Kulicke and Soffa Industries
It makes the machines that stitch hair-thin wire from a finished chip to its package, many connections a second, and sells the tools those machines wear out.
Our viewUnattractive Medium conviction
The scores
The profile · 1-4
Growth2/4 ↑
FY2025 sales were 19% below FY2017's; the current surge is cycle amplitude. A 3 would credit the four quarters of real, if narrow, growth.
Business quality3/4 →
The aftermarket annuity held while equipment revenue halved. Not 4: thermocompression and hybrid bonding are taking the densest sockets and it holds neither.
Financial position4/4 ↑
$516.6M of cash and short-term investments, no borrowings at all, capital spending under $18M a year, and the plan needs no outside money.
Execution record3/4 ↑
Beat guidance four quarters running. Not 4: a cancelled product programme, a ceased business and three restructurings in two years is not capital allocated well.
Per share capture3/4 ↓
The count fell 26% over eight years. Not 4 now: buying stopped at about $100 while stock compensation continues, so issuance exceeds repurchase.
Potential2/4 →
The upside is a cycle somebody else spends into. Not 3: nothing is committed - the newest filed order figure predates the whole upcycle.
Risk profile2/4 ↓
Three concentrations overlap: three Chinese customers at 39% of sales, 56% of revenue to China, and the cycle. One pause hits all three.
Valuation2/4 ↑
At $37 for each $1 of nine-year average operating profit the coming peak must become the new middle. Not 1: about 13 times peak-year profit.
Evidence quality3/4 →
Everything reconciles to the filings and three outside primaries back the machine. Not 4: a 14.8% customer is unidentifiable and out-years rest on two analysts.
The bet · 2-3 years
Priced for success0.6×
If it works 0.6× · if it fails you keep about 45%.
The Verdict
We rate Kulicke and Soffa unattractive. It makes the machines that wire a chip to its package, and business is booming on real data-centre demand. But its sales more than halved after the last peak, and three Chinese customers now buy a large share of everything, so the business is mixed. The price is full, not cheap: it treats today's peak profit as an ordinary year, so another slump would hurt. As a bet it is fragile: even if its newer AI line hits management's target, the company is worth less than it costs, and if that line fails about two-fifths is left.
The long caseStacked AI chips need a newer kind of bonding tool; if K&S wins enough of those sales, a small loss-making unit could become a real business.
The evidence under this - 8 claims
C02facthigh confidence
Q3 2026 net revenue was $330.409M, 122.6% above Q3 2025's $148.413M and 36.2% above Q2 2026's $242.621M.
C15management claimmedium confidence
Management says the data-centre market relies on wire bonding at least as much as smartphones and personal computers do, because most chips in a data centre - networking, communications, power management and storage - are still packaged the traditional way.
against this
- It is an estimate by the company that sells wire bonders, offered without a source, and management prefaced it with 'we estimate'. No independent measurement of wire-bonded content in a data centre was found.
- It was volunteered in prepared remarks in the quarter the shares had fallen 30% from their high, which is when a company has most reason to reframe a legacy product as an artificial-intelligence one.
C16facthigh confidence
ASMPT, the only other large supplier of both wire bonders and thermocompression bonders, independently reports the same driver: heightened power-management needs in artificial-intelligence data centres kept utilisation at leading integrated device makers elevated, supporting its mainstream business.
C05facthigh confidence
Net revenue of $654.081M in FY2025 was 19.2% below FY2017's $809.041M. Revenue reached $1,503.620M in FY2022 and fell 56.5% over the three years to FY2025.
C11facthigh confidence
Three customers were 39.0% of nine-month FY2026 net revenue - Haoseng Industrial 14.8%, Tianshui Huatian Technology 13.8% and Changjin Technology (Shanghai) 10.4%. A year earlier only Tianshui Huatian cleared the 10% disclosure line, at 10.1%.
against this
- Concentration rising in an upcycle is partly arithmetic rather than a change in the customer base: when three buyers expand capacity at once, everyone else's share of a much larger number falls below 10% and disappears from the table. The ten largest customers were 54.8% of FY2025 revenue against 53.6% in FY2024, so the broad measure barely moved before this year.
- Tianshui Huatian is China's third-largest outsourced assembler and a listed company - a substantial, identifiable end customer rather than an obscure one.
C07estimatehigh confidence
At $98.73 a share the whole enterprise costs $4.88B, which is $37 for each $1 of the operating profit K&S has averaged over its last nine fiscal years and about $13 for each $1 of the operating profit the FY2027 analyst estimate implies.
derivedEnterprise value of $4,884,097,530 from the research pack at the 2026-10-05 close, divided by the $131.8M nine-year mean in C06, gives 37.1. The FY2027 figure works back from the panel's $5.90 a share on the 53.0 million diluted shares the Q4 2026 outlook uses: about $313M of net income, grossed up at the 21% tax rate management guides to and less roughly $18M of interest income, is about $378M of operating profit, or 12.9 times.
against this
- The pack's enterprise value nets off only $320.7M of cash, less than the $516.573M of cash and short-term investments the Q3 2026 10-Q reports; netting all of it puts the enterprise at $4.69B, or 35.6 and 12.4 times. On the $5.20B market value of the equity alone the two multiples are 39.5 and 13.8.
- A nine-year average operating profit is the wrong denominator if the installed base has structurally stepped up - which is exactly what management and ASMPT both argue is happening in data-centre bonding. If the cycle's floor has risen, the historic mean understates mid-cycle earnings and 37 times is an artefact of the window rather than a finding.
- 2026-10-06editedprice re-struck at the 2026-10-05 close of $98.73, up 21.0% since the thesis struck $81.61
C04facthigh confidence
Advanced Solutions - the segment that holds K&S's thermocompression bonding, and the part of the company sold as artificial-intelligence packaging - produced $29.415M of Q3 2026 net revenue, 8.9% of the company total, at an operating loss of $12.553M.
C19management claimlow confidence
Management targets Advanced Solutions revenue above $100M in FY2026 and says thermocompression bonding revenue will be somewhere in the region of $150M to $200M in FY2027.
against this
- The two figures are not the same measure and management ran them together on the call: the prepared remarks put over $100M on the Advanced Solutions segment for FY2026, while the answer to the question put $100M on thermocompression bonding alone. The FY2027 range is therefore struck on a base that was not clearly stated.
- The segment lost $12.553M in Q3 2026 at a 42.7% negative operating margin, so a revenue target says nothing about whether the business earns anything at that size.
- $150M to $200M in FY2027 would be 14% to 19% of the whole thermocompression market ASMPT sizes at roughly $1.05B that year, in a market whose leader targets 35% to 40% and where the largest memory socket sits with a third supplier.
The company
Every chip has to be connected to the package that gets soldered onto a circuit board. K&S's main machine, the ball bonder, stitches gold, silver or copper wire from pads on the chip to leads on the package, one connection at a time. Ball bonders were 68.4% of June-quarter sales.
It also makes heavier bonders for power parts, newer machines that press a chip straight onto tiny bumps, and the capillaries, blades and wedges all of them wear out.
The evidence under this - 2 claims
C01facthigh confidence
K&S sells the machines that attach a chip to its package - ball bonders that stitch hair-thin gold, silver-alloy or copper wire from the chip to the package, wedge bonders that use ribbon or heavy wire for power parts, thermocompression and related advanced tools, and the capillaries, blades and bonding wedges those machines consume.
C03facthigh confidence
Ball Bonding Equipment produced $226.068M of Q3 2026 net revenue - 68.4% of the company's $330.409M - and $93.207M of segment operating income, against a company operating income of $68.266M.
How it makes money
Mostly by selling machines, which chip-assembly companies buy in bursts when they add factories and stop buying when they don't. Sales hit $1.50B in fiscal 2022, then fell 56.5% over three years. The steadier dollars come from the tools the machines consume: $156M to $161M a year right through that slump.
The evidence under this - 2 claims
C05facthigh confidence
Net revenue of $654.081M in FY2025 was 19.2% below FY2017's $809.041M. Revenue reached $1,503.620M in FY2022 and fell 56.5% over the three years to FY2025.
C21facthigh confidence
Aftermarket Products and Services revenue was $160.718M in FY2023, $160.009M in FY2024 and $156.129M in FY2025 - a 2.9% decline across three years in which total revenue fell 11.9% and ball bonder revenue fell 18.1% in FY2025 alone.
