Bitmain vs MicroBT: The Sub-10 J/TH Mining War in 2026

In late September 2026, Bitcoin trades around $83,600, up roughly 9 percent on the week, and the machines that secure it are back in demand. According to CoinGecko, the coin has recovered most of the ground it lost during a brutal summer for miners. Hashprice, the daily revenue a rig earns per petahash of computing power, has clawed its way back toward $40, and network difficulty has ground higher again. For the people who actually run the machines, though, the number that decides who makes money is no longer just the coin price. It is joules per terahash: the amount of electricity a rig burns to produce a unit of work. In 2026 that number crossed a line the industry had chased for a decade.

The Antminer S23, now shipping in volume, does its work at 9.5 joules per terahash, the first Bitcoin miner to slip under 10. The company that built it, Bitmain, has sold roughly four out of every five new machines on the market for years. Its only serious rival, MicroBT, sells most of the rest. That two-company grip on the hardware underpinning a trillion-dollar network is what people mean by the ASIC duopoly, and 2026 is testing it harder than any year since the two firms were founded: on the efficiency frontier, against a resurgent third contender born from Bitmain’s own boardroom feud, and under a tariff-and-export-control regime that has turned the question of where a machine gets bolted together into a competitive weapon.

A Duopoly Meets Its Stress Test

Bitcoin mining hardware is one of the most concentrated markets in technology. Industry trackers put Bitmain at roughly four-fifths of global ASIC production, around 82 percent by one 2026 estimate, with MicroBT near 15 percent and Canaan around 2 percent. That leaves three Chinese firms with close to the entire world market between them, and the top two accounting for almost all of it.

The reason the field is so narrow is the cost of entry. Designing a competitive SHA-256 ASIC means taping out a custom chip on a leading-edge process node, today TSMC’s 5 nanometer and 3 nanometer lines, then wrapping it in power supplies, control boards, cooling systems, firmware, and a global logistics and repair operation. Each generation costs tens of millions of dollars to develop and can be obsolete inside eighteen months. Only a handful of companies can absorb that risk, and only two have done it at scale, again and again, for a decade.

For most of that decade the duopoly looked settled. In 2026 three separate pressures are pushing on it at once, and each runs through the same variable: efficiency. As one recent HOGE Wire analysis argued, Bitcoin’s hashrate now climbs only as fast as the two makers can ship efficient silicon, so the race between them has become the effective ceiling on the whole network’s growth (see Bitcoin’s Hashrate Climbs Again, but Silicon Sets the Ceiling).

How Two Rivals Became One Industry

To understand why the two companies are so evenly matched on engineering, start with the fact that they are, in a sense, the same company split in two. Bitmain was founded in Beijing in 2013 by Jihan Wu, a finance-minded entrepreneur who had translated the Bitcoin white paper into Chinese, and Micree Zhan, a chip designer. Their Antminer line and their Antpool mining pool made Bitmain the dominant force in the industry within a few years.

MicroBT exists because of a falling-out inside Bitmain. Yang Zuoxing had been a chief chip designer there, working on the Antminer S7 and S9 that defined an era. He left in 2016, reportedly after a dispute over equity, and founded MicroBT in Shenzhen. The Whatsminer line he built went on to become the only credible alternative to the Antminer, and Bitmain treated it as exactly the threat it was. Bitmain first sued MicroBT and Yang in 2017 for patent infringement, but the Beijing IP Court dismissed that case in 2018 and revoked the patent in question. Yang was then detained in Shenzhen in late 2019 amid embezzlement allegations tied to his old Bitmain-linked employer. In April 2021, Bitmain went back to the Beijing IP Court with a fresh claim accusing Yang and MicroBT of misappropriating trade secrets from the S7 and S9 era, seeking as much as 99 million yuan, about $15 million. (A widely circulated summary that dates this trade-secret suit to 2025 is simply wrong; the filing was April 2021.)

Bitmain then nearly tore itself apart. A power struggle between Wu and Zhan ran from 2019 into 2021 and was settled only when Zhan bought out Wu-aligned shareholders for a reported figure in the hundreds of millions of dollars. Zhan kept Bitmain, meaning the Antminer brand and Antpool, while Wu walked away with the cloud-mining and pool assets that became Bitdeer and BTC.com. That divorce matters for 2026, because the estranged co-founder did not retire. He built a company that now makes its own chips and its own machines, and it is the clearest new threat the duopoly has faced in years.

