Bitmain vs MicroBT: Inside Bitcoin Mining’s ASIC Duopoly

Every time a new Bitcoin block is found, roughly every ten minutes, the machine that found it was almost certainly built by one of two companies. Bitmain, the Beijing-born incumbent that has shaped mining since 2013, and MicroBT, the Shenzhen challenger founded by a former Bitmain chip designer, together account for the overwhelming majority of the application-specific integrated circuits (ASICs) grinding SHA-256 hashes around the clock. Industry trackers put Bitmain at roughly four-fifths of the global rig market and MicroBT in a clear second place, with Canaan a distant third; together the three control something close to 97 percent of new supply, according to data compiled by outlets such as TheMinerMag.

That concentration would matter in any industry. In Bitcoin it matters more, because the hardware these two firms ship is the physical substrate of the network’s security budget. When you read that Bitcoin’s hashrate sits near 900 exahashes per second, you are really reading a tally of Antminers and Whatsminers plugged in from Texas to the Gulf. Choosing between them is a decision that shapes a mining operation’s cost base for years, and, in aggregate, it shapes how decentralized the supply chain behind Bitcoin’s proof of work really is.

2026 is a pointed moment to make the comparison. Bitcoin trades around $64,000, down by roughly half from its late-2025 record, per Fortune’s daily price data. Hashprice, the daily revenue a miner earns per unit of hashrate, has been stuck near $31 to $32 per petahash per day, according to the Hashrate Index roundup. And network difficulty actually fell year over year for only the second time in Bitcoin’s history, a sign of how hard the squeeze has bitten, as Blockspace reported. Thin margins, a redrawn US tariff map, and an export blacklist that has ensnared a Bitmain affiliate have turned a simple hardware question into a bet on efficiency, reliability, and geopolitics.

Why one of two companies built almost every miner

The Bitcoin mining hardware market is one of the most concentrated in technology. A handful of firms design the chips, all of them lean on the same foundry to fabricate those chips, and just two brands dominate the shelves. The rough split most analysts cite is Bitmain around 80 percent of the ASIC rig market, MicroBT around 15 percent, and Canaan around 2 percent, with everyone else fighting over the scraps. Whatever the exact figures in any given quarter, the top-heavy shape is stable: nearly every terahash of honest work securing Bitcoin runs on an Antminer or a Whatsminer.

This is why the rivalry is not a niche procurement story. The machines do only one thing, and they do it relentlessly: they compute SHA-256 hashes to propose and validate blocks on a chain whose rules, from the block subsidy to soft forks like Taproot, the miners enforce by choosing which software to run. A duopoly on that hardware concentrates an enormous amount of influence over how, and where, Bitcoin’s security is produced. Here is how the two contenders stack up at a glance before we get into the details.

Bitmain: the incumbent that industrialized mining

Bitmain was founded in 2013 by Jihan Wu, an early Bitcoin evangelist who had helped translate the Bitcoin whitepaper into Chinese, and Micree Zhan, an engineer who designed the company’s first mining chips. Their Antminer line quickly outran the field. The S9, released in 2016, became the workhorse that carried Bitcoin through the 2017 bull run; later generations, the S17, S19, and S21, each pushed efficiency lower and cemented Bitmain as the default choice for industrial miners. Alongside the hardware business, Bitmain built Antpool, still one of the largest mining pools, giving the company a presence at both ends of the mining stack.

Scale gave Bitmain pricing power. As Colin Harper, editor-in-chief of Blockspace Media, noted in Forbes, the company “has historically had the upper-hand when negotiating prices with mining clients.” For most of the last decade, if you wanted the most efficient machine available in volume, you bought Bitmain and you accepted the terms. The firm’s dominance is not an accident of branding; it reflects a repeated ability to ship the leading chip a generation ahead of rivals, then manufacture it at a scale no competitor could match.

