Quick answer: Min-entropy measures the worst case: how likely an attacker is to guess your secret on the first try. Classical min-entropy assumes the attacker holds ordinary information about your random source. Quantum min-entropy (formally, conditional min-entropy H_min(X|E)) assumes the attacker may hold quantum information, such as a system entangled with your source. Those two numbers can differ. Some classical randomness extractors that are secure against classical attackers fail against quantum ones. With NIST's final transition report (October 2026) setting RSA and ECC disallowance for 2035, every post-quantum key you generate is only as strong as the min-entropy behind it.
If you've read our Bell Test explainer, you've seen how entanglement can certify randomness. This Architect piece goes one layer down, to the metric itself: what min-entropy measures, what changes when the adversary goes quantum, and how to audit your randomness supply chain before 2030.
Executive Summary: Key Takeaways
Min-entropy, not Shannon entropy, is the number that matters for keys. A 128-bit key that takes one particular value half the time has about 65 bits of Shannon entropy but only 1 bit of min-entropy. An attacker guesses it on the first try 50% of the time.
Quantum min-entropy is well defined. It isn't an open research question. Renato Renner (ETH Zurich) formalized it in 2005, and König, Renner and Schaffner (2009) proved it equals the attacker's best possible guessing probability using any quantum measurement.
Classical guarantees don't automatically survive a quantum adversary. A 2007 result showed an extractor secure against classical memory can be broken by an attacker with a much smaller quantum memory. Use quantum-proof extractors.
Randomness has become a regulated, certifiable product. NIST completed its SP 800-90 series in September 2025. Quantinuum's software QRNG earned SP 800-90B validation in April 2025, and JPMorganChase and Quantinuum certified 71,313 bits of quantum randomness in Nature that March.
The deadlines are close. NSA's CNSA 2.0 applies to new national security system acquisitions from January 1, 2027. The EU wants high-risk systems migrated by end of 2030. NIST disallows RSA and ECC after 2035.
What is min-entropy, in plain English?
Min-entropy is the security engineer's version of "how hard is this to guess?" It ignores the average case and focuses on the single most likely outcome, because that's what a smart attacker tries first.
Take a coin. A fair coin has 1 bit of min-entropy: the best guess wins 50% of the time. Weight the coin to land heads 99% of the time and min-entropy collapses to about 0.015 bits, because the attacker just says "heads." The more lopsided the source, the lower the min-entropy and the weaker any key built from it.
For the engineers: H∞(X) = −log₂ maxₓ P(X = x). Equivalently, the attacker's best one-shot guessing probability is p_guess = 2^(−H∞). Min-entropy is always ≤ Shannon entropy, often by a wide margin.
Why not just use Shannon entropy?
Shannon entropy measures average surprise, which suits compression and communication. Attackers don't care about averages. In the 128-bit example above, the key looks rich on average (roughly 65 bits of Shannon entropy) but a single value dominates. Min-entropy exposes that weakness immediately, which is why NIST's entropy-source standard, SP 800-90B, is built around it.
What min-entropy is not
The original framing of min-entropy often blurs two things. True random number generators (TRNGs) harvest min-entropy from physical noise. Pseudorandom generators (PRNGs and NIST's DRBGs) don't create any: their output can never hold more min-entropy than their seed. A good DRBG gives computational unpredictability, meaning an attacker can't predict it without impractical computing power. That distinction becomes important once the attacker's computer is quantum.
How is min-entropy used in classical cryptography today?
Nearly every security primitive depends on it. These are the places it matters most:
Key generation. AES, RSA, ECC, and the new ML-KEM keys all start as random bits. Low min-entropy shrinks the effective key space.
Nonces and IVs. Repeated or predictable values break signatures and authenticated encryption. Leaked ECDSA nonces have exposed private keys in production systems.
One-time pads. The pad's security is exactly its min-entropy. A pad with less min-entropy than the message length isn't perfectly secret.
Seeding DRBGs. NIST SP 800-90A generators are only as strong as the entropy that seeds and reseeds them.
Randomness extraction. Raw physical noise is biased. Extractors, such as the hash functions covered by the leftover hash lemma, turn a source with k bits of min-entropy into nearly k uniform bits.
The US standards stack for this is now complete. SP 800-90A defines the DRBGs, SP 800-90B specifies how to validate entropy sources and estimate their min-entropy, and SP 800-90C, published September 25, 2025, defines how to combine them into full random bit generators (classes RBG1, RBG2, RBG3 and RBGC). For FIPS 140-3 modules, an SP 800-90B entropy validation is a standard requirement.