The Sub-10 Breakthrough: Inside the Antminer S23

Bitmain unveiled the S23 generation at its World Digital Mining Summit in May 2025 and began shipping in volume through the first half of 2026. By late summer it had gone from a launch slide to a machine you can price, order, and take delivery on through hosts and distributors worldwide. The headline is the efficiency figure. The flagship S23 Hydro produces roughly 580 terahashes per second while drawing about 5,510 watts, which works out to 9.5 joules per terahash. That is the first shipping Bitcoin miner rated under 10 J/TH, a threshold the industry had been circling for years.

The range does not stop there. The air-cooled S23 lands near 318 terahashes at around 11 J/TH for sites that cannot run water, and a dense 3U hydro chassis, sometimes listed as the S23 Hyd 3U, roughly doubles the flagship into petahash territory at about 1,160 terahashes and 11 kilowatts in a single box, holding the same 9.5 J/TH. Against the prior S21 generation, which topped out around 13.5 J/TH, Bitmain claims the S23 delivers roughly 69 percent more hashrate and about 41 percent better efficiency, gains that come from a new, denser chip generation rather than incremental tuning.

The launch also reset the commercial mood. Colin Harper, editor in chief at Blockspace Media, wrote when the S23 was revealed that the flood of efficient new machines was turning what had long been a seller’s market into a buyer’s market, handing miners the upper hand in negotiating prices for the first time in a while, a point he made in his Forbes coverage of the reveal. For an industry that spent 2025 short on top-tier hardware, more supply of far more efficient rigs is a structural shift, not a footnote.

MicroBT’s Answer: The Whatsminer M70 Series

MicroBT did not concede the frontier. It used the Bitcoin MENA conference in Abu Dhabi in December 2025 to launch the Whatsminer M70 series, built on a 5 nanometer process and targeting 12.5 J/TH on its best units. The lineup spans the same air-to-hydro spread Bitmain offers. The air-cooled M70S runs 226 to 258 terahashes at 13.5 J/TH, drawing between about 3,140 and 3,500 watts. The hydro M73S climbs to roughly 500 to 552 terahashes, and the higher-binned M73S and later hydro units reach the 12.5 J/TH target. At the top of the range sits the M79S, a hydro monster rated at 930 to 1,040 terahashes and about 14,000 watts of normal draw, with an overclocking envelope up to 20,000 watts, all at a nameplate 13.5 J/TH.

The M70 series is a large jump for MicroBT. The prior-generation M60S sat near 18.5 J/TH, so the new line closes most of the efficiency gap that had opened up during the M60 era, when Bitmain’s S21 family pulled ahead. It does not fully close it: the M79S at 13.5 J/TH is a direct competitor to Bitmain’s petahash-class 3U box, but that Antminer runs at 9.5 J/TH, and coverage of the M79S in Bitcoin News made the same point, noting the Bitmain unit delivers more hashrate at better efficiency in a comparable chassis. Where MicroBT competes hardest is on the parts of the pitch that do not show up as a single J/TH number: sticker price per terahash, availability, and a reputation for uptime and fleet-level reliability that large operators weigh heavily when they are buying tens of thousands of units. MicroBT also leaned into the same theme at launch that Bitmain did, positioning the M70 line for the institutional-scale, hydro-cooled sites that now dominate new buildout.

Head to Head, on Paper

Stacked side by side, the current flagships tell a consistent story: Bitmain owns the top of the efficiency chart, MicroBT sits a couple of joules behind on its best hydro units and further back on air, and Bitdeer’s SEALMINER A3 has quietly matched MicroBT’s best figure. All numbers below are nameplate ratings and vary with firmware, cooling, and ambient conditions.

Why a Few Joules Decide Everything Now

In a bull market with fat margins, the difference between 9.5 and 13.5 J/TH is a rounding error. In 2026 it is the difference between running and switching off. The reason is hashprice, the dollar revenue a miner earns per unit of hashrate per day. It collapsed to multi-year lows over the summer, and even after the autumn recovery it sits only around $40 per petahash per day, up about 22 percent from the low $30s a month earlier, according to Bitcoin.com News. When revenue per unit of work is that thin, the electricity a machine burns is almost the entire cost equation.

The math is simple enough to do on a napkin. A rough cash breakeven power price equals hashprice divided by 24 times the machine’s efficiency in J/TH. At a $40 hashprice, that gives the electricity ceiling below which a machine still earns more than it spends on power. The table shows how brutally efficiency compresses that ceiling. An S23 at 9.5 J/TH stays cash-positive at power prices up to roughly 17.5 cents per kilowatt-hour. A three-year-old S19-class rig at 21.5 to 29.5 J/TH needs power under about 8 cents, or under 6 cents, to clear the same bar. Whole fleets of that older hardware went dark at the summer lows; CoinShares estimated that 15 to 20 percent of the global fleet was running at a cash loss, against a weighted-average cash cost near $80,000 per Bitcoin.