The founder wars that split Bitmain in two

Bitmain’s dominance nearly unraveled from the inside. A bitter power struggle between its two co-founders erupted in 2019, when Micree Zhan was ousted while traveling, and escalated into a public and legally messy fight for control of one of crypto’s most valuable private companies. The saga ended in early 2021 with a settlement that CoinDesk reported saw Zhan buy out the camp of shareholders aligned with Wu for around $600 million, regaining control of the company.

The split cleaved Bitmain’s empire in two. Micree Zhan kept the crown jewels: the Antminer hardware business, the Antpool mining pool, and the domestic operations. Jihan Wu walked away with the overseas mining and cloud-hosting assets, which he built into Bitdeer, along with BTC.com. That divorce matters for this comparison in two ways. First, it left Bitmain focused squarely on manufacturing under Zhan’s control. Second, as we will see, it planted the seed of a third ASIC contender, because Wu’s Bitdeer has since started designing its own mining chips, turning a boardroom feud into a fresh competitive threat.

MicroBT: the challenger born from a defection

MicroBT exists because of a talent dispute inside Bitmain. Yang Zuoxing was a processor-design lead at Bitmain who contributed to the flagship S7 and S9 Antminers. He left in mid-2016 after being refused an equity stake, and within weeks founded MicroBT, launching the Whatsminer line that would become Bitmain’s most credible rival. By the 2019 to 2021 period, Whatsminer machines had earned a following among large North American miners for their build quality, and MicroBT had grown into a genuine second pole in a market Bitmain once owned outright.

That rise was contested in the courts and, at one point, in a police station. Bitmain first sued MicroBT and Yang in 2017 over patent infringement, but the Beijing IP Court dismissed the case in 2018 and revoked the patent Bitmain had relied on. In late 2019, Yang was detained by police in Shenzhen amid the dispute, on suspicion of embezzlement, as CoinDesk reported at the time. Bitmain then sued again in April 2021, seeking roughly 99 million yuan (about $15 million) and accusing Yang of taking corporate secrets after completing R&D on the S7 and S9, a characterization Yang disputed, per The Block. Yang has consistently maintained that he ended his Bitmain contract before founding MicroBT. The legal history is a reminder that the two firms are not just competitors; they share DNA, and a decade of grievance.

Antminer vs Whatsminer: the 2026 flagships

The hardware is where the rivalry gets concrete. Bitmain used the World Digital Mining Summit to unveil the Antminer S23 generation, which it began shipping in the first quarter of 2026. The headline model, the water-cooled S23 Hydro, is rated at about 9.5 joules per terahash, making it the first mass-market Bitcoin miner to break the 10 J/TH barrier, with an air-cooled S23 near 11 J/TH and a rack-mounted 3U variant that pushes past a petahash of hashrate in a single chassis, according to Harper’s Forbes breakdown. Bitmain says the S23 delivers roughly 69 percent more hashrate and is about 41 percent more power efficient than the S21 it replaces.

MicroBT answered with the Whatsminer M70 series, introduced at the Bitcoin MENA conference in Abu Dhabi, targeting around 12.5 J/TH at the top of its hydro range and spanning roughly 12.5 to 14.5 J/TH across air, hydro, and immersion units. In a Hashrate Index teardown, the air-cooled M70S is rated near 13.5 J/TH at about 250 TH/s. The table below lines up the current flagships. Treat the exact hashrate and efficiency figures as nameplate specifications that vary a little by firmware and cooling; the pattern that matters is consistent.

Read the table and the story becomes clear. At the very top of the range, Bitmain holds an efficiency lead of roughly 2 to 3 J/TH; the S23 Hydro at about 9.5 J/TH is the most efficient machine a buyer can source in volume in 2026. But MicroBT has closed most of the gap that existed a generation ago, when its M60-class units sat near 18 J/TH against Antminers already under 15. The flagship fight is now measured in single-digit joules, not double.

