Bitmain vs MicroBT: The Reshoring War Over Bitcoin’s Miners

For most of the last decade, comparing Bitmain and MicroBT came down to a single number: joules per terahash, the electricity a machine burns to produce a unit of hashing. Whoever shipped the lower figure won the order. That contest has not stopped. Bitmain’s Antminer S23 Hydro now runs at 9.5 J/TH, and Bitdeer’s SEALMINER A4 has pushed the record to 9.45. But in 2026 the number that decides a US miner’s purchase order is not printed on any spec sheet. It is the tariff line stamped at the customs house.

The backdrop is a market that leaves little room for waste. Bitcoin trades near $84,000, with a market capitalization around $1.69 trillion and roughly 20.09 million coins in circulation. Network difficulty sits at 132.76 trillion, near record territory, and hashprice, the daily revenue a miner earns per unit of hashrate, hovers around $40 per petahash per second per day, up from the low $30s a month earlier. When revenue per machine is this compressed, a cost the size of an import tariff is not a rounding error. It is the difference between a rig that pays for itself and one that does not.

The politics arrived first. In June 2024, Donald Trump told a room of mining executives that “We want all the remaining Bitcoin to be MADE IN THE USA!!!” The line read as applause bait at the time. By 2026 it had hardened into policy, delivered through a wall of reciprocal tariffs that reshaped how every mining rig enters the country. Both of the Chinese-founded companies that build almost all of the world’s Bitcoin miners are now racing to stamp those same three words on their hardware.

This is the third front of the Bitmain versus MicroBT war. The first was chip design, the race to a smaller, faster, cooler ASIC. The second was price, the discounting that turned a seller’s market into a buyer’s market as the last cycle’s oversupply cleared. The third, and the one that will decide the US market in 2026, is geography: where a machine is fabricated, where it is assembled, and which border it crosses before it reaches a data center in Texas or Georgia.

A Duopoly Built on Borrowed Silicon

Start with how concentrated this market is. According to Cambridge research cited by Cointelegraph, Bitmain accounts for roughly 82 percent of global ASIC production, MicroBT for about 15 percent, and Canaan for around 2 percent. Three companies, all founded in China, supply nearly every machine securing the Bitcoin network. That is a striking single point of dependence for a system whose entire value proposition is decentralization.

It is a dependence that proof-of-work cannot escape. Bitcoin’s security budget is bought with electricity and specialized silicon; the more efficient the fleet, the cheaper each terahash of protection. That is the opposite of proof-of-stake, where security comes from bonded capital rather than a hardware arms race. An Ethereum solo staker needs a modest computer and 32 ETH, not a fresh generation of ASICs every eighteen months. For Bitcoin, by contrast, the machine is the moat, and the companies that make the machines hold structural power over the miners who buy them.

Here is the twist that shapes everything downstream: neither Bitmain nor MicroBT actually manufactures its own chips. They design the ASICs, then hand the blueprints to Taiwan Semiconductor Manufacturing Company, or TSMC, which etches them on advanced process nodes in Taiwan. Both firms are fabless. Their rigs are assembled from silicon they do not fabricate, in a country neither controls. That borrowed foundation is exactly what tariffs and export controls have learned to squeeze.

How Washington Redrew the Map: The 2026 Tariff Stack

Before 2024, importing a mining rig into the United States meant paying a standard duty of about 2.6 percent. Freight and customs were an afterthought next to the sticker price. That era is over. After the Trump administration’s reciprocal-tariff regime took effect, the duty a rig pays now depends almost entirely on where it was last substantially assembled.

According to The Block, machines assembled in Southeast Asia, in Indonesia, Malaysia or Thailand, the very countries Bitmain and MicroBT had shifted production to in order to dodge China duties, now carry a total levy of 21.6 percent, effective 7 August 2025. Rigs of Chinese origin fare far worse: a stack of the base duty, Section 301 duties, the reciprocal tariff and a China-specific premium adds up to as much as 57.6 percent. The same machine, depending only on the flag on its final assembly line, can cost a fifth more, or well over half more, to bring into the country.

