TechChain Insights Site Visit: TSMC Supplier Kinik to Triple Diamond Disk Capacity by 2034

Kinik holds ~25% of the world diamond disk market, with higher share at the most advanced chip nodes

Kinik has an estimated 24.7% market share of the world CMP diamond disk market in 2026, up from 15.3% in 2020. Disk revenue rose from US$41m to US$105m over the same period. 3M (MMM US) remains the largest supplier at about 45%, down from above 70%, and Saesol Diamond of Korea is the third. Those three companies supply more than 80% of the market.

Market share even higher for the most advanced chip nodes.

At its largest customer TSMC (Taiwan Semiconductor Manufacturing) (2330 TT) (Please note that Kinik does not disclose specific customer details however Zero One has deduced TSMC to be their largest client), Kinik estimates that it supplies about 70% of the diamond disks used at 3nm and above 80% at 2nm. This compares to about 50% across all nodes.

Kinik is not only critical for TSMC's advanced nodes. For Intel's leading edge 18A node we estimate that Kinik supplies a substantial portion of the disks Intel uses.

We believe Kinik's market share will expand significantly in the next 2-3 years

The nodes where Kinik holds its highest market share are the leading edge ones that become the volume nodes for TSMC and Intel two to three years later. Hence Kinik's blended market share should rise as the node mix moves up the complexity curve.

Moreover, each next advanced chip node has historically required about 25% more CMP steps than the one before, from one or two layers at 0.8 micron to more than 20 at 2nm. Kinik uses that trend to estimate future nodes, whose exact layer counts are not yet known. What this means is that the number of diamond disks consumed per wafer should keep rising with each new advanced node, likely amplifying Kinik's share gains if the company can maintain its leading position in the advanced nodes.

How more complex chips turn into revenue and margin for Kinik

The link runs in four steps. 1) Each new node has needed about 25% more CMP layers, so more disks are used per wafer. 2) Kinik's share is highest at the newest nodes, so it captures a larger part of that added volume. 3) The newest nodes use more of Kinik's advanced disks, Pyradia today and equaDia next, 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. 4) Capacity growth to 2030 and beyond, helped by automation, is what allows Kinik to carry that volume. Hence node complexity drives both Kinik's volume and its product mix.

Examples of different diamond disk products

Conventional disk

Comprised of thousands to more than ten thousand small diamond grits bonded to a steel plate in a regular grid. Grit size and the bond differ by customer and by polishing step. This is the family Kinik brands DiaGrid.

Pyradia, the advanced disk

Comprised of far fewer and much larger diamonds, each a square-cut tip whose height is set individually so that every tip meets the pad at the same level. The tips are arranged in patterns tuned to the node. The leading edge chip nodes, such as 2nm, A16 and beyond, use this disk on more and more of their CMP layers.

CVD disk

This disk type is not an abrasive at all. A thin film of diamond is grown by chemical vapor deposition over a patterned substrate, which gives a fine, even surface. It is used on soft pads for the final polish step of wafer processing.

The dark and the bright

The dark finish on some disks is a protective film over the metal bond, which Kinik calls EDP. It began as a way to keep nickel and iron out of the slurry for customers that specify metal-free consumables, and the company says customer data now show better defect performance with the film as well.

Automation is a major initiative of Kinik

Kinik Company (1560 TT)'s newest product family, equaDia, is specifically designed to be assembled and inspected by machine.

Kinik's target is at least 80% automation on this new product family's line, as compared to under 50% automation on its current advanced disk line. Kinik showed Zero One several automation projects under way, each one replacing a step still done by a technician with a machine.

Automation will unlock higher capacity growth potential.

A diamond disk for an advanced node is made to a standard that takes years for a technician to learn, and the supply of technicians with that experience is finite. If capacity is to go from 75,000 disks a month to 250,000, the plants cannot staff that growth by hiring alone. Success in automating key parts of the manufacturing process offers the potential to increase Kinik's capacity growth potential by removing availability of staff limitations.

Why customers stay with Kinik

Kinik's advantage is more than the diamond itself. It lies in precision manufacturing, defect control and close work with customers on each new chip node.