This is why the fight over a few joules is not a spec-sheet vanity contest. Fred Thiel, chief executive of the large public miner MARA, has described the business bluntly, telling CoinGeek that mining is a zero-sum game: as more people add capacity, it gets harder for everybody else, margins compress, and the floor is your energy cost. The most efficient machine is the one that pushes that floor lowest, which is why the S23-versus-M70 duel maps almost directly onto the margin math that governs the whole industry. The broader August rally that lifted these numbers, which followed the Federal Reserve’s rate move, is chronicled in Buy the News: Crypto Rallied After the Fed Hiked to 4%.

The Third Contender: Bitdeer and the Ghost of a Feud

The most interesting development of 2026 is not inside the duopoly but just outside it. Bitdeer, the Nasdaq-listed miner chaired by Bitmain co-founder Jihan Wu, spent years as a customer of Bitmain and MicroBT. Then it started designing its own chips. Its self-developed SEAL silicon now powers the SEALMINER line, and the roadmap has moved fast. The second-generation SEALMINER A2, built on the SEAL02 chip, shipped around 16.5 J/TH. The SEALMINER A3 series pushed that to as low as 12.5 J/TH, with an A3 Pro Hydro rated at 660 terahashes, matching MicroBT’s best figure.

The eye-catching item is the next step. Bitdeer has teased a SEALMINER A4 Ultra Hydro rated at 886 terahashes and 9.45 J/TH, a whisker under the S23’s 9.5. If that machine reaches mass production at the stated figure, the duopoly’s efficiency crown would, for the first time, be genuinely contested by an outsider. And the outsider is no random startup: it is the vehicle of the man who co-founded Bitmain and lost the fight for it. Bitdeer is also scaling as a miner in its own right, with monthly Bitcoin production that jumped 251 percent as its self-mining hashrate climbed toward 45.7 exahashes, giving it both a captive buyer for its rigs and a proving ground for them.

Canaan, the third of the Chinese incumbents, remains a distant player at around 2 percent of the market with its Avalon line, though as a Nasdaq-listed company it is subject to US securities disclosure and has run early US assembly trials of its own. The point is that the field is no longer obviously a two-horse race. It is Bitmain and MicroBT, with a well-funded, vertically integrated Bitdeer pushing into the top tier and Canaan holding the low end.

The 2026 Network Backdrop: Recovery and the Silicon Ceiling

The hardware race is playing out against a network that spent 2026 in an unusual state. Difficulty, the self-adjusting measure of how hard it is to find a block, has climbed back to about 132.76 trillion, per CoinWarz, still roughly 15 percent below the record near 156 trillion set in late 2025 but rising again after a long stretch of cuts. Network hashrate hovers near 900 exahashes per second, with individual daily readings occasionally spiking toward a full zettahash. That recovery is itself a story about machines switching back on as the price rose, covered in Bitcoin Mining Difficulty Climbs Again as Miners Switch Back On.

Here is where the duopoly and the network meet. Difficulty is a thermostat: when efficient new rigs come online, they add hashrate, difficulty rises, and each machine earns a smaller slice of the same fixed block reward. So the S23 and M70 do not just help their individual owners; collectively they raise the bar for everyone, competing away the very margin the autumn rally handed back. And because those efficient machines can only be produced as fast as Bitmain and MicroBT can secure wafer capacity at TSMC, the growth of Bitcoin’s security budget is now effectively rate-limited by two companies’ chip orders. Capital is not the binding constraint anymore; silicon is.

Tariffs and the Geopolitics of a Joule

An efficient machine is only cheap if you can get it into your data center without paying a penalty at the border, and in 2026 that penalty grew sharply for American buyers. The United States hosts something like 40 percent of global hashrate, so US import policy shapes the whole market. Under the reciprocal tariff regime that took effect on August 7, 2025, mining rigs assembled in Indonesia, Malaysia, or Thailand, the Southeast Asian countries where Bitmain and MicroBT shifted production years ago to dodge China duties, now carry about 21.6 percent in total US import duties, according to The Block. Machines that count as China-origin can face roughly 57.6 percent once the various stacked tariffs are added up. Before all this, the standard duty on such imports was around 2.6 percent.