The person who has to price all this is Ethan Vera, chief operating officer of the mining-services firm Luxor Technology. His verdict was blunt. “At 21.6% tariffs, the U.S. is now one of the least competitive jurisdictions to bring machines in, and miners are looking at Canada,” he told The Block, adding: “We generally foresee a dampening in growth in the U.S. which will lead machines to head to overseas markets with more favorable import tariffs.” For an administration that wants Bitcoin made in America, that is an awkward result: the tariffs meant to reshore production first make it more expensive to mine on American soil at all, unless the machine itself is already American.

What a Tariff Does to the Price of a Hash

To see why this reshaped procurement, translate a tariff into the unit miners actually budget in: dollars per terahash of capacity. A modern flagship rig sells for somewhere in the region of $18 to $20 per terahash. Take a round $20 per TH as an illustration and run it through the tariff table.

The spread is enormous. A rig that lands from China costs more than half again what the same machine costs if it is bolted together in a US facility. On a 10,000-unit order, that gap runs into the tens of millions of dollars before a single hash is produced. And it compounds with the economics of a thin market: at a hashprice near $40, a 9.5 J/TH machine breaks even on power at roughly $0.175 per kWh, while a 13.5 J/TH machine breaks even near $0.123. Efficiency still sets the operating floor, but the tariff sets the entry price, and in 2026 the entry price is where deals are won and lost.

One caveat runs through the entire reshoring story, and it bites hardest right here. A zero on the tariff line applies to the finished unit assembled in the United States. It does not necessarily zero out duties on the imported components and raw materials that go into that unit. Assembly is not fabrication, and the chips at the heart of every rig still come from abroad. The made-in-America pitch shrinks the tariff bill; it does not erase it.

Bitmain’s Answer: An American Factory

Bitmain moved first and moved big. It opened its first US production line in December 2024, and has since laid out plans for a full American factory, with chip assembly slated to begin in early 2026 and full-scale manufacturing by year end, according to Cointelegraph. The company has said it will site a new US headquarters in Texas or Florida and hire about 250 workers in the first phase, in roles from manufacturing training to facility maintenance.

Irene Gao, Bitmain’s global business chief, framed the American push as “a unique opportunity,” describing the goal as faster delivery and repair for US customers. The subtext is defensive as much as opportunistic. Bitmain earns a large share of its revenue from American miners, and every point of tariff on an imported Antminer is a point of margin handed to a rival that can offer a domestic alternative. Building in the United States is how the market leader protects the market it already leads.

The move also fits a pattern the company has been signaling for a while: Bitmain wants to be seen as a partner to American energy and industrial policy, not a foreign vendor to be taxed at the dock. Whether Washington accepts that framing, given what is happening on the export-control side, is a separate question this piece returns to below.

MicroBT’s Answer: Made in the USA, and Earlier Than You Think

MicroBT is the smaller company, but on reshoring it can credibly claim to have seen the wall coming first. The firm began investing in US manufacturing back in 2021, working with a domestic partner, Synos, years before Washington raised duties on Chinese-made rigs. That head start became a marketing spearhead. On its official channel, MicroBT declared, “Made in the USA, Supported in the USA,” promising “tariff-free mining rigs, faster delivery, and superior service” from US-based facilities.

It backed the slogan with supply. In December 2024, Luxor signed a $131.4 million purchase-and-options agreement with MicroBT for thousands of WhatsMiner units, an arrangement structured partly around domestic assembly to soften the tariff blow. By September 2025, MicroBT had gone further, opening a dedicated US online shop for domestically built rigs, with output of more than 10,000 units a month at a yield rate up to 99 percent, pickup from a Delaware warehouse within 24 hours, and repair hubs on both coasts.

This is the underdog’s supply-chain jiu-jitsu. MicroBT does not always win on joules per terahash, as the spec comparison below makes clear. But if a US miner is choosing between an imported market-leading rig carrying a 21.6 percent duty and a slightly less efficient machine that ships tariff-free from a warehouse a day’s drive away, the choice is no longer obvious. Under a tariff wall, a few joules of disadvantage can be cheaper than a fifth of the purchase price.

