Kinik:The Diamond Disk Maker With ~70% of TSMC's Newest Nodes

Kinik Company (1560 TT) makes the diamond disks that recondition, or "dress", the polishing pads used in chemical mechanical planarization (CMP), the step that flattens each layer of a chip during fabrication. It is the second largest supplier of these disks in the world, behind 3M (MMM US).

Its largest business by revenue is wafers. Chipmakers run test wafers through their machines to check them, and Kinik cleans and re-polishes those used wafers so they can be used again. It also supplies new test and specialty wafers. It also makes industrial grinding wheels. Founded in 1953, the company is based in Yingge, New Taipei. Revenue was NT$8.15bn in 2025 and we forecast NT$10.29bn in 2026.

Kinik has taken share of the world diamond disk market in every year but one since 2020. By the company's own count its share rose from 15.3% in 2020 to an estimated 24.7% in 2026, while its disk revenue rose from US$41m to US$105m. 3M remains the largest supplier at about 45%, down from above 70%. Saesol Diamond of Korea is third. The three supply more than 80% of the market.

The share is higher at the most advanced nodes. A node is a generation of chip-making technology, named by a size such as 3nm, and a smaller number is newer. We estimate that Kinik supplies about 70% of the diamond disks its largest customer uses at 3nm and more than 80% at 2nm. That compares with about 50% across all of that customer's nodes. Kinik does not publish its share by node and does not name its customers. We deduce that the largest is TSMC (2330 TT), the world's largest contract chipmaker.

Kinik's own disclosure points the same way. In 3Q26, sales to its largest customer for nodes of 5nm and below and for advanced packaging, which joins several chips in one package, were 62% of disk revenue. It has also begun shipping disks for 1.4nm. Separately, we estimate that Kinik supplies a substantial portion of the disks used at 18A, the most advanced node of Intel (INTC US). Kinik has not confirmed that.

The nodes where Kinik's share is highest are the ones that become the volume nodes two to three years later. Hence its share across all nodes should rise as production moves to them, provided it keeps its position at each new node.

Four things keep customers with Kinik. 1) Each disk is built for one customer's polishing step: grit size, diamond pattern and bonding method differ by customer and by step. 2) Kinik's disks go into a node while it is still in development, about two years before volume production in the case of 1.4nm. A supplier that is not qualified then must wait for the next node. 3) Customer data show fewer defects with Kinik's protective film, which matters more as chip features shrink. 4) Kinik works with the makers of the pads and of the slurry, the abrasive liquid used in polishing, that are used alongside its disks.

A more advanced chip is polished more times. Each new node has historically needed about 25% more CMP steps than the one before, from one or two at 0.8 micron to more than 20 at 2nm. The next node, 1.6nm, adds about 20% to 25% more again, because moving the power wiring to the back of the wafer adds polishing steps there. Those counts exclude advanced packaging.

Every polishing step wears the pad, and the disk is what restores it. Disks are used up and replaced, so demand follows the number of wafers processed and the number of steps on each, not the number of machines sold.

This reaches Kinik's revenue and margin in three steps. 1) More steps per wafer mean more disks per wafer. 2) On our estimates Kinik's share is highest at the newest nodes, so it captures a larger part of the added volume. 3) Those nodes use more of Kinik's advanced disks, which earn a higher margin than conventional disks once they reach the same production scale. A new product family starts at a lower margin while its volume is small.

Advanced packaging could add to this, but most of it is not yet in revenue. Stacked chips need wafers ground very thin before bonding, and Kinik's grinding wheels for that use are in development. Hybrid bonding, which joins stacked chips directly, needs a polishing step on every bonded surface and is already used in TSMC's 3D chip stacking.

Zero One Investment Research toured Kinik's diamond disk plant at Shulin, New Taipei, and met Tony Pai, vice president and acting spokesperson. Many thanks to him for the visit. We saw the main disk families, the automation projects under way on the production floor, and the capacity plan to 2034.