That is a large enough number to reorder purchasing decisions. Ethan Vera, chief operating officer of mining services firm Luxor Technology, told The Block that at 21.6 percent tariffs, the US is now one of the least competitive jurisdictions to bring machines into, and warned that even domestic assembly will not fully fix it because the raw materials and components still come from Asia, so the machines end up carrying a higher cost regardless. Not everyone reads it as fatal. Leo Lu, chairman and chief executive of BitFuFu, offered a more optimistic view in the same reporting, arguing that US miners can stay competitive through low-cost, often renewable power and local partnerships in states like Texas, Oklahoma, and Colorado. Both can be true at once: the tariff raises the landed cost of a rig, and cheap power can offset it, which only sharpens the premium on efficiency because an efficient machine wrings more revenue out of each of those cheap kilowatt-hours.

Made in America? The Reshoring Theater

The obvious response to a border tax is to move production inside the border, and both makers have. Bitmain launched its first US production line in December 2024, delivering an initial batch of Antminer S21 Pro units, and signaled plans to open a full American factory, reportedly eyeing Texas or Florida, hiring local staff to speed deliveries and repairs, as documented by TheMinerMag. MicroBT has pursued the same path through partners; The Block noted that Luxor has been helping clients secure machines through onshore manufacturing arrangements, including a domestic production deal with MicroBT.

The catch is the difference between assembly and fabrication. What happens on a US line is the final putting-together of boards, power supplies, and enclosures. The part that actually defines a mining chip, the advanced semiconductor fabrication, still happens at TSMC in Taiwan, with some capacity at Samsung in South Korea. The wafers, the boards, and many components still originate in Asia and remain dutiable. So a machine that is technically assembled in America is a partial hedge against tariffs, not an escape from them, and it does nothing to change the single most important dependency in the entire industry: every serious ASIC, from every maker, is fabricated by the same handful of foundries. If TSMC has a bad quarter or a geopolitical shock, Bitmain, MicroBT, and Bitdeer all feel it at once. The reshoring is real, but it is also, in part, theater staged for a tariff schedule.

The Export-Control Shadow: Sophgo and the Entity List

Tariffs are a cost. Export controls are a wall, and Bitmain has already brushed against one. In January 2025, the US Commerce Department added Sophgo, a chip-design company affiliated with Bitmain, to its Entity List. The trigger was a TSMC-made chip that matched a Sophgo design and turned up inside Huawei’s Ascend 910 artificial intelligence processor, a discovery that prompted TSMC to halt shipments and alert authorities. Sophgo denied any business relationship with Huawei. What makes it relevant to the mining duopoly is the ownership: Micree Zhan, Bitmain’s controlling co-founder, indirectly holds a roughly 23 percent stake in Sophgo through an investment vehicle.

The listing does not ban Antminers, and Bitcoin ASICs are not the target of US chip sanctions. But it is a warning shot about where Bitmain wants to go next. Sophgo represents Bitmain’s ambition to sell AI and general-purpose silicon, not just mining chips, and that ambition runs straight into the most sensitive corner of US-China technology policy. A company whose founder is caught in an export-control action is a company whose access to leading-edge fabrication is politically fragile, and leading-edge fabrication is exactly what its mining business also depends on. It is one more reason the geopolitics of chips now sits at the center of a business that used to be about nothing but hashes, and it lands amid a broader reset of US crypto and technology policy detailed in Crypto’s Countdown Resets: New Deadlines After CLARITY Failed.

What Buyers Actually Weigh

For an operator deciding what to deploy in 2026, the J/TH figure is the headline but not the whole invoice. The real decision balances several factors, and the right answer changes with the size of the buyer, the cost of their power, and where their site sits relative to a border.

• Efficiency (J/TH): the single biggest driver of breakeven power price and useful life. The more expensive your electricity, the more an S23’s edge is worth.
• Upfront price per terahash: MicroBT and Bitdeer often undercut Bitmain on sticker price, and the S23’s efficiency commands a premium that only pays back at higher power costs or over longer holds.
• Landed cost: tariffs and shipping can add double-digit percentages, so where a unit is assembled now materially changes the total.
• Availability and lead times: flagship rigs sell in batches and pre-orders; a machine you cannot receive for two quarters is a different investment than one on the shelf.
• Reliability and serviceability: uptime, warranty terms, and access to local repair matter enormously at fleet scale, and are central to MicroBT’s pitch.
• Cooling model: air, hydro, and immersion each imply different site buildout; the most efficient figures generally require water.
• Resale value and firmware: a machine’s second-hand price and its openness to tuning affect the return long after purchase.

