Can the world’s fastest supercomputer hack Bitcoin? The claim that one Bitcoin miner is “1,000 times behind” a modern supercomputer arises from comparing one machine rated in petahashes with another rated in exaFLOPS—but the units cannot be divided into a meaningful speed ratio. Bitcoin miners perform SHA-256 hashes; scientific supercomputers are ranked by floating-point operations. A machine can dominate one benchmark and be poorly suited to the other. A 51% attack is a real theoretical risk, but a powerful scientific computer does not automatically possess the specialized hashing hardware needed to mount one.

For broader security context, a Cambridge Centre for Alternative Finance study estimated the combined top 500 supercomputers at only 4.7 petahashes per second of SHA-256 work under a deliberately simplified, CPU-based conversion. Its authors cautioned that GPUs, CPU architecture and the fundamentally different benchmarks make the approximation uncertain. Nevertheless, it found those machines collectively represented about 0.0006% of Bitcoin’s estimated hash rate at the end of 2024. That historical network-level illustration does not answer the individual-machine benchmark question; it is relevant only to the separate 51% security analysis.

Editor’s clarification — October 27, 2026, updated: This article originally compared the entire Bitcoin mining network against one supercomputer; the intended headline comparison is one supercomputer installation versus one individual Bitcoin mining machine. More fundamentally, our question is about raw computational activity over time—the processing cycles, parallel work and computational throughput that a machine can sustain each second—and the electrical power and energy efficiency required to achieve that performance. We understand that the machines execute different programs and compute different operations. The purpose is a fundamental compute-performance and power-efficiency comparison, not a claim that one workload directly substitutes for the other. We use “compute cycles per second” as an intuitive description of that underlying question, not as an established common benchmark: processor clock cycles per second, floating-point operations per second (FLOP/s) and completed SHA-256 hashes per second (H/s) are different quantities. A single clock cycle can support multiple operations in parallel, while a hash comprises many internal operations. Thus 2.198 EFLOP/s and 1.35 PH/s cannot be converted into an apples-to-apples cycle count or a valid 1,628× performance advantage without measurements of the actual hardware work, operating clocks, parallelism and energy consumption under a comparable benchmark. A meaningful comparison could report each system’s architecture, clocked processing activity, measured workload-specific throughput, watts and energy per completed task, with limitations clearly labeled. We retain the wider Bitcoin network analysis solely to assess the economics and feasibility of a majority-hashpower attack, which is a separate question from the performance of one machine.

The intended one-machine comparison: LineShine versus a 1.35 PH/s Bitcoin ASIC

The world’s fastest single TOP500 supercomputer installation, LineShine, recorded 2.198 exaFLOP/s on the June 2026 HPL benchmark. A high-output individual Bitcoin mining unit, the MicroBT WhatsMiner M79S, is specified at 1.35 petahashes per second (1,350 TH/s) of SHA-256 hashing, drawing about 20 kW. This is the intended comparison: one supercomputer system against one 3U hydro-cooled ASIC mining chassis. LineShine is physically a large multi-rack installation with millions of cores, not a single-chip device, so even the word “machine” describes very different physical scales.

Numerically, 2.198 × 10¹⁸ divided by 1.35 × 10¹⁵ is approximately 1,628. That is the source of a provocative “roughly 1,000×” headline, but the quotient is not a valid speed or security comparison: its units are floating-point operations per SHA-256 hash, not a dimensionless performance ratio. The calculation does not show LineShine mines Bitcoin 1,628 times faster. To determine actual mining performance, LineShine would need a measured end-to-end Bitcoin double-SHA-256 benchmark, including its software, CPU instruction throughput, memory and communication overhead. We have not found a verified LineShine SHA-256 benchmark and therefore cannot claim which of these two systems would perform more Bitcoin hashes per second.

The 1.35 PH/s M79S figure is a manufacturer/distributor specification rather than a verified third-party measurement of an installed unit, and it is not a claim about the entire Bitcoin network. By contrast, the network is estimated in exahashes per second because it comprises many independent miners. For an actual 51% attack, the relevant denominator is total active network hashrate, not one M79S machine. That is why the rest of this report discusses network-scale proof-of-work security after first establishing the correct one-machine framing.

1. What exactly is the world’s fastest supercomputer?

The June 2026 TOP500 ranking lists China’s LineShine at the National Supercomputing Centre in Shenzhen as No. 1, with a measured 2.198 exaFLOP/s on the High Performance LINPACK (HPL) benchmark. The former leader, El Capitan at Lawrence Livermore National Laboratory, ranks second at 1.809 exaFLOP/s. El Capitan is a relevant U.S. example, but it should no longer be described as the fastest in the June 2026 list.