Density is the other half of the story the spec sheet tells. A single S23 Hyd 3U chassis packs more than a petahash of hashrate into a rack slot, and MicroBT’s M79S answers with roughly a petahash of its own. That matters because a modern mining facility is constrained less by floor space than by power delivery and heat rejection: packing more terahashes behind each megawatt of electrical service lets an operator wring more revenue from a fixed grid connection. The generational jump from the S21 XP and M60 class, which sat in the low-to-mid teens of joules per terahash, to today’s single-digit and low-teens flagships is what makes fleet upgrades pencil out even when Bitcoin itself is trading sideways.

Efficiency is (almost) the only number that matters

Why obsess over a couple of joules per terahash? Because in a mature mining operation, electricity is the dominant recurring cost, and efficiency sets how much revenue each megawatt-hour of power can generate. Hashrate Index’s own tiering makes the point in dollars: at current hashprice, machines under 14 J/TH earned roughly $112 of revenue per megawatt-hour in early August 2026, those in the 19 to 25 J/TH band earned about $61, and legacy units above 25 J/TH earned closer to $42, per the Hashrate Index roundup. The efficient machine does not just cost less to run; it survives price drawdowns that shut the old machine off entirely.

You can turn efficiency into a breakeven power price with a simple back-of-envelope formula. At a hashprice H (in dollars per petahash per day) and an efficiency e (in joules per terahash), the electricity price at which a machine merely covers its energy bill is about H divided by (24 times e) dollars per kilowatt-hour. Plugging in the current hashprice near $31.73 produces the illustrative breakeven prices below. These ignore hardware cost, hosting fees, and pool fees, so they are a ceiling, not a promise; for the full picture of cash cost versus all-in cost, see our field guide on how to read a Bitcoin miner’s margins.

The spread is the whole argument. A miner paying 8 cents per kilowatt-hour can run an S23 Hydro comfortably but would bleed cash on an old S19 XP. That is why efficiency, and by extension the Bitmain versus MicroBT choice at the top of the range, translates directly into which machines stay powered when the market turns. MARA chief executive Fred Thiel has framed the same reality bluntly, calling mining “a zero-sum game” in which “margins compress, and the floor is your energy cost,” in comments to CoinGeek.

Efficiency tells only half the profit story; the other half is hashprice, the market value of the work these machines do. Hashprice folds the Bitcoin price, network difficulty, and transaction fees into a single number, and when it sits near $31 to $32 per petahash per day, as it has through the summer of 2026, even efficient rigs earn thin returns and inefficient ones earn nothing. That is the backdrop against which the Bitmain versus MicroBT decision is made: in a high-hashprice boom almost any machine prints money and buyers chase availability, while in a squeeze like this one the efficiency gap between an S23 and a three-year-old rig is the difference between staying on and switching off.

Air, hydro, and immersion: the cooling arms race

Efficiency gains have forced a change in how machines are cooled, and this is one area where the two firms have converged. For years, air cooling with a pair of loud fans was the default. As chips packed more transistors and drew more power per chassis, air struggled to keep them from throttling, so both Bitmain and MicroBT now sell three cooling formats: traditional air, hydro (water piped directly through the unit), and immersion (the whole board submerged in a dielectric fluid). Hydro and immersion let a machine sustain full clocks without thermal throttling, which is why the most efficient flagships on both sides, the S23 Hydro and the Whatsminer M73S+, are water-cooled.

The tradeoff is infrastructure. Hydro and immersion demand a facility built for liquid: manifolds, dry coolers, pumps, and fluid handling that an air-cooled shed does not need. That favors large, well-capitalized operators over hobbyists, and it raises the switching cost of moving between vendors, because a site plumbed for one manufacturer’s hydro form factor is not trivially repurposed for another’s. Cooling choice, in other words, is becoming as strategic as chip choice, and both Bitmain and MicroBT are racing to lock operators into their respective high-density ecosystems.

Reliability, resale, and firmware: the buyer’s real math

Spec sheets sell machines, but total cost of ownership keeps operators loyal. This is where MicroBT’s pitch is sharpest. Reviewing the M70S for Hashrate Index, Luxor Technology research associate Kaan Farahani wrote that the machine emphasizes “operational stability and fleet-level reliability through build quality and predictable performance,” targeting miners who “prioritize long-duration infrastructure performance over short-term efficiency gains.” Whatsminers have a long-standing reputation for tolerating heat, dust, and voltage swings, and many large fleets keep buying them precisely because uptime, not the last joule, drives realized returns.