Where the money comes from, stream by stream
| Stream | Share of sales (Q3 FY2026) | How it earns it |
|---|---|---|
| Ball Bonding Equipment | 68.4% | Wire bonders that stitch chips to their packages, sold mostly to Chinese and other Asian assembly houses as they add capacity. |
| Aftermarket Products and Services | 10.8% | Capillaries, blades, wedges, spares and service consumed by the installed machines, bought steadily whatever the cycle. |
| Advanced Solutions | 8.9% | Thermocompression and die-attach systems for advanced logic and memory packaging; growing, still loss-making. |
| Wedge Bonding Equipment | 7.8% | Heavy-wire and ribbon bonders for power chips in cars, industrial equipment and batteries. |
| All Others | 4.0% | Dispensing equipment and the run-off of the ceased Electronics Assembly business, too small to report separately. |
Who runs it
Raj Talluri took over as chief executive on 1 September 2026, after running battery maker Enovix and Micron's mobile chip business; he has no record here yet. Lester Wong, finance chief since 2018, ran the company in between and sold half his shares near the high. The yearly bonus now rides on operating profit and gross margin, with 15% on personal goals for executives; the long-term shares pay only if the stock beats other chip companies. Nothing pays for growth.
The evidence under this - 6 claims
C33facthigh confidence
Dr Raj Talluri became president and chief executive on September 1, 2026, joining from Enovix where he was chief executive; the shares fell 9.5% in the session after the August 17, 2026 announcement. Lester Wong reverts to chief financial officer.
C57facthigh confidence
Lester Wong joined K&S in 2011 as general counsel, became chief financial officer in December 2018 and was interim chief executive from October 28, 2025 until September 1, 2026, after Fusen Chen retired. Robert Chylak, at K&S since 1980, has been chief technology officer since December 2019.
C53facthigh confidence
As of December 8, 2025 the ten directors, nominees and current executive officers owned 651,409 shares, 1.2% of the company; the largest current holder among them was Nelson Wong with 156,369. The retired chief executive Fusen Chen held 2.2% outside that group, and BlackRock (15.7%) and Vanguard (6.2%) were the only holders above 5%.
C54facthigh confidence
The fiscal 2025 cash bonus pool paid half on net income and half on operating margin, excluding one-time items: reported net income of $0.2M and operating margin of -0.49% became $25.3M and 2.6% after excluding $87.5M of Electronics Assembly cessation charges, paying 39.98% of target. All fiscal 2025 PSUs vest on three-year total shareholder return relative to a GICS index, capped at target when absolute return is negative; the CEO and CFO take 60% of equity as PSUs.
against this
- The adjustment turned a loss year into a bonus: without excluding the cessation charges, reported net income of $0.2M sat below the $16.1M threshold and nothing would have paid.
- No metric pays on revenue growth, the thermocompression ramp or cash conversion, so the receivables build in C28 costs the plan nothing.
C55facthigh confidence
The new chief executive's annual bonus targets 110% of a $750,000 salary, paid between 0% and 200% on corporate results whose metrics the compensation committee sets, and he must hold shares worth three times salary within five years. On 13 October 2026 the committee changed the plan for fiscal 2027: the two measures became operating profit and gross margin, half each, and executive awards add a 15% personal-goal share (85% corporate).
sourceCEO employment agreement, Exhibit 10.1 to Form 8-KprimaryK&S Form 8-K, Item 5.02: amended and restated Incentive…primary
- 2026-10-19editedthe 13 October 2026 8-K replaced the bonus measures (net income and operating margin) with operating profit and gross margin for fiscal 2027 and added a 15% individual-goal share for executives
C23facthigh confidence
Lester Wong, then interim chief executive and chief financial officer, sold 60,000 shares at $121.7092 on June 18, 2026 under transaction code S, an open-market sale, leaving 50,743 shares - 54.2% of the holding, sold within two weeks of the twelve-month high.
- 2026-10-06editedthe Form 4's Rule 10b5-1 field read: no plan claimed; the claim text is unchanged
Why it wins, and who could beat it
Why customers come back
A bonder is tuned into a customer's production line: the force, heat and vibration for each part are proven on chips already selling, and the consumables are matched to that machine. Changing supplier means proving the yield all over again, so a customer running well does not switch.
The evidence is the consumables line. It fell only 2.9% across three years, while total sales fell 11.9% in fiscal 2025 alone, because installed machines keep needing tools even when nobody buys new ones.
The evidence under this - 2 claims
C21facthigh confidence
Aftermarket Products and Services revenue was $160.718M in FY2023, $160.009M in FY2024 and $156.129M in FY2025 - a 2.9% decline across three years in which total revenue fell 11.9% and ball bonder revenue fell 18.1% in FY2025 alone.
C22inferencemedium confidence
That steadiness is what a switching cost looks like from outside: a bonder is qualified into a customer's production process and the capillaries and wedges it consumes are matched to that machine and that process, so the installed base keeps buying tools, spares and service in the years it stops buying machines.
derivedC21 - an aftermarket line that moved 2.9% across three years while the equipment lines around it fell by five to six times as much - together with C01, the fact that the same company supplies both the machine and its consumables.
against this
- Aftermarket stability may be a property of the installed base rather than of any advantage K&S holds: spares demand tracks how hard existing machines are running, which is a utilisation fact, and any equipment maker with a large installed base shows the same pattern.
- K&S's own 10-K names PECO, Disco and Small Precision Tools as consumables competitors, and states that its FY2025 aftermarket gross margin fell partly on 'unfavourable pricing from bonding tools' - which is the filing saying the consumables business does not hold price.
- K&S itself sells aftermarket products for its peers' equipment, which shows the consumables are not fully locked to the machine that uses them.
The threat
The newest packaging is moving away from wire. Stacked AI chips and high-bandwidth memory are joined by pressing the chip onto bumps, and in June SK hynix ordered 44.2 billion won of those machines from Hanmi. K&S has no disclosed position in that memory.
Next comes hybrid bonding, copper joined straight to copper, which lifted BE Semiconductor's orders 7.7% in a quarter.
The evidence under this - 3 claims
C36factmedium confidence
SK hynix placed an order worth 44.2 billion won with Hanmi Semiconductor in June 2026 for thermocompression bonders for its sixth-generation high-bandwidth memory.
C35inferencemedium confidence
K&S holds no disclosed position in high-bandwidth memory bonding. Management describes its memory business as focused on NAND 'for now' and its stacked-DRAM work as still in development, while the thermocompression sockets at the high-bandwidth memory makers sit with Hanmi and ASMPT.
derivedC34 and S04d, where management names NAND as the memory focus and describes vertical wire work on new forms of stacked DRAM as ongoing development; and S12, SK hynix placing a 44.2 billion won order with Hanmi for HBM4 thermocompression bonders in June 2026. No K&S filing or release names a high-bandwidth memory customer or programme.
against this
- Absence of disclosure is not absence of position: equipment makers routinely do not name memory customers, and K&S names none of its customers anywhere except in the concentration note, where they are identified only because accounting rules require it.
- Management does say the vertical wire team works closely with memory customers on new forms of stacked DRAM, so there is a programme, just not a disclosed win. And high-bandwidth memory is one socket - being outside it is not the same as being outside the artificial-intelligence build.
C20facthigh confidence
BE Semiconductor reported Q1 2026 orders of EUR 269.7M, up 7.7% on the previous quarter primarily because of a significant increase in bookings for hybrid bonding systems - the die-attach method that replaces wire bonding at the highest interconnect densities.
The rivals: all private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| ASMPT thermocompression bondersPriced against it: Listed abroad | Ball Bonding Equipment, Wedge Bonding Equipment, Aftermarket Products and Services and All Others | On sale | Listed abroad | |
| Hanmi Semiconductor TC Bonder, at SK hynixPriced against it: Listed abroad | Advanced Solutions | On sale | Listed abroad | |
| BE Semiconductor (Besi) hybrid bonding systemsPriced against it: Listed abroad | Advanced Solutions | On sale | Listed abroad |
How big it could get
How big this could get
SEMI, the industry's trade body, expects $6.7B of spending on chip-assembly machines in 2026, up 9.6%, reaching $8.6B by 2028. That is the whole of what K&S sells into.
The pressing machines are a small, fast-growing slice: about $760M in 2025 and $1.6B by 2028, by the estimate of ASMPT, the leader, which aims to keep 35% to 40% of it.
The evidence under this - 2 claims
C17factmedium confidence
SEMI, the industry's trade association, forecasts that assembly and packaging equipment sales rise 9.6% to $6.7B in 2026 after growing 20.8% in 2025, and reach $8.6B by 2028 - inside a total equipment market it forecasts at a record $165.9B in 2026.
sourceSEMI - Year-End Total Semiconductor Equipment Forecast,…primarySEMI - Year-End Forecast, total equipment and the 2028…primarySEMI - Year-End Forecast, total equipment sales for 2026primary
against this
- SEMI's out-year numbers move a great deal. Its December 2025 vintage put total 2026 equipment at $145B; seven months later it put the same year at $165.9B, a 14.4% upward revision - so a forecast two years out is not a fact about 2028.