Exa means 10¹⁸. A floating-point operation is a numerical calculation, often an addition or multiplication on approximate real-number values. HPL measures sustained performance solving large dense linear-algebra problems. These are valuable for simulations, weather and physics, but they are not the bitwise SHA-256 compression operations Bitcoin miners repeat.

2. How much computing does Bitcoin actually perform?

Bitcoin’s aggregate network hashrate was estimated at approximately 1,022 EH/s on a 30-day average as of October 26, 2026, with a noisier daily estimate around 1,132 EH/s, according to Blockchain.com-derived historical network statistics. An exahash per second means 10¹⁸ hash attempts each second. Hashrate is estimated from observed block production and difficulty, not measured by directly counting every miner’s silicon output.

A single SHA-256 proof-of-work attempt applies Bitcoin’s double-SHA-256 function to an 80-byte block header. The miner repeatedly varies a nonce or other block-template material until the resulting digest is numerically below the network’s difficulty target. Bitcoin’s developer documentation explains this block-header process. Most competitive miners use application-specific integrated circuits (ASICs) designed for the relevant bitwise operations; they do not need to perform massive floating-point matrix calculations.

Metric	Individual WhatsMiner M79S	LineShine supercomputer
Primary workload	Double SHA-256 proof-of-work	General scientific numerical computing
2026 measurement	1.35 PH/s (specified, individual chassis)	2.198 EFLOP/s HPL (June TOP500)
Units	SHA-256 hashes per second	Floating-point operations per second
Direct ratio valid?	No: different operations, architectures and benchmarks
51% threshold	Attacker needs more than half of combined mining hash power for sustained majority control	HPL result says nothing directly about attainable SHA-256 mining share

3. Why 1,000× is not a valid comparison

One cannot calculate 2.198 EFLOP/s ÷ 1,022 EH/s and call the answer a security ratio. Both numbers contain the prefix exa, but a FLOP is not a SHA-256 hash. This is like comparing a cargo train’s ton-kilometers with an aircraft’s passenger-kilometers: each measures real work, but they measure different tasks. A scientifically useful comparison must benchmark actual double-SHA-256 throughput on the supercomputer’s hardware, account for GPUs and CPU instructions, and compare energy, utilization and network difficulty.

The Cambridge study’s technical appendix approximated 4.7 PH/s for all 500 top supercomputers combined, using a reference CPU and a FLOPS-to-hash conversion. At 1,022 EH/s, that historical estimate would equal about 0.00046% of today’s network (4.7 PH/s ÷ 1,022,000 PH/s × 100). That illustration is not a new benchmark: the 2024 hardware population and 2026 network measurement are different dates, and the conversion is deliberately approximate. The study itself notes that accelerators could change the result. Still, nothing in the TOP500’s HPL ranking establishes a viable 51% attack.

4. What is a 51% attack?

Bitcoin’s security depends on the relative cost of producing valid proof-of-work. In a majority-hashpower attack, an adversary with sustained control of more than half of total mining hash power can statistically outpace honest miners and attempt to replace a recent chain segment. That creates the possibility of double-spending the attacker’s own coins, excluding transactions or disrupting confirmations. Less than 50% can sometimes succeed in limited reorganizations, but success becomes much less reliable as honest confirmations accumulate. The phrase “51% rule” is a shorthand for a probabilistic security threshold, not an absolute impossibility theorem.

Crucially, even a majority attacker cannot manufacture a valid signature for another user’s wallet, spend someone else’s UTXOs without authorization, create coins outside the issuance rules or force independently validating full nodes to accept invalid blocks. Nodes verify consensus rules regardless of the number of hashes an attacker generates. These limits are documented in Bitcoin’s developer guide and Mastering Bitcoin’s mining-and-consensus chapter. A 51% attack is dangerous to transaction finality and censorship resistance; it is not a universal wallet-decryption or protocol-rewriting capability.

5. How much hash power would an attacker need?

If honest miners currently produce H hashes per second, an attacker joining with newly added capacity A needs A > H to control more than 50% of the combined total, because A/(H+A) must exceed 0.5. With H ≈ 1,022 EH/s, the attacker would need more than 1,022 EH/s of additional sustained hashing, assuming honest capacity stays constant. If instead the attacker seizes or coordinates already-existing miners, the requirement is control of more than half of the existing combined network, roughly above 511 EH/s at that baseline. These are different attack models and should never be conflated.