Bitmain counters with ubiquity. Because Antminers are everywhere, they enjoy the deepest secondary market, the most third-party repair expertise, and the widest firmware support, including tuning tools from firms like Braiins and LuxOS that let operators underclock or overclock to chase the best power price. Resale liquidity is a real financial variable: a machine you can offload quickly when a cycle turns is worth more than an equally efficient one that is hard to sell. The honest summary is that Bitmain tends to win the spec race and the resale market, while MicroBT competes on durability, service, and, increasingly, supply-chain location. Neither advantage is decisive on its own, which is exactly why the duopoly persists.

The buyer’s market has sharpened these secondary factors into real leverage. When rigs were scarce, an operator took whatever warranty, lead time, and payment schedule the manufacturer offered. With demand soft in 2026, buyers can negotiate longer warranties, staggered deposits, and firmer delivery dates, and they weigh a vendor’s repair turnaround and spare-parts availability as seriously as its efficiency rating. For a large fleet, a machine that is one joule worse but ships with local US service and next-week delivery can beat a marginally better rig stuck in a customs queue behind an unpredictable tariff bill.

Boiled down, the choice tends to sort by power price and priorities. Operators with very cheap, abundant power and purpose-built liquid-cooling infrastructure usually reach for Bitmain’s most efficient hydro units, where every fractional joule compounds across tens of thousands of machines. Operators facing higher or more variable power costs, or running mixed air-cooled sites, often value the Whatsminer’s durability, simpler serviceability, and US-made availability more than a two-joule spec advantage. Neither is a wrong answer; they optimize for different constraints, which is exactly why both firms keep selling everything they can build.

Price wars and the buyer’s market

The April 2024 halving cut the block subsidy to 3.125 BTC and, with hashprice grinding lower ever since, flipped the balance of power between manufacturers and miners. For most of Bitmain’s history, new rigs sold out on allocation and buyers competed for slots. In 2026 that has inverted. As Colin Harper put it in Forbes, compressed margins and higher import costs have “flipped a seller’s market into a buyer’s market.” Bitmain has responded with price cuts and payment installments to stimulate demand, moves that would have been unthinkable when it could name its terms.

TheMinerMag analysts flagged a subtler shift in how the newest hardware is used. Rather than fueling raw network expansion the way past upgrades did, the S23 generation is likely to be deployed to replace older machines inside existing fleets, marking, in their words, “a new era where efficiency gains are about surviving squeezed margins, not just chasing growth,” per CoinDesk. For MicroBT, the buyer’s market is an opening: when operators are cost-sensitive and skeptical of paying a premium for the last joule, a slightly less efficient but cheaper and hardier machine becomes easier to justify. The macro backdrop only sharpens the calculus, and readers weighing a purchase around key catalysts may find our September FOMC market-reaction playbook a useful companion.

The tariff earthquake: made in China, assembled in America

The single biggest force reshaping the Bitmain versus MicroBT contest in 2026 is not a chip; it is a tariff schedule. A wave of US trade measures since early 2025 has pushed import duties on mining hardware sharply higher, whether the machines ship from China directly or from assembly hubs in Malaysia, Thailand, and Indonesia. With ASIC miners also swept into steel, aluminum, and copper measures as metal-heavy derivative products, the landed cost of a rig has become a moving target that can swing a purchase decision more than a two-joule efficiency edge.

Both firms responded by moving final assembly onshore. Bitmain established its first US production line in December 2024, and vice president Irene Gao told Forbes the company is “working to expand its manufacturing capabilities in the United States.” TheMinerMag documented the logistics of that pivot, reporting that Bitmain shipped roughly 187 tons of Antminer components to a Delaware affiliate over a matter of weeks to feed US assembly, according to TheMinerMag. MicroBT went further in its messaging, marketing “Made in the USA, Supported in the USA” rigs from local manufacturing and repair centers that it says deliver “tariff-free mining rigs, faster delivery, and superior service,” on its official WhatsMiner account. Canaan, the number-three player, has run early-stage US trials as well.