In practice, the largest institutional buyers with sub-5-cent power and long horizons tend to chase the absolute efficiency leader, while operators with cheaper marginal decisions or tighter capital often find MicroBT’s or Bitdeer’s price-per-terahash more compelling. The duopoly holds not because one maker is best on every axis, but because between them they cover almost every combination a buyer might want.

After the Block Reward: AI, Fabs, and the 2028 Clock

Two forces will shape the next round of this rivalry. The first is artificial intelligence, and it cuts both ways. Bitmain, through Sophgo, wants to sell AI silicon; Bitdeer runs an AI cloud business alongside its mining. At the same time, their own customers, the public miners, are pivoting hard into AI and high-performance computing, which means the same megawatts and the same TSMC wafers are being fought over by two industries at once. That competition for scarce compute capacity is increasingly a boardroom-level decision at every large miner, a dynamic explored in Bitcoin Hashrate Growth Is a Boardroom Call in 2026. If AI buyers outbid miners for leading-edge wafers, the pace of ASIC efficiency gains could slow, or the makers could simply prioritize higher-margin AI chips.

The second force is the halving clock. In 2028 the block subsidy falls again, from 3.125 to 1.5625 Bitcoin, cutting the reward side of every miner’s revenue in half unless price or fees rise to compensate. Thiel has argued the math gets very tough after 2028 without sustained annual price growth, and that miners will increasingly need to own or partner with power generation rather than simply plug into a grid. In that world, the efficiency of the machine is not a nice-to-have; it is the precondition for surviving the next subsidy cut. Every joule the S23, the M70, and the SEALMINER shave off today buys their owners a little more runway toward a future where the block reward alone no longer pays the bill.

The Bottom Line

Bitmain still leads. It holds the efficiency crown with the first sub-10 J/TH machine, and it still sells roughly four of every five new rigs. But 2026 is the first year in a long while when that lead looks genuinely contestable. MicroBT has closed most of the gap with the M70 series and competes hard on price and reliability. Bitdeer, run by Bitmain’s estranged co-founder, has matched MicroBT’s best efficiency and teased a machine that would edge past the S23. Tariffs and export controls have redrawn the map of what a machine actually costs to land and run, and the whole industry now leans on the same foundries and, increasingly, competes with AI for their output. For a miner, the lesson of the year is simple and unforgiving: in a market where hashprice is thin and difficulty is climbing, the spec that matters is joules per terahash, and the cheapest joule wins.

Frequently Asked Questions

Is Bitmain or MicroBT better in 2026?

Bitmain still leads on raw efficiency. Its Antminer S23 Hydro runs at 9.5 J/TH, the first Bitcoin miner under 10, while MicroBT’s best Whatsminer M70 hydro units land around 12.5 J/TH. MicroBT competes on price, availability, and a reputation for uptime, so the better choice depends on your power cost: the more you pay per kilowatt-hour, the more Bitmain’s efficiency edge is worth.

What is the most efficient Bitcoin miner right now?

As of late 2026 the Antminer S23 Hydro and its denser 3U version are the most efficient shipping machines at 9.5 J/TH. Bitdeer has teased a SEALMINER A4 rated at 9.45 J/TH, which would edge past it, but that model is not yet in mass production.

What does J/TH mean and why does it matter?

J/TH stands for joules per terahash, the electricity a miner burns to produce one terahash of computing work. A lower number means less power for the same output. Because electricity is a miner’s largest ongoing cost, a lower J/TH rating directly lowers the Bitcoin price at which a machine stops being profitable, which is why efficiency decides which rigs survive a downturn.

How do US tariffs affect the price of a mining rig?

Since August 2025, mining machines assembled in Indonesia, Malaysia, or Thailand face about 21.6 percent in US import duties, and China-origin rigs can face roughly 57.6 percent, up from about 2.6 percent before. Those levies land on top of the sticker price, which is why Bitmain and MicroBT have both moved some assembly to the United States, even though the chips themselves are still made in Asia.

Who founded MicroBT, and how is it related to Bitmain?

MicroBT was founded in 2016 by Yang Zuoxing, a former chief chip designer at Bitmain who left after a dispute over equity. Bitmain later sued him and MicroBT twice, including an April 2021 trade-secret claim seeking up to 99 million yuan (about $15 million). The two firms have been direct rivals ever since.

By Marcus Okafor, senior mining correspondent at HOGE Wire.