lenders and suppliers are paid on a fixed schedule, while ASE’s return arrives only when the new lines run.

There is a second, quieter cash drain. Trade receivables, money customers owe ASE, rose from NT$143.3 billion to NT$174.5 billion in the quarter, and inventories from NT$86.3 billion to NT$99.7 billion. Fast growth ties up cash before it comes back. We saw the same pattern at Wiwynn, where sales doubled but cash went out (the Wiwynn Analysis), and at Meta, where spending outran cash (the Meta Pulse).

One more clock is built into this. Spending on plant hits the profit line later, as depreciation, spread over the years a tool or building is used (capex vs depreciation). ASE’s depreciation and amortisation was NT$22.3 billion in the quarter, up from NT$17.0 billion a year earlier. The bigger the build, the higher that bill climbs. ASE’s bet is that better prices and fuller, more advanced lines keep the margin rising anyway.

Why are testers growing while wirebonders stand still?

ASE reports how many of its two main machine types it runs. Wirebonders are the older workhorses of packaging: they join a chip to its base with fine gold or copper wires. Testers check that a finished chip works. The counts tell a clear story.

ASE added 1,765 testers in a year and 483 in the last quarter alone (our arithmetic), while its wirebonder count went slightly backwards. In the third quarter it spent more on test equipment than on packaging equipment. Testing revenue was NT$26.3 billion, up about 43% from a year earlier (our arithmetic).

This fits what is happening to AI chips. A big AI processor with stacked memory around it is expensive to make and expensive to throw away, so it is tested longer, at more stages and at higher power than a phone chip. That eats tester hours. This week, reports said TSMC was weighing higher prices for its advanced packaging as AI chip tests run longer (the CoWoS price Wire), and Nvidia has paid to secure test capacity directly (the KYEC Wire).

In plain English: the new floor at ASE is filling with test machines, and test time is one of the scarcest things in the AI chip chain. Whoever makes those testers, the probe cards and the handlers that feed them is paid as the floor fills.

What did ASE say about LEAP and next year?

LEAP is ASE’s name for its leading-edge advanced packaging and test work, the part that serves AI and high-end computing chips. Tung raised the 2026 LEAP revenue view from US$3.7 billion to US$4.0 billion. The target for 2027 is to double LEAP revenue, which would mean about US$8 billion on the new base (our arithmetic). Its earlier target was US$3.5 billion.

ASE also gave targets for two newer lines. Its full-process advanced packaging, where it does the whole job end to end rather than only the back steps, is on track for about US$300 million this year and should pass US$1 billion next year, with yields at mass-production levels, Tung said. Co-packaged optics, which places light-based links right beside the chip to move data with less power, should bring more than US$500 million next year. For background on how chips are stacked and joined, see SoIC vs CoWoS and what SoIC is.

Commercial Times reported that ASE said advanced packaging capacity is still short of demand, that the 2027 advanced business should keep doubling, and that next year’s capital spending could rise further. Each of those targets needs the same thing: more floor, built and fitted on time.

Our math: what do the numbers say?

• Cash gap: NT$125.9 billion paid for plant and equipment minus NT$50.5 billion from operations is about NT$75.4 billion in one quarter (our arithmetic).
• Borrowing: NT$100.9 billion of net new borrowing covered that gap plus the NT$29.0 billion dividend (our arithmetic on reported cash flows).
• Build pace: plant and equipment on the balance sheet rose about NT$104 billion, or 20%, in three months (our arithmetic).
• Equipment pace: US$2.1 billion of equipment spending in Q3 against about US$0.78 billion a year earlier is about 2.7 times (our arithmetic).
• Computing work: 34% of NT$151.5 billion is about NT$52 billion, against 25% of NT$100.3 billion, about NT$25 billion, a year earlier (our arithmetic).
• Q4 range: group revenue up 6% to 8% gives about NT$259 billion to NT$264 billion; ATM up 11% to 13% gives about NT$168 billion to NT$171 billion (our arithmetic).
• Margin path: a 30.6% ATM gross margin plus one to two points gives about 31.6% to 32.6% for the fourth quarter (our arithmetic).

What this tells you: ASE has more orders than floor, it can charge more for the floor it has, and it is borrowing to build more. Its profit is real and rising. But the next leg of growth is limited by how fast new plants can be finished and fitted, not by whether customers want the work.

Who gets paid first?

First, the toolmakers — on order and delivery. ASE spent US$2.1 billion on equipment in one quarter, more of it on testing than on packaging. Makers of testers, probe cards, handlers, bonders and inspection tools are paid when the machines ship and are accepted, before a single new chip is billed.

Second, the builders — on progress. Thirteen new buildings and eight converted sites mean steady payments to construction firms, clean room fitters and the makers of air, water and power systems inside the plant. Their work is paid stage by stage.

Third, the lenders — every quarter. Banks and bondholders are paid interest on the extra NT$100.9 billion borrowed this quarter whether the new lines start on time or late.

Fourth, ASE — as lines start. ASE is paid per chip packaged and tested. With prices firm and margins above 30%, each new line is worth more than it used to be. But it earns nothing until it runs, while its depreciation starts soon after.

Last, the chip buyers — when chips ship. Nvidia, AMD and the cloud companies get their chips only after packaging and testing are done. Where that step is short, they pay more or wait (the packaging gate; the CoWoS Analysis).

One-line rule: at ASE, toolmakers, builders and lenders are paid as the floor is built; ASE is paid as the lines start; chip buyers get their chips last. The stuck step is the new floor.

So what is the stuck step?