There is a catch that no marketing slogan escapes: assembly is not fabrication. The advanced chips at the heart of every Antminer and Whatsminer are still etched overseas, and US assembly reduces but does not eliminate tariff exposure, because metal inputs remain dutiable. The onshoring race has coincided with the United States becoming the center of gravity for hashrate; VanEck’s Matthew Sigel pegged the US share of global hashrate near 31.6 percent in mid-2025, per CryptoSlate. For a running calendar of the policy deadlines that keep resetting this map, our regulatory countdown tracks what is next.

The TSMC chokepoint and the Sophgo blacklist

Here is the fact that unifies the entire market: Bitmain and MicroBT both fabricate their leading chips at the same foundry, Taiwan’s TSMC. That shared dependence is a single point of failure for the whole industry. If access to TSMC’s advanced nodes were disrupted, by geopolitics, by capacity allocation to higher-margin AI customers, or by export controls, neither the Antminer nor the Whatsminer roadmap would be immune. The duopoly at the brand level sits on top of a monopoly-adjacent chokepoint at the fabrication level.

That risk stopped being theoretical for Bitmain. Sophgo, a chip-design firm tied to Bitmain co-founder Micree Zhan, was added to the US Commerce Department’s Entity List in January 2025 after investigators found a TSMC-made chip matching a Sophgo design inside a Huawei AI processor, part of a batch of more than two dozen China-based entities blacklisted, as the South China Morning Post reported. An Entity List placement bars a firm from receiving US-linked technology exports without a license that is generally denied. Sophgo denied any Huawei relationship, and the action does not ban Antminers. But it is a flashing warning about how exposed Bitmain’s wider chip ambitions are to US export controls, and how tightly the company’s fortunes are bound to a foundry an ocean away.

Beyond Bitcoin: the AI-chip temptation

The Sophgo episode also exposes a strategic fork between the two rivals. Bitmain has long harbored ambitions beyond Bitcoin, dating back to its earlier attempts at AI accelerators, and Sophgo represents an affiliated bet that the same design and foundry relationships that produce mining chips could produce chips for artificial intelligence. That is a lucrative adjacency, given the scramble for AI compute, but it is also precisely the arena where US export controls bite hardest. MicroBT, by contrast, has stayed conspicuously focused on Bitcoin mining hardware, a narrower but less politically radioactive lane.

The irony is that the miners themselves are the ones chasing AI most aggressively. Public operators are converting mining sites into AI and high-performance-computing data centers because the revenue per megawatt is higher, a pivot that has become the dominant story in the sector. The chipmakers face a version of the same temptation from the supply side, weighing whether to serve a cyclical Bitcoin market or a booming AI one with the same scarce foundry capacity. Where mining chips end and AI accelerators begin is getting blurry, and the firm that masters both could reshape the economics of the whole sector.

Bitdeer, SEALMINER, and the case for a third force

A two-horse race can become a three-horse race, and the most credible challenger comes from inside the family. Bitdeer, the company Jihan Wu carried out of the 2021 Bitmain split, has moved from operating and hosting rigs to designing its own. Its SEAL mining chips and SEALMINER machines are an explicit attempt to break the Bitmain and MicroBT lock on the market, and Bitdeer has told CoinDesk it wants to shake up ASIC manufacturing rather than merely buy from it. In a neat historical twist, the boardroom war that once threatened Bitmain has produced a new rival led by its own estranged co-founder.

Whether Bitdeer, Canaan, or a newer entrant can meaningfully dent the duopoly is an open question. Designing a competitive chip is hard; manufacturing it at scale, at yields that beat Bitmain’s, is harder; and doing so while depending on the same TSMC capacity everyone else wants is harder still. For now the practical choice facing a miner remains Antminer or Whatsminer, but the emergence of a third serious designer is the healthiest thing that could happen to a market this concentrated, and it is worth watching closely.