- The category boundary is not the one a reader assumes: much advanced-packaging spend on redistribution layers and interposers is classified as wafer-fab equipment rather than assembly and packaging, so this line understates total spending on advanced packaging.
C18factmedium confidence
ASMPT, the leader in thermocompression bonding, sizes the entire global market for it at about $760M in 2025 rising to $1.6B in 2028, and says it targets a 35% to 40% share of that market.
sourceASMPT Limited - 2025 annual results announcementprimaryASMPT Limited - 2025 annual results, target share of the TCB…primary
against this
- It is one participant's estimate of its own market, published while raising it - ASMPT revised the number upward, so the direction of its bias is toward a larger market rather than a smaller one, which makes it a conservative bound for the argument used here.
- A market growing at a 30% compound rate to 2028 is growing far faster than the assembly and packaging category around it, so thermocompression is the fast part of a slow market rather than a stagnant niche.
Who has signed
Almost nothing on file. The last reported orders on hand were $245.3M at October 2025, less than one quarter at today's $375M rate, and the latest quarterly filing gives no figure at all. Management says orders now reach into early 2027, which it says does not usually happen.
The evidence under this - 2 claims
C30facthigh confidence
The Q3 2026 10-Q states no backlog figure anywhere. The last filed backlog is $245.280M at October 4, 2025, up 65.1% from $148.585M a year earlier - less than one quarter of the $375M rate now being guided.
C14management claimmedium confidence
Management says utilisation is above 95% in China and around 90% in memory and general semiconductor, that purchase orders now extend into fiscal Q2 2027 which does not usually happen, and that Chinese customers continue to build factories.
against this
- The 10-K's own risk factors say purchase orders may be revised or cancelled without penalty and that the industry is subject to double-booking and unexpected build-ups of inventory in the supply channel. An order book that extends unusually far is exactly what double-booking looks like from the supplier's side.
- No backlog figure appears in the Q3 2026 10-Q, so none of this is checkable against a filed quantity.
Where the growth is coming from, business by business
Where the growth is coming from
| Business | Share of revenue | Growth: last quarter | Last half-year | Last year |
|---|---|---|---|---|
| Ball Bonding Equipment Segment | 68% | +197.5% | +171.5% | +66.5% |
| Rest of the company | 23% | +21.6% | -4.9% | -3.8% |
| Advanced Solutions | 9% | +172.1% | +89.5% | +29.6% |
| Whole company | 100% | +122.6% | +84.6% | +33.0% |
The filed split through 2026-07-04. These rates are not carried forward: the reading below says why.
How we read it: The blend. Ball Bonding's 172% is a cycle rebounding from FY2025's trough, when revenue sat 19% below FY2017 after falling 56.5% from FY2022; management calls that peak extreme and analysts model FY2028 lower.
The results
The numbers
Is it growing?Fast
Sales grew 44% over the twelve months to Jul 2026, faster than the −7.4% in the year to Oct 2025.
Is it profitable?Healthy
Operating margin is 13%, up from 5.3% in the year to Sep 2023.
Does growth pay off?Pays off
Running costs take 35¢ of each sales dollar, up from 19¢ in the year to Oct 2021.
Does it earn good returns?Decent
Each dollar put into the business earns 11% a year after tax, down from 39% in the year to Oct 2021.
Is it making cash?Thin
It turned 4.2% of sales into free cash, $40.0M over the last twelve months.
How steady is it?Swingy
Yearly sales growth has ranged from −51% to +144% over the last 5 years, with sales falling in 4 of them.
Are shares being diluted?Shrinking
The share count fell 16% in 4.8 years, and pay in shares costs 2.9% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in October.
| Oct 2021 | Oct 2022 | Sep 2023 | Sep 2024 | Oct 2025 | 12 mo to Jul 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $1.52B | $1.50B | $742M | $706M | $654M | $950M |
| Sales growth | +144% | −0.9% | −51% | −4.9% | −7.4% | +44% |
| Profit | ||||||
| Gross margin | 46% | 50% | 48% | 38% | 42% | 48% |
| Operating margin | 27% | 31% | 5.3% | −13% | −0.5% | 13% |
| Net margin | 24% | 29% | 7.7% | −9.8% | 0.0% | 12% |
| What growth costs, per sales dollar | ||||||
| Research and development | 9.1% | 9.1% | 19% | 21% | 23% | 17% |
| Selling, general and admin | 9.7% | 9.3% | 21% | 23% | 26% | 18% |
| Pay in shares | 1.0% | 1.3% | 3.1% | 3.8% | 4.4% | 2.9% |
| Capital spending | 1.5% | 1.5% | 6.0% | 2.3% | 2.6% | 1.5% |
| Cash | ||||||
| Cash from operations | $300M | $390M | $173M | $31.0M | $114M | $54.0M |
| Free cash flow | $277M | $367M | $129M | $14.9M | $96.4M | $40.0M |
| Balance sheet | ||||||
| Cash and short-term investments | $740M | $776M | $759M | $577M | $511M | $517M |
| Returns and shares | ||||||
| Return on capital | 39% | 37% | 2.6% | - | - | 11% |
| Shares (diluted) | 63.5M | 61.2M | 57.5M | 55.6M | 53.2M | 53.4M |
What growth costs
Most of what K&S spends on research, selling and running the company stays roughly the same whatever it sells. So when sales recover, each extra sale is cheap and profit margins widen fast, as they are now. The catch runs the other way: when sales fell by half after 2022, the same costs pushed the company into a loss.
The evidence under this - 2 claims
C58facthigh confidence
Per sales dollar, R&D ran 9.1 cents in FY2021 and FY2022, rose to 22.9 in FY2025 and fell to 17.0 over the four quarters to July 2026; selling, general and administrative ran 9.7, 9.3, 25.6 and 18.0. In dollars both moved far less: R&D from $137.5M in FY2021 to $161.5M, SG&A from $147.1M to $170.7M. Operating margin went 31.3% in FY2022, -0.5% in FY2025, 13.2% over the four quarters.
sourceForm 10-K, fiscal year ended September 30, 2023primaryForm 10-K, fiscal year ended October 4, 2025primaryForm 10-Q, quarter ended July 4, 2026primary
against this
- The leverage runs both ways: when revenue halved from FY2022 to FY2024, the same fixed costs turned a 31.3% margin into an operating loss (C06).
- Advanced Solutions carries the largest segment R&D, $47.4M in the nine months to July 2026, and loses money (C04), so part of the cost base is spending for a market that has not yet paid for it.
C06estimatehigh confidence
Across its last nine fiscal years, FY2017 to FY2025, K&S earned an average of $131.8M of operating profit a year - a range running from a $92.496M loss in FY2024 to a $470.072M profit in FY2022.
derivedThe mean of the nine filed annual operating income figures: $113.083M, $166.632M, $21.610M, $58.509M, $412.447M, $470.072M, $39.437M, -$92.496M and -$3.224M, summing to $1,186.070M over nine years.
against this
- A nine-year window is a choice. Taking twelve years back to FY2014 lowers the average to $113.0M; taking the last five years raises it to $165.2M. The nine-year window spans two full cycles and is the least flattering of the three to the argument being made against it, so the shorter window is the fair test and it moves the multiple from 30 to 24 rather than removing it.
- The perimeter is not constant across the window: the Electronics Assembly business was ceased in FY2025 and Project W was cancelled in FY2024, so part of the loss years is exit cost rather than trading. Adding back the FY2024 impairment of $39.817M and the FY2025 restructuring would lift the average by roughly $5M a year.
The last report (5 Aug)
Quarter ended July 2026, filed figures.
123%
Sales growth
against the same quarter a year earlier
36.2%
Growth from the quarter before
against 21.5% the quarter before that
21¢
Operating profit
per dollar of sales
The third quarter of fiscal 2026, to 4 July, was the boom in numbers: sales of $330.4M, up 122.6% from a year earlier, and a guide of about $375M for the next quarter. The growth came from ordinary chips for data centres, networking and power.
The AI packaging line, Advanced Solutions, brought in $29.4M, 8.9% of sales, and lost $12.6M. Cash lagged too: over nine months operating profit swung to $124.7M while cash from operations fell 55.7%, the working capital a fast ramp consumes.