Assembly Is Not Fabrication: The TSMC Chokepoint

Strip the reshoring headlines down and one fact stays untouched: the most advanced Bitcoin mining chips in the world are made in Taiwan. Bitmain, MicroBT and Bitdeer all rely on TSMC to fabricate their ASICs on leading-edge process nodes. A plant in Texas that screws heatsinks onto boards and slots them into chassis is real, useful industrial activity, but it sits downstream of the step that actually matters. The wafer is still Taiwanese.

That makes TSMC the true single point of failure in Bitcoin’s supply chain, more than any one rig maker. A serious disruption to Taiwan’s foundries, whether from a natural disaster, an export restriction or a geopolitical shock, would throttle new-rig production for all three companies at once, no matter where their assembly lines sit. Reshoring assembly diversifies the last mile of the supply chain. It does nothing about the chokepoint at the source.

It also puts a ceiling on the made-in-America story. Until a company can fabricate mining ASICs on US soil, and none does today at the volumes required, an American Antminer is an American-assembled Antminer with a Taiwanese brain. That distinction is exactly what a customs classification, or a future export rule, can be written to exploit.

Export Controls Enter the Picture: The Sophgo Entity List

Tariffs are a tax. Export controls are a wall, and Bitmain has already hit one. In January 2025, the US Commerce Department added Sophgo, a chip-design firm affiliated with Bitmain, to its Entity List, after a TSMC-made chip matching a Sophgo design was found inside Huawei’s Ascend 910B artificial-intelligence processor, according to the South China Morning Post. Firms on the Entity List cannot receive US-origin goods or technology without a license that is generally denied. Sophgo said it “has never been engaged in any direct or indirect business relationship with Huawei.”

The listing does not ban Antminers, and Bitmain is a separate legal entity from Sophgo. But it exposes the risk. Bitmain’s ambitions run well beyond Bitcoin: like its rivals, it has eyed the far larger market for AI and high-performance computing silicon, the same market where US export controls are tightest. The moment a China-linked chip designer touches the AI supply chain, it enters a regulatory arena with sharper teeth than any mining tariff.

This is where the hardware war and the compute wars converge. The infrastructure being built to verify AI workloads and settle machine-to-machine payments, examined in our look at opML and the agent economy, runs on exactly the accelerators export policy is designed to fence off. For a mining-hardware maker, moving into AI silicon is the obvious growth path and the fastest route into Washington’s crosshairs. MicroBT, which has stayed closer to its mining core, carries less of this particular exposure, another quiet advantage in a market where policy is now part of the product.

The Specs Still Matter: 2026’s Flagship Rigs

None of this means efficiency stopped mattering; it means efficiency is now read through a second lens, namely where the machine comes from. Here is where the three main flagship lines stand in 2026: Bitmain’s Antminer S23 Hydro, MicroBT’s WhatsMiner M70 series, and Bitdeer’s SEALMINER A4. The final column translates joules into economics with an illustrative cash-breakeven power price at a hashprice near $40.

Two things jump out. First, the efficiency crown is no longer Bitmain’s alone. Bitdeer’s SEALMINER A4 Ultra Hydro, launched in April 2026 on the company’s SEAL04 chips, reaches 9.45 J/TH, a hair below the S23’s 9.5, and it was reported as an industry efficiency record at debut. Second, MicroBT’s best machines sit around 12.5 to 13.5 J/TH, roughly 3 joules behind the leaders. On a pure spec sheet, that is a real gap. Filtered through the tariff table, it is a gap MicroBT can close, or even reverse, with a domestic assembly line.

The Third Company Nobody Predicted: Bitdeer

The most interesting player in the 2026 hardware war owes its existence to a corporate divorce. Bitmain was co-founded in 2013 by Jihan Wu and Micree Zhan. A power struggle tore the company apart, and in an early-2021 settlement Zhan kept Bitmain, the Antminer brand and the Antpool mining pool, while Wu left to build Bitdeer. Five years on, Wu’s company is not just a miner but a chipmaker in its own right, trading on Nasdaq as BTDR and headquartered in Singapore.