Every polishing pad in a fab is dressed by a disk, so a fab uses disks continuously. Kinik makes several kinds.

The disks Kinik supplies to its largest customer's fabs in Arizona and Japan are made in Taiwan. The company said it supplies overseas fabs from its Taiwan plants and serves China separately, from a plant in Yixing. Chip production is spreading to the US and Japan, but this part of the supply chain still sits in Taiwan.

Kinik has raised its disk capacity plan. It now targets more than 80,000 disks a month by the fourth quarter of 2026 and 110,000 by the fourth quarter of 2027, up from more than 70,000 and 90,000 to 100,000. It was already at 75,000 in October. Management told us that total disk capacity could exceed 150,000 a month by 2030, against a previous plan of more than 120,000 in Taiwan and more than 10,000 in Yixing. The board has approved buying a site in New Taipei to add disk capacity and to make diamond wheels for advanced packaging. At our visit the company forecast about 250,000 a month by 2034.

Demand has been growing faster than Kinik plans for. Kinik plans on disk demand growing 20% to 25% a year because it does not know its customers' future demand in detail. Disk revenue rose 36% on the year in 3Q26. The raised plan moves toward that demand. We understand why a supplier keeps its guidance moderate, and from what we saw, we believe the capacity Kinik builds after 2027 is likely to be higher than the plan.

Our forecasts reflect that view. We forecast disk revenue growing 40% in 2027, 32% in 2028 and 25% in 2029. The raised plan supports 2027: reaching 110,000 by the end of the year lifts average capacity by about 50%. From 2028 our forecasts sit above the plan, which on a straight line to 150,000 by 2030 lifts average capacity by about 23% in 2028 and 11% in 2029. Disk revenue in 2029 would be 2.3x our 2026 figure, from about twice the 2026 capacity and a larger share of higher-priced advanced disks. These numbers are ours. Kinik has not endorsed them and gives no financial forecast.

Automation is how Kinik gets there. A disk for an advanced node is made to a standard that takes a technician years to learn, and the supply of such technicians is limited. The plants cannot go from 75,000 disks a month to 250,000 by hiring alone. Kinik's newest disk family, equaDia, is designed for machine assembly and inspection. The target is at least 80% automation on that line, against under 50% on the current advanced disk line.

Kinik showed us several automation projects on the floor, each replacing a step still done by hand, and told us that AI is part of this work. We see it as an example of how AI can help a manufacturer automate steps that used to depend on skilled technicians. For Kinik the gain is capacity as much as cost.

Wafers face a harder limit. Capacity rose from 300,000 pieces a month in 2025 to 400,000 in June 2026 and is expected to reach 430,000 by the third quarter of 2027, the maximum in the two existing plants. Kinik is raising prices on selected wafer products by about 10% from the fourth quarter of 2026. A third plant of about 300,000 is in planning and not confirmed. Our wafer forecasts need that plant to be producing by 2029, with higher prices and a richer product mix carrying growth until then.

Kinik's gross margin was 42.2% in 3Q26, against 32.4% in 3Q25. Operating margin rose from 15.9% to 27.0% over the same period. In the first nine months of 2026, revenue rose NT$1,501.9m on the year and gross profit rose NT$1,033.5m, so each extra dollar of revenue brought 69 cents of gross profit.

Two things explain it. 1) Disks earn more than the company average: Kinik says conventional disks earn above 60% gross margin and the disk unit about 50%, and disks are the fastest-growing unit. 2) Plants carry fixed costs, so added volume in disks and wafers raises the margin on every unit. The wafer plants ran full at 360,000 pieces a month in 2Q26, and the capacity added since has gone to higher-priced reclaimed wafers for advanced nodes.

We forecast gross margin of 40.2% in 2026, 45.0% in 2027 and 47.0% in 2028. That assumes each extra dollar of revenue in 2027 brings 66 cents of gross profit, slightly less than in 2026, as depreciation on new plant rises. Capital expenditure was NT$1,219.2m in the first nine months of 2026, more than in all of 2025.