The new floor: buildings that are finished, clean rooms that are fitted and tools that are installed and qualified. Not demand, which ASE says is ahead of supply. Not money, which it can borrow. Four things make it stick:

• Buildings take longer than tools. A tester can be ordered and shipped in months. A new plant takes far longer, and ASE says its current wave of buildings runs to 2028.
• Spending runs ahead of capacity. ASE paid NT$125.9 billion for plant and equipment in a quarter, but that money only becomes revenue once the tools are running.
• Qualification takes time. New lines for AI chips must pass the customer’s checks before they run in volume. ASE’s full-process line only reaches mass-production yields next year.
• People are part of the floor. ASE’s headcount rose by 4,788 to 119,346 in the quarter (our arithmetic). New lines need trained staff as well as machines.

This is the bind our stuck-step Analysis describes: once money and orders are in place, the bind moves to the slowest physical input. At ASE that input is finished, fitted floor space.

Which dates test this read next?

• Late 2026 — Small volumes of co-packaged optics due to start, as ASE said in July.
• Early December 2026 — ASE’s November sales. A test of the 11% to 13% packaging and test growth guided for the fourth quarter.
• First quarter 2027 — ASE’s fully automated panel-level packaging line due to start production, as it said in July. Also ASE’s fourth-quarter results and its 2027 spending plan.
• 2027 — LEAP revenue targeted to double from about US$4.0 billion; full-process above US$1 billion; co-packaged optics above US$500 million.
• 2028 — The current wave of new buildings is meant to support ASE to here, with some reaching into 2029.

What would change this view?

• Floors arrive faster. If ASE beats its own guidance again because new buildings open early, the stuck step loosens and the payment order shortens.
• Prices soften. If packaging and test prices stop rising, the margin case weakens and the borrowing looks heavier.
• AI orders slow. If computing stops gaining share of ASE’s packaging and test revenue, demand rather than floor space becomes the question.
• Customers build their own. If TSMC or chip designers keep more packaging and test in-house, ASE’s new floor fills more slowly.

Practical takeaways

• ASE’s third-quarter revenue rose 44.8% to a record NT$244.2 billion; packaging and test grew 51% with a 30.6% margin
• It raised 2026 spending to US$11.5 billion and borrowed NT$100.9 billion in the quarter to build
• Testers are up 25% in a year while wirebonders are flat; Q3 test equipment spending beat packaging
• Fourth-quarter guidance: packaging and test up 11% to 13%, margin up one to two points
• Toolmakers, builders and lenders are paid now; ASE is paid as lines start; stuck step: the new floor

FAQ

What did ASE report on 29 October 2026?
ASE Technology Holding, the world's largest chip packaging and test house, reported third-quarter 2026 revenue of NT$244.2 billion, up 44.8% from a year earlier and 27.8% from the second quarter. Its packaging and test business (ATM) grew 51% to NT$151.5 billion with a 30.6% gross margin. Net profit for shareholders was NT$27.0 billion, up 148%.

How much is ASE spending, and how is it paying?
On the 29 October call ASE raised its 2026 capital spending plan from US$10.5 billion to US$11.5 billion. In the third quarter it paid NT$125.9 billion for property and equipment while its operations brought in NT$50.5 billion; net new borrowing was NT$100.9 billion, and its net debt to equity ratio rose from 0.47 to 0.66 in three months.

What is the stuck step?
The new floor: the buildings, clean rooms and tool installs that turn spending into working capacity. ASE said in July that demand was not its worry; its own pace of building, installing and ramping was. Its test machine count rose 25% in a year while wirebonders were flat, which shows where the new floor space is going.

Is this investment advice?
No. Education only. It maps who gets paid and which step is stuck. It is not a recommendation to buy or sell any security, and we leave out share prices, broker ratings and price targets on purpose.

Keep reading on the map

• Taiwan Markets hub — country page
• TSMC Q3 and CoWoS · Glass cloth and AI boards · Taiwan’s server mix — recent Taiwan Pulses
• Taiwan’s US parks · Chiayi packaging land · September exports record · Taipower and AI halls
• SK hynix Q3 and HBM4 · Meta’s capex and cash — other markets
• Who tests AI chips? · SoIC vs CoWoS · What is SoIC? · Capex vs depreciation
• TSMC’s CoWoS stuck step · Wiwynn’s working capital · Where does the stuck step move?
• Semiconductors on the stack · All Markets doors — Taiwan, Korea, US, China, Singapore, Australia, Malaysia

Not investment advice. Do your own research.

References

1. ASE Technology Holding (PR Newswire), unaudited consolidated financial results for the third quarter of 2026, 29 Oct 2026. prnewswire.com — revenue, margins, segment tables, capex, testers and wirebonders, balance sheet and cash flow.
2. Central News Agency (CNA), ASE third-quarter profit second-highest; fourth-quarter revenue seen up 6% to 8%, 29 Oct 2026. cna.com.tw — the fourth-quarter guidance given by CFO Joseph Tung on the call.
3. Central News Agency (CNA), ASE raises advanced packaging and test revenue view, lifts capex again, 29 Oct 2026. cna.com.tw — LEAP revenue, full-process and CPO targets, the US$11.5 billion capex and its split, and the margin comments.
4. Commercial Times (CTEE), ASE sees Q4 packaging and test revenue up 11% to 13%, 29 Oct 2026. ctee.com.tw — advanced packaging still short of demand and the 30.6% margin not being a ceiling.
5. Central News Agency (CNA), ASE September and third-quarter revenue hit records, 12 Oct 2026. cna.com.tw — the July guidance for the third quarter and the earlier US$10.5 billion capex estimate.
6. BusinessNext (Bnext), eight key points from ASE's second-quarter call, 30 Jul 2026. bnext.com.tw — the 13 new buildings and 8 acquired sites, negative free cash flow for a time, and execution as the main risk.