What it did to the thesisIt confirmed the boom is real and comes from ordinary chips, not the AI packaging line, which is still small and losing money. Cash lagged as the ramp ate working capital. It shows peak profits, not a new normal, so the view stays Unattractive.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $1.20 against $1.06 expected; $1.07 as filed, the gap being what the company adds back.
The evidence under this - 7 claims
C02facthigh confidence
Q3 2026 net revenue was $330.409M, 122.6% above Q3 2025's $148.413M and 36.2% above Q2 2026's $242.621M.
C10management claimhigh confidence
K&S guides Q4 2026 revenue to about $375M plus or minus $20M, gross margin of 48%, earnings of $1.29 a share and $1.42 a share excluding stock compensation and other adjustments.
sourceQ3 2026 results release, Exhibit 99.1 to Form 8-KprimaryQ3 2026 earnings call transcript, held August 6, 2026primary
against this
- K&S exceeded its own revenue outlook in each of the first three quarters of FY2026, so this guide is more likely a floor than a ceiling and the real run rate may be above $375M.
- It covers one quarter. Management guides one quarter at a time and explicitly declines to characterise fiscal 2027, so the guide carries no information at all about the period this thesis actually turns on.
C34management claimmedium confidence
In Q3 2026 general semiconductor revenue rose 52.6% sequentially to $227.2M, memory rose 8.8% to $34.0M and automotive and industrial rose 9% to $24.2M. Management says its memory business is focused on NAND and that about 40% of the NAND market now goes to data centres.
C15management claimmedium confidence
Management says the data-centre market relies on wire bonding at least as much as smartphones and personal computers do, because most chips in a data centre - networking, communications, power management and storage - are still packaged the traditional way.
against this
- It is an estimate by the company that sells wire bonders, offered without a source, and management prefaced it with 'we estimate'. No independent measurement of wire-bonded content in a data centre was found.
- It was volunteered in prepared remarks in the quarter the shares had fallen 30% from their high, which is when a company has most reason to reframe a legacy product as an artificial-intelligence one.
C04facthigh confidence
Advanced Solutions - the segment that holds K&S's thermocompression bonding, and the part of the company sold as artificial-intelligence packaging - produced $29.415M of Q3 2026 net revenue, 8.9% of the company total, at an operating loss of $12.553M.
C28facthigh confidence
Over the first nine months of FY2026 revenue rose 62.1% to $772.655M and operating income turned from a $4.112M loss to a $124.652M profit, while cash from operations fell 55.7% to $46.554M from $105.003M.
C29inferencehigh confidence
That cash gap is the working capital a fast ramp consumes, not a deterioration in collection - and the prior-year comparison flatters it, because the nine months to June 2025 released working capital as the business shrank and carried a $75.987M gain on ceasing the Electronics Assembly business.
derivedC28, plus the balance-sheet movements behind it and the prior-year income statement. Receivables at $329.498M were 99.7% of the quarter's revenue against 103.4% at the FY2025 year end, so the ratio improved while the absolute balance nearly doubled.
against this
- Working capital consumed by a ramp is returned only if the ramp stops gently. If orders are cancelled rather than tapering, the inventory built for them is written down rather than sold, which is what happened to the spares inventory when the Electronics Assembly business was ceased.
- Half the receivable sits with two counterparties, so the aggregate ratio being healthy does not mean the exposure is.
What management did with the money
It stopped buying back shares as they rose: $96.5M at about $40 in fiscal 2025, $0.5M at about $100 last quarter. It holds $516.6M of cash and no borrowings.
Insiders in 2026Q3: 8 purchases and 0 sales, from their Form 4 filings.
The evidence under this - 2 claims
C25facthigh confidence
K&S repurchased 2.4 million shares for $96.5M in FY2025, about $40 a share. In Q3 2026 it repurchased 5.0 thousand shares for $0.5M, about $100 a share.
C27facthigh confidence
Cash and short-term investments were $516.573M at July 4, 2026 against shareholders' equity of $911.957M, and the company carries no borrowings: interest expense was $34 thousand in the quarter against $4.529M of interest income.
The price, and where we disagree
In short. Full. The price pays for today's peak profit becoming an ordinary year, at a company whose ordinary years have included sales halving.
Share price
$98.73
close,
Price for each $1 of yearly profit
45.1
its 6-year normal 18.1
Company value for each $1 of cash profit
30.4
its 6-year normal 10.9
The shares have more than doubled in a year on the AI packaging story. The question is whether $375M a quarter is a plateau or a peak, and it gets its first answer on 18 November, when K&S gives its outlook for fiscal 2027.
Until then management will not describe the year. It calls the $1.5B years of 2021 and 2022 'extreme levels' produced by a pandemic.
The evidence under this - 3 claims
C07estimatehigh confidence
At $98.73 a share the whole enterprise costs $4.88B, which is $37 for each $1 of the operating profit K&S has averaged over its last nine fiscal years and about $13 for each $1 of the operating profit the FY2027 analyst estimate implies.
derivedEnterprise value of $4,884,097,530 from the research pack at the 2026-10-05 close, divided by the $131.8M nine-year mean in C06, gives 37.1. The FY2027 figure works back from the panel's $5.90 a share on the 53.0 million diluted shares the Q4 2026 outlook uses: about $313M of net income, grossed up at the 21% tax rate management guides to and less roughly $18M of interest income, is about $378M of operating profit, or 12.9 times.
against this
- The pack's enterprise value nets off only $320.7M of cash, less than the $516.573M of cash and short-term investments the Q3 2026 10-Q reports; netting all of it puts the enterprise at $4.69B, or 35.6 and 12.4 times. On the $5.20B market value of the equity alone the two multiples are 39.5 and 13.8.
- A nine-year average operating profit is the wrong denominator if the installed base has structurally stepped up - which is exactly what management and ASMPT both argue is happening in data-centre bonding. If the cycle's floor has risen, the historic mean understates mid-cycle earnings and 37 times is an artefact of the window rather than a finding.
- 2026-10-06editedprice re-struck at the 2026-10-05 close of $98.73, up 21.0% since the thesis struck $81.61
C10management claimhigh confidence
K&S guides Q4 2026 revenue to about $375M plus or minus $20M, gross margin of 48%, earnings of $1.29 a share and $1.42 a share excluding stock compensation and other adjustments.
sourceQ3 2026 results release, Exhibit 99.1 to Form 8-KprimaryQ3 2026 earnings call transcript, held August 6, 2026primary
against this
- K&S exceeded its own revenue outlook in each of the first three quarters of FY2026, so this guide is more likely a floor than a ceiling and the real run rate may be above $375M.
- It covers one quarter. Management guides one quarter at a time and explicitly declines to characterise fiscal 2027, so the guide carries no information at all about the period this thesis actually turns on.
C40management claimhigh confidence
Management will not characterise fiscal 2027 as a whole. It says visibility is usually not even as far out as the first half of fiscal 2027, calls the $1.5B revenue of 2021 and 2022 'extreme levels' produced by a pandemic, and will give more colour on the second half at the November call.
against this
- Declining to guide a full year is K&S's standing practice - it guides one quarter at a time - so the refusal carries less information than it appears to.
- In the same answers management said it feels 'pretty confident' the first half of fiscal 2027 will be strong and, in the prepared remarks, that it continues to anticipate above-average demand in fiscal 2027. The reticence is about the second half only.
The betPriced for success
If it works
0.6×
from today's value, at the stated multiple
If it fails
keep 45%
of today's market value: the rest of the company, less its debt
How we read the multiple
Counting its cash, the whole company costs $4.88B. That is about $13 for each $1 of operating profit analysts expect in fiscal 2027, but $37 for each $1 of the $131.8M it has averaged over nine years, a range from a loss to a $470M profit.
That gap is the whole argument, and the fiscal 2027 forecast simply assumes the current $375M of quarterly sales holds all year.
Against its own history. Dear against its own normal even while profit is near a cyclical high, so the price leans entirely on the high lasting.
The evidence under this - 3 claims
C07estimatehigh confidence
At $98.73 a share the whole enterprise costs $4.88B, which is $37 for each $1 of the operating profit K&S has averaged over its last nine fiscal years and about $13 for each $1 of the operating profit the FY2027 analyst estimate implies.
derivedEnterprise value of $4,884,097,530 from the research pack at the 2026-10-05 close, divided by the $131.8M nine-year mean in C06, gives 37.1. The FY2027 figure works back from the panel's $5.90 a share on the 53.0 million diluted shares the Q4 2026 outlook uses: about $313M of net income, grossed up at the 21% tax rate management guides to and less roughly $18M of interest income, is about $378M of operating profit, or 12.9 times.
against this
- The pack's enterprise value nets off only $320.7M of cash, less than the $516.573M of cash and short-term investments the Q3 2026 10-Q reports; netting all of it puts the enterprise at $4.69B, or 35.6 and 12.4 times. On the $5.20B market value of the equity alone the two multiples are 39.5 and 13.8.