What changes when the attacker is quantum?
Two things change, and only the first gets much attention.
Threat 1: Quantum computers attack the math around your randomness
Shor's algorithm breaks the factoring and discrete-log problems behind RSA and elliptic-curve cryptography (ECC). Randomness provides one of the clearest examples. Dual_EC_DRBG, the NIST generator removed from SP 800-90A after the 2013 Snowden disclosures, relies on an elliptic-curve relationship. Anyone who knows the secret link between its two curve points can predict its output. A large enough quantum computer could compute that link. This is the generator at the center of the 2015 Juniper ScreenOS backdoor incident. Any randomness design whose unpredictability depends on discrete-log hardness falls when Shor arrives.
Grover's algorithm gives a quadratic speedup for brute-force search, nominally halving symmetric key strength. NIST still treats AES-128 as meeting its lowest post-quantum category, because Grover parallelizes poorly. NSA's CNSA 2.0 requires AES-256 anyway. Longer keys need more min-entropy to fill them.
Threat 2: The attacker holds quantum information about your source
This shift changes the metric itself. In the classical model, the attacker's knowledge (side information) is classical: logs, timing data, a leaked partial seed. In the quantum model, the attacker may hold a quantum system correlated with your source: a photon from the same emission, or a system entangled with the device that generated your key. They can delay measuring it until they learn more, then choose the best measurement.
Quantum min-entropy measures your secret's unpredictability against that stronger attacker. It's written H_min(X|E), where X is your secret, and E is everything the adversary holds, quantum or classical.
For the engineers: Classical conditional min-entropy is H∞(X|E) = −log₂ Σₑ P(e) maxₓ P(x|e). The quantum version replaces "look up the best guess given e" with "perform the best quantum measurement on E": H_min(X|E) = −log₂ p_guess(X|E), optimized over all measurements (König, Renner and Schaffner, IEEE Trans. Inf. Theory, 2009). For fully quantum states it can even be negative: for a maximally entangled pair of qubits, H_min(A|B) = −1.
That negative value is not a curiosity. It captures something classical models can't express: an entangled attacker can know more than everything about your system in a precise sense, which is exactly the leverage an eavesdropper on a quantum channel is trying to gain.
Quantum vs. classical min-entropy: what are the key differences?
Classical min-entropy | Quantum min-entropy | |
|---|---|---|
Attacker's side information | Classical data | Quantum states, including entanglement |
Attacker's strategy | Best guess given the data | Best quantum measurement, chosen later |
Range | Always ≥ 0 | Can be negative for quantum states |
Extractor guarantee | Any classically secure extractor | Only quantum-proof extractors |
Where it's used | SP 800-90B, FIPS 140-3, TRNG design | QKD security proofs, QRNG certification, device-independent protocols |
The row that should worry engineers is the extractor row. In 2007, Gavinsky, Kempe, Kerenidis, Raz and de Wolf showed that an extractor secure against an attacker with classical memory can be broken by an attacker with a much smaller quantum memory. The good news is that the workhorses still hold up. Two-universal hashing (the leftover hash lemma) is secure against quantum side information (Renner, 2005; Tomamichel et al., 2011), and Trevisan's extractor is quantum-proof (De, Portmann, Vidick and Renner, 2012). The lesson: check that your extractor's proof covers quantum adversaries. A security proof against classical attackers alone doesn't guarantee it.
How do QRNGs turn quantum min-entropy into a product?
Quantum random number generators exploit outcomes that physics says are unpredictable, such as which way a photon goes at a beam splitter, when it arrives, or the phase of vacuum noise. The real question for a buyer is not "is it quantum?" but "how much do I have to trust the device?" The market spans three trust levels:
Trusted-device QRNGs model the hardware and bound its min-entropy. They're fast (Gbps-class) and cheap, and they're shipping in phones. The device must behave as modeled.
Semi-device-independent QRNGs trust part of the setup (say, the detector) and certify the rest, trading some speed for weaker assumptions.
Device-independent QRNGs certify min-entropy from a Bell-inequality violation, even if an adversary built the hardware. They're the strongest and slowest, and still mostly operated by labs and governments.
Certified randomness from quantum computers adds a fourth model: a remote quantum computer produces outputs that a classical supercomputer later verifies could not have been faked.
Real-world case studies: who is putting quantum min-entropy to work?
