ASIC Chip Foundries Explained: TSMC, Samsung, SMIC

Every Bitcoin ASIC begins life as a design, but it becomes real hardware only inside a semiconductor foundry. Having ASIC chip foundries explained clarifies one of the least visible yet most decisive parts of the mining supply chain: the small group of fabrication plants capable of producing the advanced chips that power modern miners. Foundry access, process nodes, and capacity allocation shape what hardware exists, when it ships, and how efficient it is. This explainer is factual and contains no investment advice.

What a chip foundry is and why it matters

A foundry is a semiconductor fabrication plant: an enormous, ultra-clean facility that turns silicon wafers into finished chips through hundreds of precise manufacturing steps. ASIC designers such as Bitmain, MicroBT, and Canaan do not own foundries. They are fabless: they design the chip, then contract a foundry to manufacture it. This split between design and fabrication is standard across the entire chip industry.

The reason it matters for mining is concentration. Only a few foundries in the world can produce chips at the advanced process nodes that competitive miners require. That scarcity means mining-chip supply depends on a handful of companies, and those companies prioritize their largest, most reliable customers. The downstream effects of this dependency, on availability and price, are the subject of the explainer on ASIC hardware shortage cycles.

The major foundries that make mining chips

Three names dominate the conversation. TSMC, the Taiwan Semiconductor Manufacturing Company, is the largest and most advanced contract foundry in the world and has historically fabricated chips for leading mining-hardware makers. Samsung Foundry, the contract-manufacturing arm of Samsung, is the principal alternative at advanced nodes and has produced mining chips as well. SMIC, the Semiconductor Manufacturing International Corporation based in China, serves as a domestic option, though it lags the leaders on the most advanced nodes and faces export-control constraints on the most advanced fabrication equipment.

The choice of foundry is consequential. A design fabbed at a leading-edge node will generally be more power-efficient than the same logic at an older node. That efficiency translates directly into joules per terahash, the figure that decides whether a miner remains profitable as difficulty rises. The competitive landscape among the designers who depend on these foundries is mapped in the ASIC manufacturer landscape for 2026.

How designers compete for capacity

Foundry capacity at advanced nodes is finite and booked far ahead. Mining-chip designers compete for it against the entire electronics industry: smartphone processors, automotive chips, and AI accelerators all want the same leading-edge wafers. During capacity crunches, mining orders can be deprioritized behind larger, steadier customers. This competition for fab slots is one of the structural reasons miner production cannot respond quickly to demand spikes, a dynamic traced through the full ASIC supply chain.

The competition is uneven by design. A foundry prefers customers who book steady, predictable volume year after year over those whose orders swing wildly with a volatile coin price. Mining demand is famously cyclical, so mining-chip designers are not always the most attractive customers from a fab’s perspective. The largest designers mitigate this by committing to capacity well ahead and by being significant enough buyers to command attention, but smaller players can find themselves at the back of the queue when capacity tightens. This is one reason the mining-chip business has consolidated around a few large designers who can credibly commit to foundry volume.

Why a single foundry dependency is risky

When a chip designer relies heavily on one foundry, it inherits that foundry’s constraints and risks. A capacity shortage, a yield problem on a new node, a natural disaster, or a geopolitical disruption at a single fab can stall an entire generation of miners. The semiconductor industry has seen all of these, from earthquakes affecting Taiwanese production to droughts threatening the water-intensive fabrication process. For mining specifically, any hiccup at a key foundry ripples downstream into delayed launches and tighter supply months later.

Designers manage this by qualifying chips at more than one foundry where possible, but dual-sourcing advanced nodes is expensive and not always feasible, since each foundry’s process differs and a design must be adapted to each. The practical result is that mining-hardware availability is hostage to the health and priorities of a very small number of fabs. Understanding this concentration helps explain why miner launches sometimes slip and why supply can stay tight even when demand is strong, a pattern that feeds directly into the swings covered in the explainer on hardware shortage cycles. It also underlines why where finished machines are assembled, examined in the piece on where Bitcoin ASICs are made, is a separate question from where their chips are fabricated.