- A nine-year average operating profit is the wrong denominator if the installed base has structurally stepped up - which is exactly what management and ASMPT both argue is happening in data-centre bonding. If the cycle's floor has risen, the historic mean understates mid-cycle earnings and 37 times is an artefact of the window rather than a finding.
- 2026-10-06editedprice re-struck at the 2026-10-05 close of $98.73, up 21.0% since the thesis struck $81.61
C06estimatehigh confidence
Across its last nine fiscal years, FY2017 to FY2025, K&S earned an average of $131.8M of operating profit a year - a range running from a $92.496M loss in FY2024 to a $470.072M profit in FY2022.
derivedThe mean of the nine filed annual operating income figures: $113.083M, $166.632M, $21.610M, $58.509M, $412.447M, $470.072M, $39.437M, -$92.496M and -$3.224M, summing to $1,186.070M over nine years.
against this
- A nine-year window is a choice. Taking twelve years back to FY2014 lowers the average to $113.0M; taking the last five years raises it to $165.2M. The nine-year window spans two full cycles and is the least flattering of the three to the argument being made against it, so the shorter window is the fair test and it moves the multiple from 30 to 24 rather than removing it.
- The perimeter is not constant across the window: the Electronics Assembly business was ceased in FY2025 and Project W was cancelled in FY2024, so part of the loss years is exit cost rather than trading. Adding back the FY2024 impairment of $39.817M and the FY2025 restructuring would lift the average by roughly $5M a year.
C09estimatemedium confidence
The FY2027 revenue estimate of $1,517.0M is only 1.1% above what the Q4 2026 outlook of $375M produces if simply repeated four times. The panel is not modelling further growth - it is modelling the current quarterly rate holding for a year, and then easing.
derived$375M multiplied by four is $1,500M; the panel's $1,517.0M is 1.13% above it. The FY2028 estimate of $1,461.7M is 3.6% below FY2027.
against this
- K&S's fiscal year ends in early October, so FY2027 runs from October 2026 to October 2027 and contains a seasonally weaker December quarter that management has said always carries some seasonality. Holding $375M flat across a year that includes that quarter therefore embeds growth in the other three, not a flat line.
- Two contributors is a thin panel and one of them may simply not have updated after the August print. A two-analyst mean is weak evidence about what anyone expects.
Priced against its rivals
None of the listed rivals above sells the same thing to the same buyers as a large part of what it does. A price comparison needs at least two, so the price is judged against its own history instead.
Private or foreign rivals that cannot be priced: ASM Pacific Technology (ASMPT Limited), BE Semiconductor Industries N.V. (Besi), Han's Laser Technology Co., Ltd., Disco Corporation, Panasonic Holdings Corporation, Yamaha Robotics Holdings Co. Ltd., Hanwha Precision Machinery Co., Ltd., Hesse GmbH, PECO Co., Ltd., Small Precision Tools Co., Ltd.
The bet: what it pays if it works
This bet is on one business only: Advanced Solutions, the newer bonding machines sold for artificial-intelligence chips, which still loses money. Wire bonders and spare parts are counted separately, at what the thesis thinks a normal year earns, and they cover about half of today's price.
Even if Advanced Solutions reaches the top of management's own target, the company is worth less than it costs, so the bet is expensive. If it comes to nothing, shareholders keep a debt-free core worth about half of what the company costs today.
| Advanced Solutions revenue in two to three years | Points of the 2028 thermocompression bonding market needed | At the stated multiple |
|---|---|---|
| $175.00M | about 11 at 100% kept | 0.5× |
| $200.00M (the big win) | about 12 at 100% kept | 0.6× |
| $320.00M (market forecast) | about 20 at 100% kept | 0.7× |
The leader sizes thermocompression at $1.6B by 2028; held there, a 20% share, above the roughly 15% K&S's own FY2027 range implies, is about 150 tools and $320 million of revenue, with hybrid bonding counted at nothing. On that forecast, $320.00M pays 0.7× from today's value: an Expensive bet on the market case. Advanced Solutions would be worth about $1.20B, 0.4× what the price pays for it.
To pay 2× (a Fair bet) it would need about $2.07B of Advanced Solutions revenue (about 129 points of the 2028 thermocompression bonding market); to pay 5× (a Cheap bet), about $5.98B of Advanced Solutions revenue (about 373 points of the 2028 thermocompression bonding market), at the same multiple.
Each row is Advanced Solutions alone, plus the rest of the company (Ball Bonding Equipment, Wedge Bonding Equipment, Aftermarket Products and Services and All Others) counted at about $2.01B.
The steps it has to take
- Done Advanced Solutions selling thermocompression tools for revenue
- In progress Advanced Solutions above $100M of FY2026 revenue due 2026-11-18
- Not started A hybrid bonding tool delivered to a customer due 2027-04
- Not started The expanded Singapore Advanced Solutions plant complete due 2027-10
- Not started FY2027 thermocompression revenue of $150M to $200M reported due 2027-11
- Not started A high-bandwidth memory bonding win at a named memory maker due 2028-09
- Not started Advanced Solutions at an operating profit on more than $200M of annual revenue due 2028-11
The detail behind the label
Rating
Expensive
the big win's payoff against the 2× and 5× lines
Cheap bet at about
≈ $16.57
83% below the price (close, )
The option
0.3×
what Advanced Solutions is worth if it works, against what the price pays for it
Market case
0.7×
an expensive bet if the market forecast holds
The evidence under this - 5 claims
C04facthigh confidence
Advanced Solutions - the segment that holds K&S's thermocompression bonding, and the part of the company sold as artificial-intelligence packaging - produced $29.415M of Q3 2026 net revenue, 8.9% of the company total, at an operating loss of $12.553M.
C19management claimlow confidence
Management targets Advanced Solutions revenue above $100M in FY2026 and says thermocompression bonding revenue will be somewhere in the region of $150M to $200M in FY2027.
against this
- The two figures are not the same measure and management ran them together on the call: the prepared remarks put over $100M on the Advanced Solutions segment for FY2026, while the answer to the question put $100M on thermocompression bonding alone. The FY2027 range is therefore struck on a base that was not clearly stated.
- The segment lost $12.553M in Q3 2026 at a 42.7% negative operating margin, so a revenue target says nothing about whether the business earns anything at that size.
- $150M to $200M in FY2027 would be 14% to 19% of the whole thermocompression market ASMPT sizes at roughly $1.05B that year, in a market whose leader targets 35% to 40% and where the largest memory socket sits with a third supplier.
C18factmedium confidence
ASMPT, the leader in thermocompression bonding, sizes the entire global market for it at about $760M in 2025 rising to $1.6B in 2028, and says it targets a 35% to 40% share of that market.
sourceASMPT Limited - 2025 annual results announcementprimaryASMPT Limited - 2025 annual results, target share of the TCB…primary
against this
- It is one participant's estimate of its own market, published while raising it - ASMPT revised the number upward, so the direction of its bias is toward a larger market rather than a smaller one, which makes it a conservative bound for the argument used here.
- A market growing at a 30% compound rate to 2028 is growing far faster than the assembly and packaging category around it, so thermocompression is the fast part of a slow market rather than a stagnant niche.
C27facthigh confidence
Cash and short-term investments were $516.573M at July 4, 2026 against shareholders' equity of $911.957M, and the company carries no borrowings: interest expense was $34 thousand in the quarter against $4.529M of interest income.
C07estimatehigh confidence
At $98.73 a share the whole enterprise costs $4.88B, which is $37 for each $1 of the operating profit K&S has averaged over its last nine fiscal years and about $13 for each $1 of the operating profit the FY2027 analyst estimate implies.
derivedEnterprise value of $4,884,097,530 from the research pack at the 2026-10-05 close, divided by the $131.8M nine-year mean in C06, gives 37.1. The FY2027 figure works back from the panel's $5.90 a share on the 53.0 million diluted shares the Q4 2026 outlook uses: about $313M of net income, grossed up at the 21% tax rate management guides to and less roughly $18M of interest income, is about $378M of operating profit, or 12.9 times.
against this
- The pack's enterprise value nets off only $320.7M of cash, less than the $516.573M of cash and short-term investments the Q3 2026 10-Q reports; netting all of it puts the enterprise at $4.69B, or 35.6 and 12.4 times. On the $5.20B market value of the equity alone the two multiples are 39.5 and 13.8.