At an illustrative mining efficiency of 8.9 joules per terahash, an additional 1,022 EH/s of purpose-built mining equipment would draw approximately 9.10 gigawatts of chip-level power (1,022 million TH/s × 8.9 J/TH). This excludes fans, pumps, transformers and other facility overhead; it also assumes enough ASICs, electrical capacity and mining infrastructure can be obtained. The efficiency is an illustrative leading-generation parameter, not the network-wide average. Hardware acquisition, manufacturing, logistics and power delivery make the attack more than a pure mathematics problem. For background on ASIC trends, see our research into the path toward 1 J/TH.

6. Why Bitcoin remains difficult to attack—and what could still go wrong

The pro-Bitcoin case is strong: specialized ASICs already deploy immense SHA-256 throughput worldwide, and full nodes independently enforce valid transaction signatures, block subsidies and other consensus rules. Scientific exascale hardware is not an economical shortcut around either protection. Bitcoin also spreads mining across jurisdictions and operators, making coordinated acquisition of majority hashing expensive and operationally conspicuous. Those features support resilience, but they do not establish that attacks are impossible.

The most credible concerns are mining-pool coordination and concentration, compromised mining infrastructure, energy-supply shocks, chain reorganizations, fee-market weakness and economic incentives rather than a headline FLOPS number. A pool’s reported share does not automatically mean it owns the ASICs of all participants; miners may switch pools, though such switching is not instantaneous or guaranteed. A separate October 2026 Cambridge security-budget analysis warns that declining block subsidies could weaken future incentives unless transaction-fee revenue or other economic factors compensate. Its researchers found that transaction fees averaged only 0.67% of miner revenue in the 90 days through October 4. That deserves monitoring, not dismissal.

7. What about quantum computers?

Quantum computing is a different threat model. A sufficiently capable fault-tolerant quantum computer might eventually threaten exposed public-key signatures, while Grover’s algorithm offers only a quadratic theoretical speedup for brute-force hash search and must contend with large quantum error-correction overhead. Today’s HPL supercomputers are not quantum computers, and FLOPS do not quantify their ability to break signatures. Bitcoin could adopt quantum-resistant signature schemes through carefully coordinated protocol changes, but the engineering and governance challenges are significant. This report does not claim a guaranteed date for such a threat.

8. Technical glossary and reproducible checks

FLOP/s: floating-point operations per second; commonly used for scientific computing. H/s: cryptographic hashes per second; Bitcoin mining’s relevant throughput. ASIC: specialized chip built for a narrow task. Difficulty: Bitcoin’s adjustable target controlling how hard it is to find a valid block. Confirmation: a block added on top of a transaction’s block, increasing the work needed to reverse it. Reorganization: replacement of a previously accepted branch by another valid branch with more accumulated work. Full node: software independently checking consensus validity. IT load: power consumed by computing equipment, distinct from facility overhead.

Worked calculation: 1,022 EH/s = 1,022,000 PH/s. A 4.7 PH/s hypothetical cluster represents 4.7 / 1,022,000 = 0.00000460 of the total, or 0.000460%. For a new adversary to become the majority while honest miners stay at 1,022 EH/s, it must add more than 1,022 EH/s. The attacker could not accomplish this merely by pointing an HPL benchmark at Bitcoin: its true SHA-256 performance must be measured.

Conclusion: Bitcoin is not 1,000× behind the supercomputer

The apparent “1,000×” single-machine ratio fails on units and workload. The latest TOP500 leader excels at scientific floating-point computation; Bitcoin’s network is a distributed, specialized SHA-256 engine. A majority attack remains theoretically possible, but the evidence does not support the suggestion that a supercomputer could effortlessly hack Bitcoin. More importantly, even majority hashpower does not grant control over private keys or the rules enforced by independent nodes. The most productive security debate focuses on measurable mining concentration, attack economics, fee incentives, and continued improvement of network decentralization—not a misleading exaFLOPS-versus-exahashes headline.

For related reading, see BitcoinVersus.Tech’s FLOPS vs Hashrate lesson, 2026 supercomputer ranking and coverage of mining-pool participation.

Editor’s Note: This textbook-style report favors a technically grounded understanding of Bitcoin security, without claiming invulnerability. Benchmark rankings, Cambridge approximations and contemporary network estimates are distinguished. Two authentic, topic-specific YouTube videos are embedded. Support independent technical reporting with a 5,000-satoshi donation by scanning the QR code below.

Scan the QR code using a compatible Bitcoin wallet while viewing the article on another device.