Each disk is built for one customer's polishing step. Grit size, diamond pattern and the bonding method differ by customer and by CMP step. Hence a disk is designed for one customer's process rather than sold off the shelf.

Defect control. Kinik's EDP film was first developed to keep metal out of the slurry. Customer data now show better defect performance with the film as well, which matters more as chip features shrink.

Working on the next node years ahead. Kinik's disks go into a customer's node while it is still in development. Kinik reported its first sales of disks for TSMC's 1.4nm (A14) node this quarter, about two years before that node is expected to reach volume production. At Intel, Kinik won its position at 18A during the node's development, years before volume production.

A total solution. Kinik works with the makers of the slurry and pads used alongside its disks, so that the whole CMP step performs as one system for the customer.

We believe these positions are hard for a new entrant to take. A supplier that is not qualified while a node is in development must wait for the next node to compete. New suppliers, including those in China, remain much smaller than Kinik and 3M.

2026 production capacity exceeding original targets

Kinik production capacity in October exceeded its original year-end 2026 target. Diamond disk capacity was previously expected to reach 70,000 units a month by 4Q26, from 60,000 in 2Q26. The company has already reached 75,000 a month run rate in October.

Further production capacity expansion in process for 2027

Adding tools to its two plants. Kinik is adding tools to its two existing plants which is expected to take production capacity to 90,000 to 100,000 units a month in 2027.

Disks Kinik supplies to TSMC's fabs in Arizona and Japan are made in Taiwan. This shows how Taiwan remains a critical point for production capacity despite efforts by TSMC and others to replicate their chip supply chain to the U.S. The company said it supplies overseas fabs from the Taiwan plants and serves China separately.

Source: Company disclosures and Zero One site visit. Zero One Investment Research.

Wafers business segment: Major production capacity also in process

Kinik's wafer business segment, which reclaims and supplies 12 inch test and specialty wafers to the fabs, is the larger of Kinik's two semiconductor businesses at 43% of second quarter revenue. Its capacity was 300,000 pieces a month in 2025, reached 400,000 in June 2026 across the two wafer plants at Zhubei and Zhunan, and is expected to reach 430,000 in 2027. Beyond that, Kinik is planning a third wafer plant, in the Hsinchu area, that could add about 300,000 pieces a month three to five years out. The plan is still in progress and not yet confirmed. The company wants it to be a brand new factory with much higher automation, to reduce the labour needed per wafer and raise throughput. If it is built, wafer capacity could rise to around 730,000 a month, against 300,000 in 2025.

Advanced packaging: what is in production and what is in development

Advanced packaging could become a further growth driver for Kinik, but most of it is not yet in revenue.

Stacked chips, such as TSMC's CoWoS, 3D stacked chips and high bandwidth memory (HBM), need wafers ground very thin before they are bonded. Kinik already sells grinding wheels for wafer thinning, but to silicon carbide and power chip customers and, in small amounts, for conventional packaging. Wheels for advanced packaging are under development.

Hybrid bonding, which joins stacked chips directly, needs a CMP step on every bonded surface. It is already used in TSMC's 3D chip stacking. For HBM, memory makers still use other bonding methods, and Kinik is not sure when hybrid bonding will arrive there.

Source: Zero One site visit. Zero One Investment Research.

Valuation and peer multiples

At NT$858 Kinik trades at 72x trailing twelve month EPS of NT$11.93 and 58x the NT$14.84 of 2026 consensus earnings.

The peer median sits at 23x trailing and 20x forward. Price to book of 14.4x is the highest in the set and more than five times the peer median of 2.6x. On enterprise value to EBITDA, which strips out the cash on each balance sheet, Kinik trades at 50.7x trailing and 36.6x forward against a peer median of 13.3x and 10.6x, and is again the highest in the set.

The closest comparison on the trailing multiple is Entegris at 73x. Entegris Inc (ENTG US) is an American CMP consumables supplier whose product position most resembles Kinik's. However, wafer reclaim specialists trade at lower multiples, for example Phoenix Silicon Intl (8028 TT) at 44x and RS Technologies (3445 JP) at 14x. Kinik's multiple thus sits closer with a peer CMP consumables supplier rather than wafer reclaim companies.

Source: Company filings and consensus estimates. Zero One Investment Research.