Bitdeer’s SEAL chips have turned it into a genuine third force. The SEALMINER A2 shipped around 16.5 J/TH; the A3 Pro Hydro reached 12.5 J/TH; and the A4 Ultra Hydro, on SEAL04 silicon, now leads the field at 9.45 J/TH. For the geopolitics that define this piece, Bitdeer carries a structural advantage the two Chinese giants do not: it is a US-listed, Singapore-based company. In a market where a machine’s passport increasingly sets its price, that is not a footnote.

The feud that birthed Bitdeer was not the only courtroom fight in this industry’s past. MicroBT itself was founded in 2016 by Yang Zuoxing, a former Bitmain chip-design lead who worked on the S7 and S9 Antminers before leaving. Bitmain later pursued him: after an earlier patent suit was dismissed in 2018, it filed a trade-secret case at the Beijing IP Court in April 2021, seeking about 99 million yuan, roughly $15 million, over the S7 and S9 designs. These blame wars, where former partners contest who owns an idea after a company fractures, are a recurring feature of crypto, from mining rigs to the finger-pointing that follows a protocol exploit, a dynamic we traced in the contested autopsies of crypto’s biggest hacks.

The Economics That Make Tariffs Bite

Why does a hardware surcharge matter so much now, when miners have absorbed rising costs for years? Because the margin cushion is thin. Difficulty, the measure of how hard it is to find a block, sits at 132.76 trillion, near all-time highs, and it keeps grinding upward as new machines come online. Every uptick divides the same block rewards among more hashing power, lowering revenue per rig. Bitcoin’s difficulty adjustment is one of the most elegant mechanisms in the protocol, if not a flawless one; it still carries a long-known timewarp quirk that developers are working to close, as we explained in our piece on the difficulty bug and the 2026 fight to fix it.

On the revenue side, miners lean heavily on the block subsidy rather than transaction fees, which have stayed low for long stretches. That raises a structural question about the network’s long-term security budget, the total reward that pays miners to protect Bitcoin, which shrinks with every halving. We unpacked that tension in Taproot’s fee paradox. The short version: as the subsidy falls, either fees must rise or the dollar cost of each unit of security must fall, and cheaper, more efficient machines are one of the few levers that push that cost down.

Put the two sides together and the tariff math becomes decisive. With hashprice near $40 and difficulty at record levels, a US miner buying an imported rig at a 21.6 percent premium starts the race several lengths behind a competitor who bought the same class of machine domestically. In a bull market, that gap is survivable. In a drawdown, it is the line between operators who keep the lights on and operators who go dark.

Winners, Losers, and the Miner’s Real 2026 Decision

So how should a US miner actually choose in 2026? The old checklist, lowest joules per terahash and best price per terahash, has grown a third line: lowest landed cost after duties. The decision now runs roughly like this.

• Total cost of ownership, not sticker price. Add the tariff to the purchase price, then spread it over the machine’s productive life alongside power and hosting. A cheaper-to-buy imported rig can be the more expensive rig once the duty is amortized.
• Energy cost stays king. No amount of tariff engineering saves an operator paying above-market power. This is the counterpoint from Leo Lu, chairman and chief executive of BitFuFu, who argues US miners can offset pricier equipment with cheap and increasingly renewable power, built on local partnerships across Oklahoma, Texas and Colorado.
• Delivery and service. Domestic assembly is not only about duties; it shortens lead times and repair cycles. A rig sitting in a Delaware warehouse today can beat a better rig stuck in transit for a quarter.
• Political durability. Tariff schedules and entity lists can change with an executive order. A supplier whose advantage rests on a single policy is a riskier bet than one whose supply chain is genuinely diversified.

Run that framework and the likely winner is not a single company but a posture: whoever can put a competitive machine on US soil, tariff-free, with fast service. Bitmain has the best chips and the deepest resources to build domestically. MicroBT has the earliest US manufacturing footprint and the most aggressive made-in-America pitch. Bitdeer has the efficiency crown and the cleanest corporate passport. The miner’s job is to match those profiles to a specific site, power contract and time horizon, not to crown a universal champion.

What Could Break the Reshoring Story

Several things could scramble this picture, and any serious buyer should price them in.