- A nine-year average operating profit is the wrong denominator if the installed base has structurally stepped up - which is exactly what management and ASMPT both argue is happening in data-centre bonding. If the cycle's floor has risen, the historic mean understates mid-cycle earnings and 37 times is an artefact of the window rather than a finding.
- 2026-10-06editedprice re-struck at the 2026-10-05 close of $98.73, up 21.0% since the thesis struck $81.61
Where we disagree with the market
The market believes
K&S is an AI packaging winner in an upcycle, which is why the shares have doubled in a year.
We believe
The data-centre demand for wire bonding is real, but the price pays for the top of a cycle that has halved sales before.
Coverage leads with the AI story, and so does the company. But the AI line is 8.9% of sales and loses money. What is booming is old-fashioned wire bonding, sold largely to three Chinese assemblers named in the filing.
We would be wrong if sales hold at $375M a quarter or more through fiscal 2027 and the AI line turns a profit.
What analysts say, and the evidence
Analysts positive
33%
of 3, 1 Oct
Four months earlier
33%
of 3, 1 Jun
The evidence under this - 5 claims
C04facthigh confidence
Advanced Solutions - the segment that holds K&S's thermocompression bonding, and the part of the company sold as artificial-intelligence packaging - produced $29.415M of Q3 2026 net revenue, 8.9% of the company total, at an operating loss of $12.553M.
C11facthigh confidence
Three customers were 39.0% of nine-month FY2026 net revenue - Haoseng Industrial 14.8%, Tianshui Huatian Technology 13.8% and Changjin Technology (Shanghai) 10.4%. A year earlier only Tianshui Huatian cleared the 10% disclosure line, at 10.1%.
against this
- Concentration rising in an upcycle is partly arithmetic rather than a change in the customer base: when three buyers expand capacity at once, everyone else's share of a much larger number falls below 10% and disappears from the table. The ten largest customers were 54.8% of FY2025 revenue against 53.6% in FY2024, so the broad measure barely moved before this year.
- Tianshui Huatian is China's third-largest outsourced assembler and a listed company - a substantial, identifiable end customer rather than an obscure one.
C15management claimmedium confidence
Management says the data-centre market relies on wire bonding at least as much as smartphones and personal computers do, because most chips in a data centre - networking, communications, power management and storage - are still packaged the traditional way.
against this
- It is an estimate by the company that sells wire bonders, offered without a source, and management prefaced it with 'we estimate'. No independent measurement of wire-bonded content in a data centre was found.
- It was volunteered in prepared remarks in the quarter the shares had fallen 30% from their high, which is when a company has most reason to reframe a legacy product as an artificial-intelligence one.
C09estimatemedium confidence
The FY2027 revenue estimate of $1,517.0M is only 1.1% above what the Q4 2026 outlook of $375M produces if simply repeated four times. The panel is not modelling further growth - it is modelling the current quarterly rate holding for a year, and then easing.
derived$375M multiplied by four is $1,500M; the panel's $1,517.0M is 1.13% above it. The FY2028 estimate of $1,461.7M is 3.6% below FY2027.
against this
- K&S's fiscal year ends in early October, so FY2027 runs from October 2026 to October 2027 and contains a seasonally weaker December quarter that management has said always carries some seasonality. Holding $375M flat across a year that includes that quarter therefore embeds growth in the other three, not a flat line.
- Two contributors is a thin panel and one of them may simply not have updated after the August print. A two-analyst mean is weak evidence about what anyone expects.
C10management claimhigh confidence
K&S guides Q4 2026 revenue to about $375M plus or minus $20M, gross margin of 48%, earnings of $1.29 a share and $1.42 a share excluding stock compensation and other adjustments.
sourceQ3 2026 results release, Exhibit 99.1 to Form 8-KprimaryQ3 2026 earnings call transcript, held August 6, 2026primary
against this
- K&S exceeded its own revenue outlook in each of the first three quarters of FY2026, so this guide is more likely a floor than a ceiling and the real run rate may be above $375M.
- It covers one quarter. Management guides one quarter at a time and explicitly declines to characterise fiscal 2027, so the guide carries no information at all about the period this thesis actually turns on.
What could go wrong, and how we would know
The risks
- Three customers in China. Haoseng, Tianshui Huatian and Changjin were 39.0% of nine-month sales, and two of them owe about half the money due to K&S.
- Missing the next technology. No disclosed high-bandwidth memory win, and its first hybrid bonding tool is only due in the first half of fiscal 2027.
- A patchy strategic record. A cancelled project, a closed business and three restructurings since March 2024, now under a chief executive who started in September.
The evidence under this - 6 claims
C11facthigh confidence
Three customers were 39.0% of nine-month FY2026 net revenue - Haoseng Industrial 14.8%, Tianshui Huatian Technology 13.8% and Changjin Technology (Shanghai) 10.4%. A year earlier only Tianshui Huatian cleared the 10% disclosure line, at 10.1%.
against this
- Concentration rising in an upcycle is partly arithmetic rather than a change in the customer base: when three buyers expand capacity at once, everyone else's share of a much larger number falls below 10% and disappears from the table. The ten largest customers were 54.8% of FY2025 revenue against 53.6% in FY2024, so the broad measure barely moved before this year.
- Tianshui Huatian is China's third-largest outsourced assembler and a listed company - a substantial, identifiable end customer rather than an obscure one.
C12facthigh confidence
Two customers held 49.8% of accounts receivable at July 4, 2026 - Haoseng Industrial 26.1% and Tianshui Huatian 23.7% - against a receivable balance of $329.498M, so roughly $164M is owed by two counterparties.
sourceForm 10-Q, quarter ended July 4, 2026primaryQ3 2026 results release, Exhibit 99.1 to Form 8-Kprimary
against this
- The receivable balance is not stretched in aggregate: $329.498M is 99.7% of the quarter's revenue, against $183.538M on $177.558M, or 103.4%, at the FY2025 year end. Days outstanding improved slightly while revenue nearly doubled, so this is concentration rather than a collection problem.
- No bad-debt charge or allowance increase is disclosed in the Q3 2026 10-Q.
C35inferencemedium confidence
K&S holds no disclosed position in high-bandwidth memory bonding. Management describes its memory business as focused on NAND 'for now' and its stacked-DRAM work as still in development, while the thermocompression sockets at the high-bandwidth memory makers sit with Hanmi and ASMPT.
derivedC34 and S04d, where management names NAND as the memory focus and describes vertical wire work on new forms of stacked DRAM as ongoing development; and S12, SK hynix placing a 44.2 billion won order with Hanmi for HBM4 thermocompression bonders in June 2026. No K&S filing or release names a high-bandwidth memory customer or programme.
against this
- Absence of disclosure is not absence of position: equipment makers routinely do not name memory customers, and K&S names none of its customers anywhere except in the concentration note, where they are identified only because accounting rules require it.
- Management does say the vertical wire team works closely with memory customers on new forms of stacked DRAM, so there is a programme, just not a disclosed win. And high-bandwidth memory is one socket - being outside it is not the same as being outside the artificial-intelligence build.
C32management claimmedium confidence
Management says it has increased wire bonding capacity fourfold from about two quarters earlier, plans to deliver a hybrid bonding tool to a customer in the first half of fiscal 2027, and expects the expanded Singapore Advanced Solutions plant to be complete in the second half of fiscal 2027.
sourceQ3 2026 earnings call transcript, held August 6, 2026primaryQ3 2026 results release, Exhibit 99.1 to Form 8-Kprimary
against this
- The release and the call give different dates for the same Singapore plant: the release says the second half of fiscal 2027, the prepared remarks say the company continues to target completion by the first fiscal half of 2027. Both are in documents published the same week.
- Quadrupling capacity into a surge is what a supplier does at the top of a cycle, and the cost of it stays in the base when the orders stop. Management also said it does not see a $450M quarterly run rate and that the company has supported $400M quarters before, which caps how much of the new capacity is being used.
C42facthigh confidence
The strategic record is worse than the guidance record. K&S ceased its Electronics Assembly equipment business in FY2025, cancelled Project W in FY2024 with inventory write-downs, ran restructuring programmes in March 2024, September 2024 and September 2025, and reported an FY2024 operating loss of $92.496M including $39.817M of impairment.
C33facthigh confidence
Dr Raj Talluri became president and chief executive on September 1, 2026, joining from Enovix where he was chief executive; the shares fell 9.5% in the session after the August 17, 2026 announcement. Lester Wong reverts to chief financial officer.