Kinik has one convertible bond outstanding, and the share price is above its conversion price. The bond was issued in September 2026 with a face value of NT$1.0bn and raised NT$1,236.8m, because investors paid above face value for the right to convert. It pays no interest and can be converted into shares from 4 December 2026 at NT$708.3 a share. Full conversion would add 1.41m shares, 0.9% of the shares in issue.

Our earnings per share are diluted for it. We use 154.2m shares from 2027, which includes the shares from the bond and from employee stock options, and we assume the bond converts in 2027. If it does not convert, Kinik repays NT$1.0bn in September 2029. It held NT$3,148.6m of cash at 30 September 2026.

We set the target by holding today's multiple and moving it one year forward. At NT$726 the shares trade on 36.3x our 2027 diluted earnings per share of NT$20.01. In a year, investors will be looking at 2028. The same multiple on our 2028 figure of NT$25.66 gives NT$931, and we set the target price at NT$930, 28% above the share price for long-term holders of the shares.

The method assumes the multiple neither rises nor falls. If it slipped to 33x as growth slows, the shares would be worth about NT$847. At the high in early October the shares traded at NT$1,040, or 52x our 2027 figure. The target is 46x our 2027 earnings per share and 30x our 2029 figure of NT$30.53.

Kinik's raised capacity plan alone supports a value close to our target. On that plan, revenue grows 25% in 2027 and 13% in 2028, slightly more than we forecast in 2027 and less in 2028. We estimate 2028 earnings per share of about NT$24.8 on that basis, and the same multiple gives about NT$899. The rest of the target rests on our view that capacity keeps rising faster than the plan after 2027.

Kinik trades well above other listed makers of chip materials: 54x trailing earnings against a peer median of 21x. The target of 46x our 2027 earnings is outside the peer range. The reason is growth. We forecast Kinik's earnings per share growing 55% in 2026, 42% in 2027 and 28% in 2028. If Kinik's growth slowed, its multiple would fall toward the peers'.

Three things support owning the shares over years. 1) Kinik is second to 3M overall but, on our estimates, the leader at the newest nodes of its largest customer, which is where the market is growing. 2) Suppliers are qualified node by node, years ahead, so a position is hard to take away once won. 3) Its disks are made in Taiwan and shipped to fabs in the US and Japan, so it is part of the chip supply chain those countries are trying to secure.

The shares have fallen 30% from the high of NT$1,040 on 5 October, and are still 136% above the NT$308 of a year ago. Kinik's most recent monthly revenue, NT$927.5m in October, was its highest on record and 30.9% above October 2025.

We would change our view if Kinik lost its position at a new node, if its largest customer delayed its node schedule, or if the capacity we forecast did not arrive.

We forecast revenue growth of 23% in 2027 and 19% in 2028, with operating margin rising from 24.5% in 2026 to 32.0% in 2028. For 2026 we use the nine months reported plus fourth-quarter revenue of NT$2,810m, 4% above the third quarter. That needs November and December to average NT$941m, against NT$927.5m in October. Operating expenses grow more slowly than revenue. We forecast capital expenditure of NT$2.0bn in 2027 and NT$2.3bn in 2028 to pay for the capacity above.

Five risks matter most. 1) One customer. Sales to the largest customer for nodes of 5nm and below and advanced packaging were 62% of disk revenue in 3Q26, so a change in that customer's node schedule or choice of suppliers would reach Kinik directly. 3M, the larger supplier, could also compete harder for the newest nodes. 2) Our capacity view. From 2028 our forecasts need more capacity than the company has announced. On its raised plan alone, revenue growth would be closer to 13% in 2028 and 8% in 2029.

3) The wafer ceiling. Without a third plant, wafer volume stops growing after 2027. 4) Downturns still reach Kinik. In 2023 revenue fell 7.6%, gross margin fell from 36.6% to 30.4% and net profit fell 32%. 5) Disk production is concentrated in Taiwan. The full income statement, balance sheet and cash flow statement follow.