• Policy reversal. Tariffs imposed by executive action can be lifted, adjusted or widened the same way. A truce with Beijing could erase the domestic advantage quickly; an escalation could push duties even higher.
• Component-level duties. If customs authorities tighten the rules on imported boards and chips inside a US-assembled rig, the tariff-free claim narrows. Assembly-only reshoring is legally thinner than full fabrication.
• More entity-list actions. The Sophgo case shows how fast an affiliate can be caught in export-control machinery. A direct action against a major manufacturer, however unlikely today, would be far more disruptive than any tariff.
• The Taiwan question. Everything upstream still runs through TSMC. Any shock to Taiwan’s foundries would hit all three companies at once and make the whole made-in-America debate moot until fabrication itself is diversified.
• Gray-market chip flows. High duties create incentives for rerouting, transshipment and misdeclared origin. Enforcement will chase the arbitrage, and honest buyers can get caught in the compliance crossfire.

None of these is a reason to wait. They are reasons to treat supplier diversity, not just spec-sheet superiority, as a core part of a mining operation’s risk management.

The Bottom Line

For a decade, the Bitmain versus MicroBT rivalry was a story about silicon: who could squeeze more hashing out of a watt. That contest is closer than ever, with Bitmain and Bitdeer trading blows below 10 J/TH and MicroBT pressing from just behind. But the decisive battleground has shifted from the lab to the customs house. In 2026 the question that moves a purchase order is no longer only how efficient a machine is, but where it was made and what it costs to bring home.

That shift rewards different strengths than the old one. It favors the company with a US assembly line as much as the one with the best chip, and it hands a quiet edge to Bitdeer, a US-listed outsider born from Bitmain’s own breakup. It also exposes an uncomfortable truth beneath the made-in-America slogans: as long as every rig’s brain is fabricated in Taiwan, the reshoring is only skin deep. For miners, the takeaway is practical. Read the spec sheet, then read the customs schedule, and never assume the cheapest machine to buy is the cheapest machine to run.

Frequently Asked Questions

Is Bitmain or MicroBT better in 2026?

It depends on the site. Bitmain’s Antminer S23 Hydro leads on efficiency at 9.5 J/TH, while MicroBT’s best machines run around 12.5 J/TH but reach US buyers with a strong domestic-assembly and service pitch. After US tariffs of up to 57.6 percent on imported rigs, the cheaper machine to own is often the one that ships tariff-free from a US line, not simply the one with the lowest joules per terahash.

How much are US tariffs on Bitcoin mining machines?

As of August 2025, ASIC miners assembled in Southeast Asia carry a total US duty of about 21.6 percent, while rigs of Chinese origin can face up to about 57.6 percent once base duties, Section 301 tariffs, the reciprocal tariff and a China premium are stacked. The standard duty before this regime was roughly 2.6 percent. Machines assembled in the United States can avoid duty on the finished unit.

Are Bitmain and MicroBT miners made in the USA?

Increasingly, the final assembly is. Bitmain opened its first US production line in December 2024 and plans a full American factory, and MicroBT has run US assembly since 2021 and opened a US online shop for domestically built rigs in 2025. But the ASIC chips inside every rig are still fabricated by TSMC in Taiwan, so made-in-America today means American-assembled, not American-fabricated.

What is the most efficient Bitcoin miner in 2026?

The most efficient flagship in 2026 is Bitdeer’s SEALMINER A4 Ultra Hydro at 9.45 J/TH, narrowly ahead of Bitmain’s Antminer S23 Hydro at 9.5 J/TH. Both broke the sub-10 joules-per-terahash barrier, a milestone the previous generation of machines could not reach.

Why was Sophgo added to the US Entity List?

In January 2025, the US Commerce Department added Sophgo, a chip-design firm affiliated with Bitmain, to its Entity List after a TSMC-made chip matching a Sophgo design was found inside Huawei’s Ascend 910B AI processor. Sophgo denied any business relationship with Huawei. The action does not ban Antminer sales, but it shows how Bitmain’s AI-chip ambitions are exposed to US export controls.

By Daniel Vega, mining and semiconductors correspondent at HOGE Wire.