What process nodes mean

Process node, expressed in nanometers, is shorthand for a manufacturing generation. Smaller numbers, such as 5nm or 3nm, indicate more advanced processes that pack more transistors into the same area and run them at lower power. Older nodes, like 16nm or 28nm, are cheaper and more widely available but less efficient.

For mining, node choice is a direct lever on competitiveness. Modern high-efficiency units rely on advanced nodes to reach the low joules-per-terahash figures that keep them profitable. The relationship between node, efficiency, and a machine’s useful economic life ties directly into how long a unit stays worth running, a topic covered in the explainer on ASIC miner lifespan. Buyers comparing efficiency across the current SHA-256 lineup can use the Bitcoin mining hardware hub as a reference.

Why geography and policy shape foundry access

Foundries are not evenly distributed. The most advanced fabrication is concentrated in Taiwan and South Korea, with growing investment in the United States and Japan. This concentration makes the mining supply chain sensitive to geopolitics, export controls, and trade policy. Restrictions on advanced fabrication equipment, for example, limit how quickly some regions can close the node gap.

For mining specifically, this means a chip designer’s options can narrow or widen with policy shifts that have nothing to do with cryptocurrency. A designer cut off from a leading foundry must either use an older node, accepting worse efficiency, or wait for capacity elsewhere. These constraints feed directly into the geography of where finished miners are assembled, which is examined in the piece on where Bitcoin ASICs are made.

How foundry economics reach the buyer

Most buyers never see a foundry, but they feel its effects in three ways. Availability: when fab capacity is tight, lead times stretch. Price: advanced-node wafers are expensive, and that cost flows into the unit price. Efficiency: the node a machine was built on largely determines its power draw and therefore its operating cost over years of use.

The efficiency gap between foundry generations is also why mining hardware depreciates the way it does. A unit built on a leading-edge node when it launched becomes a prior-generation unit the moment a competitor ships a chip on a newer, more efficient node. The earlier machine still works, but its joules-per-terahash figure now looks worse by comparison, and its resale value falls accordingly. In this sense, the foundry roadmap, the schedule on which TSMC and Samsung introduce new nodes, indirectly sets the depreciation clock for the entire mining-hardware market. Buyers who understand the node a unit was built on can better anticipate how quickly it will be outclassed.

Coin Web Mining operates downstream of all of this, as an independent reseller on a slim margin over distributor cost. It does not design chips or contract foundries; it sources finished units and resells them. But the foundry layer explains much of why current-generation hardware is priced and stocked the way it is, and why efficiency gaps between generations are so pronounced. The operating-cost consequences of those efficiency differences are quantified in the electricity cost analysis, and live availability appears in the catalog.

What to watch in the foundry layer

For observers tracking the industry, a few signals matter. New node introductions by TSMC and Samsung tend to precede a new generation of more efficient miners. Capacity expansions or new fab construction can ease future shortages. Export-control changes can reshape which designers have access to which nodes. And foundry pricing trends feed into miner costs a year or more later, given the long pipeline.

None of this should be read as a prediction or as guidance to buy at any moment. It is context. Understanding that a small number of foundries gate the entire mining-hardware industry makes the recurring shortages, price swings, and efficiency leaps far more predictable. Industry data services and trade press, including Hashrate Index and The Block, track how foundry developments ripple into mining-hardware availability, while manufacturer pages document the nodes their chips are built on.

References

• Antminer chip and product specifications — Bitmain
• Whatsminer hardware specifications — MicroBT
• Semiconductor and mining-supply reporting — The Block
• Mining-hardware efficiency and market data — Hashrate Index

The foundry layer explains the efficiency gaps between generations. To compare current units on power and hashrate, the Coin Web Mining catalog lists live specs and lead times, with escrow on first orders.