What we still do not know
- Whether $375M a quarter is a plateau or a peak. Management says it usually cannot see even six months ahead.
- How many orders K&S holds today. The last filed figure is from October 2025.
- Whether K&S can win a place in high-bandwidth memory, where rivals already hold the machine orders.
The evidence under this - 4 claims
C40management claimhigh confidence
Management will not characterise fiscal 2027 as a whole. It says visibility is usually not even as far out as the first half of fiscal 2027, calls the $1.5B revenue of 2021 and 2022 'extreme levels' produced by a pandemic, and will give more colour on the second half at the November call.
against this
- Declining to guide a full year is K&S's standing practice - it guides one quarter at a time - so the refusal carries less information than it appears to.
- In the same answers management said it feels 'pretty confident' the first half of fiscal 2027 will be strong and, in the prepared remarks, that it continues to anticipate above-average demand in fiscal 2027. The reticence is about the second half only.
C30facthigh confidence
The Q3 2026 10-Q states no backlog figure anywhere. The last filed backlog is $245.280M at October 4, 2025, up 65.1% from $148.585M a year earlier - less than one quarter of the $375M rate now being guided.
C35inferencemedium confidence
K&S holds no disclosed position in high-bandwidth memory bonding. Management describes its memory business as focused on NAND 'for now' and its stacked-DRAM work as still in development, while the thermocompression sockets at the high-bandwidth memory makers sit with Hanmi and ASMPT.
derivedC34 and S04d, where management names NAND as the memory focus and describes vertical wire work on new forms of stacked DRAM as ongoing development; and S12, SK hynix placing a 44.2 billion won order with Hanmi for HBM4 thermocompression bonders in June 2026. No K&S filing or release names a high-bandwidth memory customer or programme.
against this
- Absence of disclosure is not absence of position: equipment makers routinely do not name memory customers, and K&S names none of its customers anywhere except in the concentration note, where they are identified only because accounting rules require it.
- Management does say the vertical wire team works closely with memory customers on new forms of stacked DRAM, so there is a programme, just not a disclosed win. And high-bandwidth memory is one socket - being outside it is not the same as being outside the artificial-intelligence build.
C36factmedium confidence
SK hynix placed an order worth 44.2 billion won with Hanmi Semiconductor in June 2026 for thermocompression bonders for its sixth-generation high-bandwidth memory.
What would change our view
- Unattractive→AttractiveThe peak proves to be the new middle, at a mid-cycle multiple
It needs all of:
- ○Quarterly sales stay at $375M or more through all of fiscal 2027, so the level is a base, not one good year.checked at the next results, due by 18 Nov
- ○Its three largest customers fall below 25% of sales without sales falling, or China falls below 40% while sales still grow.checked at the next results, due by 18 Nov
- ○The whole company costs about 15 times its nine-year average operating profit of $131.8M, against about 37 times today.about ≈$39.92 today, 60% below the last close
- Reopens the researchThe memory win arrives
It needs:
- ○K&S names a high-bandwidth memory win at a memory maker, and its AI packaging line makes a profit on more than $200M of yearly sales.watched as it happens
What is coming
- November 2026 FY2026 Form 10-K: The first backlog figure since October 2025, so whether orders run ahead of the $375M quarterly rate, and the year's customer and China concentration.
Every dated event ahead
- November 2026 FY2026 Form 10-K: The first backlog figure since October 2025, so whether orders run ahead of the $375M quarterly rate, and the year's customer and China concentration.
- 18 November 2026 Q4 FY2026 results and the first outlook touching fiscal 2027: Whether the December quarter is guided at the $375M rate and first-half fiscal 2027 strength is repeated: a plateau, or a peak.
- January 2027 First proxy statement under the new chief executive: Whether fiscal 2027 pay adds a growth or thermocompression metric, or keeps adjusted profit and relative shareholder return only.
- April 2027 First hybrid bonding tool delivered to a customer: Whether K&S has a second advanced-packaging line at all, against a market where BESI already ships and delivers.
Since we wrote this
Written
6 Sep
Price the thesis read
$98.73
5 Oct
- 19 OctNewsBonus plan switched to operating profit and gross marginThe 13 October 8-K changed what executives are paid on for fiscal 2027; nothing pays on growth, so the view held.
- 6 OctPrice moveThe price rose 21%; the argument did not moveThe close moved from $81.61 to $98.73 since the price claims were struck, past the 12% drift line.
- 5 OctThesis reviewThe bet now names what it ridesThe owner asked that every bet name the macro trend behind it, after BWXT's bet was missing its chief executive's case for nuclear over gas.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 6 SepThesis writtenUnattractive, medium convictionThe research expected an overpriced AI story and found real demand, confirmed by a rival. What is overpriced is the cycle: a peak priced as the new normal.
Also checked 3 times with nothing new to act on.
Sources 36
Where this came from
Sources - 36
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2025-11-20Item 1 business and named competitors; segment and end-market revenue tables; backlog at both year ends; ten-largest-customer and China concentration; buyback totals; equity compensation; the Project W cancellation, the Electronics Assembly cessation and three restructuring programmesbacksC01C05C13C21C30C31C39C42C56C58C60
- S02primarycompany filing2026-08-06the customer concentration note naming three customers and their share of revenue and of receivables; China revenue share for the quarter and nine months; and the absence of any backlog figurebacksC02C03C04C11C12C13C30C52C56C58
- S03primarycompany release2026-08-05the income statement, balance sheet and cash flow statement for the quarter and nine months; the Q4 2026 outlook; the non-GAAP reconciliation; the buyback in the quarterbacksC02C10C12C25C26C27C28C32C41C42C61
- S04primarytranscript2026-08-06prepared-remarks figures with no filed home - end-market revenue and sequential growth, utilisation rates, the fourfold capacity increase; and the question section, where three analysts pressed on fiscal 2027, capacity and the wire-bonding data-centre claim
We estimate that the data center market relies on wire bonding technology at least as much as, if not more than traditional semiconductor markets such as smartphones and PCs.
backsC10C15C19C32C40 - S04bprimarytranscript2026-08-06the utilisation figures and the unusual order-book reach, in answer to an analyst rather than in prepared remarks
both the utilization rates are extremely high. In China, it's over 95%. End markets, both memory and general semi is around 90%. We are also seeing a lot of inbound POs even go extending into Q2. Usually, that doesn't happen for us.
backsC14 - S04cprimarytranscript2026-08-06the end-market split of the quarter, which appears in no filing - K&S discloses end markets annually
General semiconductor revenue increased by 52.6% sequentially to $227.2 million, driven by higher capacity and technology requirements for both ball bonding and Advanced Solutions segments. ... Memory shipment increased by 8.8% sequentially to $34 million ... Automotive and industrial demand improved by 9% sequentially to $24.2 million
backsC34 - S04dprimarytranscript2026-08-06management's own statement of where its memory business sits - NAND rather than high-bandwidth memory - which is the basis of the threat reading
we're focused on NAND for now and 40% of the NAND market now goes towards data centers
backsC34 - S05primarycompany release2025-11-19the fiscal 2025 full-year figures, the fiscal 2025 buyback of 2.4 million shares for $96.5M, and the Q1 2026 outlook that opened this fiscal yearbacksC25C41
- S06primarycompany release2026-08-17the appointment of Dr Raj Talluri as president and chief executive effective September 1, 2026, and the confirmation that Lester Wong reverts to chief financial officerbacksC33C57
- S07primarycompany filing2026-06-22the transaction code, share count, price and remaining holding for the June 2026 sale - read from the filing rather than from the pack's aggregatebacksC23C24
- S08primaryindustry data2026-07-14the size and growth of the whole market K&S sells into, measured by the industry's own trade association rather than by the company
Assembly and packaging equipment sales, which increased 20.8% in 2025, are projected to rise 9.6% to $6.7 billion in 2026
backsC17 - S08bprimaryindustry data2026-07-14the 2028 end point for the assembly and packaging category, and the total equipment figure it sits inside
test equipment reaching $20.8 billion and assembly and packaging equipment reaching $8.6 billion
backsC17 - S08cprimaryindustry data2026-07-14the total equipment market the assembly and packaging line sits inside, which is what makes the 9.6% against 23.2% comparison possible
Global sales of total semiconductor manufacturing equipment by original equipment manufacturers (OEMs) are forecast to reach a record high of $165.9 billion in 2026
backsC17 - S09primarycompany release2026-03-03the size of the thermocompression bonding market, from the company that leads it - the one number K&S's own Advanced Solutions case has to fit inside
It revised its estimated TCB Total Addressable Market (TAM) from about US$760 million in 2025 to US$1.6 billion in 2028, a CAGR of 30 percent.