What the duopoly means for Bitcoin’s security

Step back from the spec sheets and the concentration raises a question about Bitcoin itself. Proof of work is supposed to be permissionless and decentralized, yet the machines that produce it come from two firms, both reliant on one foundry, increasingly steered by one country’s trade policy. That does not threaten Bitcoin’s consensus rules, which miners cannot rewrite, but it does concentrate the supply chain that underpins the network’s hashrate, and supply chains are where real-world leverage lives. A shock to Bitmain, to MicroBT, or to TSMC would ripple through the security budget of a trillion-dollar network.

The concentration is geographic as well as corporate. Chip fabrication clusters in Taiwan, chip design sits in a few Chinese firms, and, increasingly, final assembly and the hashrate itself sit in the United States. Each link in that chain is a potential pressure point, whether a tariff, an export control, or a natural disaster. Bitcoin’s defenders argue that this is exactly why the network’s permissionless design matters: no single company controls the protocol, so a disruption at one supplier reshuffles market share rather than halting the chain. Skeptics counter that a hardware supply chain this narrow is a soft target that the protocol’s decentralization does not fully cover.

The counterweight is Bitcoin’s self-correcting economics. When prices fall below the cost of production, higher-cost machines power down, hashrate declines, and difficulty adjusts lower to keep blocks coming, which is exactly what the rare year-over-year difficulty drop of 2026 reflects. Efficient hardware from either vendor is what lets the network keep humming through those squeezes, and the next block-subsidy halving in 2028 will tighten the screw again, a horizon we examine through the lens of Bitcoin’s stock-to-flow model. The Bitmain versus MicroBT contest, then, is not a sideshow to Bitcoin’s story. It is one of the load-bearing walls, and in 2026 it is being rebuilt in real time.

Frequently Asked Questions

Is Bitmain or MicroBT better in 2026?

There is no single winner. Bitmain’s Antminer S23 line leads on raw efficiency, with the S23 Hydro rated near 9.5 J/TH, while MicroBT’s Whatsminer M70 series (around 12.5 to 14.5 J/TH) trades a little efficiency for a reputation for durability and, increasingly, US-assembled machines. Operators with cheap power and dense liquid cooling often favor Bitmain; those prioritizing uptime, service, and resale predictability often favor MicroBT.

What is the difference between Antminer and Whatsminer?

Antminer is Bitmain’s brand and Whatsminer is MicroBT’s. Both are SHA-256 ASIC miners for Bitcoin, both fabricate their chips at TSMC, and both now sell air, hydro, and immersion models. The practical differences are efficiency (Bitmain is usually a step ahead at the top of the range), firmware and tuning support, warranty and repair networks, resale liquidity, and price.

How much market share does Bitmain have?

Industry trackers put Bitmain at roughly 80 percent of the Bitcoin ASIC market, with MicroBT the clear number two near 15 percent and Canaan third; together the three account for something close to 97 percent of new rig supply. Exact figures vary by quarter and by who is counting, but the top-heavy shape is stable.

Why are Bitmain and MicroBT building factories in the United States?

US tariffs on Chinese-made and Southeast-Asia-assembled hardware have climbed sharply since early 2025, so both firms have opened or expanded US assembly to cut import duties and shorten delivery times. Bitmain started a US production line in December 2024, and MicroBT markets US-made, tariff-free rigs. The chips themselves are still fabricated overseas at TSMC, so assembly reduces but does not erase tariff exposure.

What is the Sophgo Entity List issue and how does it affect Bitmain?

Sophgo, a chip-design affiliate tied to Bitmain co-founder Micree Zhan, was added to the US Commerce Department’s Entity List in January 2025 after a TSMC-made chip matching a Sophgo design was found inside a Huawei AI processor. The action does not ban Antminers, but it underscores how exposed Bitmain’s broader chip ambitions are to US export controls and to the industry’s shared dependence on TSMC.

By Yuki Tanaka, senior mining and markets correspondent at HOGE Wire.