backsC18 - S09bprimarycompany release2026-03-03independent corroboration, from the other half of the wire-bonding duopoly, that AI data-centre demand lifts conventional bonding rather than only advanced packaging. ASMPT's own segment is named SEMI, which is not the trade association of the same name
Heightened power management needs in AI data centre kept utilisation at leading global IDMs elevated, supporting SEMI's mainstream business, while SMT won additional orders to meet rising data-transmission requirements for base stations.
backsC16 - S09cprimarycompany release2026-03-03how much of that small market the incumbent leader intends to hold, which bounds what is available to a challenger
The Group continues to target a 35 to 40 percent share of the TCB market, supported by its industry leading technologies and deep engagements with its broad AI customer base.
backsC18 - S10primarycompany release2026-04-23the attacker's own account of hybrid bonding demand - the die-attach method that displaces wire bonding at the highest interconnect densities
Orders of € 269.7 million were up 7.7% vs. Q4-25 primarily due to a significant increase in bookings for hybrid bonding systems.
backsC20 - S11secondaryvendor data2026-09-06Research pack, KLIC, 2026-09-06content/KLIC/research-pack.jsonprice, enterprise value, returns, the analyst estimate panel, the insider-filing aggregate, FINRA short interest, and the filer's own XBRL segment tablebacksC03C04C05C08C24C26C33C41
- S12secondarynews2026-06-08SK hynix orders HBM4 TC bonders from Hanmi Semiconductor Hanmi Semiconductor / SK hynixevidence that the high-bandwidth-memory bonding socket sits with a Korean thermocompression supplier rather than with a wire-bonding one. Reporting of a disclosed contract rather than the disclosure itself, so it is recorded as secondary
SK hynix orders new equipment worth 44.2 bln won for next-gen HBM4
backsC36 - S13primarycompany filing2025-12-17the confirmation test for the only competitor K&S names that files with the SEC: Nordson's own description of its Advanced Technology Solutions segment is dispensing and test and inspection, not bondingbacksC31
- S14primarycompany filing2026-08-17what the incoming chief executive is paid for: a $14M new-hire grant and a $6M annual grant, both majority performance share units vesting on total shareholder return relative to a semiconductor peer group over three years, plus base salary and bonus targetbacksC43C55
- S15primarycompany release2026-07-29the leader's half-year account of thermocompression demand in logic and in high-bandwidth memory, and the state of its hybrid bonding platform (sampling with logic and memory customers); read through ir_fetch
Notably, in July 2026 the Group received new bulk orders for more than 50 of its C2S TCB tools from OSAT customers. ... In memory, even as the Group continued to secure repeat orders from HBM manufacturers, the timing of customers’ new tool purchase decisions remains dependent on HBM4 product rollout schedules.
backsC44 - S16primarycompany release2026-07-29ASMPT - Secures significant wins for more than 50 chip-to-substrate thermo-compression bonding tools ASMPTthe leader's claim to the chip-to-substrate logic socket K&S's fluxless tools sell into, and its restated 2028 market size; read through ir_fetch
This major milestone affirms ASMPT’s position as the primary supplier and Process of Record (POR) for C2S TCB applications as a powerful enabler for next-generation bonding for larger compound dies. ... With growing AI investment and increasing packaging intensity, ASMPT estimates that the TCB Total Addressable Market will expand beyond US$1.6 billion by 2028.
backsC44 - S17primaryindustry data2025-08-18an independent analyst's 2030 size for thermocompression (11.6% a year from 2025) and hybrid bonding equipment (21.1% a year). yolegroup.com answered 202 to ir_fetch and press_fetch; the authorised reprint was read through ir_fetch
TCB will reach $936 million by 2030, fueled by integration in memory and AI platforms, Hybrid bonding will grow to $397 million, with fine-pitch, high-density interconnects essential for advanced 3D integration.
backsC45 - S18primarycompany release2026-06-18the hybrid bonding leader's installed-base forecast to 2030 in units, its timing for hybrid bonding in high-bandwidth memory, and its own reading that large logic packages require fluxless thermocompression; PDF fetched with ir_fetch's client and read with pdftotext
Estimated 1,150 to 2,200 systems cumulatively by 2030 vs. 960 to 2,000 systems Investor Day 2025 ... HBM4e is insertion point for Hybrid Bonding in HBM stacking ... Hybrid and TCB will coexist ... “Chip On Wafer” and “On Substrate” interconnect requires fluxless TCB due to increased size of components/interposers
backsC46 - S19secondarynews2026-07-07the memory makers' reported delay of hybrid bonding, the JEDEC height relaxation behind it, and the knock-on demand for wide thermocompression bonders; reporting of Korean press, so secondary; read through ir_fetch
Although hybrid bonding had once been expected to debut with HBM4, both companies ultimately continued using conventional thermo-compression (TC) bonding. Industry forecasts now point to 16-high HBM4E (the seventh-generation HBM) as the earliest likely adoption point.
backsC47 - S20secondarynews2026-06-09a reported price per high-bandwidth-memory thermocompression bonder and units per order, and SK hynix's reported DRAM wafer capacity plan to 2030; the same event as S12, so the same independence group; read through ir_fetch
A separate report from Hankyung notes that the contract value represents around 7.7% of Hanmi Semiconductor’s 2025 revenue. Based on an estimated unit price of roughly 3 billion won per TC bonder, the deal is believed to cover around 15 units of equipment, the report adds.
backsC48 - S21primarycompany release2026-03-24the company's own account of where its thermocompression share sits (leading-edge logic), the offer into memory, the capacity spend and the FY2026 growth target; read through ir_fetch
These unique fluxless capabilities, combined with an array of material handling configurations, have driven share gains in leading-edge logic and are now being offered for leading-edge memory applications. ... The Company is investing in capital expenditures to expand TCB production and continues to anticipate its TCB business will grow approximately 70% sequentially in fiscal year 2026.
backsC49 - S22secondaryvendor data2026-10-01FMP key-metrics-ttm, quote and analyst-estimates for ASMPT (0522.HK), BESI (BESI.AS) and Hanmi Semiconductor (042700.KS), fetched 2026-10-01 Financial Modeling Prepthe bonding peers' trailing company value to sales and the consensus revenue and operating profit that put a forward multiple on each, through scripts/api/fmp.py
0522.HK evToSalesTTM 2.7764823119410615; BESI.AS evToSalesTTM 20.6645326301732; 042700.KS evToSalesTTM 46.03890756592997
backsC50 - S23primarycompany filing2026-01-22beneficial ownership of directors and officers as of December 8, 2025 and the holders above 5%; the executive officers and their tenure; the fiscal 2025 cash incentive plan's metrics, scales and 39.98% payout; the relative-TSR PSU designbacksC53C54C57
- S24primarycompany filing2023-11-16the FY2021 to FY2023 income statements (R&D and selling, general and administrative by year) and the February 2023 acquisition of Advanced Jet AutomationbacksC56C58
- S25primaryother2025-06-30Automotive Electronics Council - AEC-Q006 Rev-B, qualification requirements for components using copper wire interconnections Automotive Electronics Councilthe automotive industry's own rule for when copper-wire qualification results can be reused between assembly sites: same wire bond equipment and settings, same capillary type
For the following processes, equipment incl. process settings shall be the same: Wire bond process
backsC59 - S25bprimaryother2025-06-30Automotive Electronics Council - AEC-Q006 Rev-B, Appendix 2 tools and the two qualification options Automotive Electronics Councilhow long a copper-wire requalification runs: either the AEC-Q100 stresses with analytical tests, or twice the AEC-Q100 stress duration
Option 2: 2X stress duration (based on AEC -Q100/101 requirements) followed by ATE
backsC59 - S25cprimaryother2026-03-24Automotive Electronics Council - AEC-Q100 Rev-J1, stress test qualification for integrated circuits Automotive Electronics Councilthe base requalification rule and the 1,000-hour and 1,000-cycle stress durations that AEC-Q006's 1X and 2X are multiples of
Requalification of a device is required when the supplier makes a change to the product and/or process
backsC59 - S26secondaryvendor data2026-10-05Financial Modeling Prep - KLIC daily closes 2019-2026 and dated analyst rating changes Financial Modeling Prepthe closes behind the 2020-2022 round trip and the 2026 drawdown, and the dates Needham and B. Riley cut their ratings in 2022backsC60C61
- S27primarycompany filing2026-10-13the fiscal 2027 bonus measures: operating profit and gross margin at 50% each in place of net income and operating margin, and a 15% individual-goal share for the executive teambacksC55