US: JPMorganChase and Quantinuum certify randomness on a quantum computer
In Nature (March 26, 2025), JPMorganChase, Quantinuum, Argonne and Oak Ridge National Laboratories and UT Austin reported the first certified randomness generated on a quantum computer. They used Quantinuum's 56-qubit System Model H2, accessed remotely, and certified 71,313 bits of entropy. Verification ran on DOE supercomputers with a combined sustained 1.1 exaFLOPS. Separately, on April 2, 2025, Quantinuum's Quantum Origin became the first software QRNG to earn NIST SP 800-90B entropy-source validation. That puts quantum-derived min-entropy inside a FIPS-compatible compliance boundary.
Europe: ID Quantique, Quside and the EU's 2030 deadline
Geneva-based ID Quantique (IDQ), a University of Geneva spin-off founded in 2001 with nearly 300 granted and pending patents, came under IonQ's control when IonQ completed its acquisition of a controlling stake on May 6, 2025. Barcelona's Quside extended its Series A beyond €10 million, backed by Trumpf Venture and Bullnet Capital, to scale QRNGs for data centers. Policy is setting the pace. The EU's Coordinated Implementation Roadmap (June 2025) requires national PQC plans and pilots by the end of 2026 and high-risk migrations by the end of 2030.
Asia: QRNG in consumer phones and sovereign quantum security
In South Korea, SK Telecom and Samsung shipped the Galaxy Quantum 5 with IDQ's 2.5 × 2.5 mm QRNG chip. It feeds authentication and encryption for banking, payments, and biometrics through Samsung Knox. In India, QNu Labs raised $22.8 million in a Series A1 (September 2026) led by the National Quantum Mission and Speciale Invest, with Sony Innovation Fund participating, to scale its QRNG, QKD and PQC products. In China, Pan Jianwei and Zhang Qiang's team at USTC built a public randomness beacon with a device-independent QRNG as its entropy source (2023).
The investment landscape
QRNG is no longer a lab curiosity. Segment data from Precedence Research implies a market of roughly $1 billion in 2026, with segment growth rates of 21–24% a year through 2035. The fastest growth is in quantum-computing applications, cloud delivery, and banking, financial services, and insurance (BFSI). Treat any single market estimate cautiously, since analyst figures for this category vary widely. The more reliable signals are structural:
Consolidation. IonQ's purchase of IDQ puts QRNG, quantum key distribution (QKD), and quantum computing under one roof.
Sovereign capital. India's National Quantum Mission is leading rounds directly, and EU programs fund quantum communications infrastructure.
Software delivery. Quantum Origin shows min-entropy can be sold as validated software, not only as hardware.
Compliance pull. CNSA 2.0, NIST IR 8547 and the EU roadmap turn randomness from a technical detail into an audit item.
For family offices and LPs, the investable layer is less about "who has the most qubits" and more about who owns the certification. Certification means validated entropy sources, quantum-proof extraction, and audit tooling that satisfies a FIPS 140-3 lab.

Risk assessment: red flags in quantum randomness claims
"Quantum" without a min-entropy bound. Ask for the SP 800-90B validation certificate or a published security proof. Quantum-origin noise isn't automatically secure.
Statistical tests offered as proof. Passing statistical test suites like Dieharder or NIST STS shows the output looks random. It says nothing about what an attacker knows. Min-entropy needs a model or a certificate.
Classical-only extractor proofs. Confirm the post-processing is quantum-proof (two-universal hashing, Trevisan).
Unmodeled classical noise. Electronic and thermal noise inside a QRNG may be observable or controllable by an attacker. Rigorous designs count only the quantum part as min-entropy.
No health tests. SP 800-90B requires continuous health tests, such as repetition-count and adaptive-proportion tests, to catch a source that degrades in the field.
Strategic recommendations for executives and investors
Inventory your entropy sources alongside your crypto inventory. Most PQC programs map algorithms and certificates. Add every RNG, HSM entropy source, and cloud key-management service to the same map.
Require SP 800-90B and SP 800-90C alignment in procurement. Make validated entropy a line item in HSM, IoT, and cloud contracts. That covers QRNG vendors and conventional TRNGs alike.
Match the QRNG trust model to the asset. Trusted-device QRNGs suit volume uses like phones and IoT. Reserve device-independent or certified randomness for root keys, lotteries, auditable public randomness, and high-value financial operations.
Plan to 2027, not 2035. CNSA 2.0's January 1, 2027 gate hits US defense suppliers first. NIST's final IR 8547 (October 27, 2026) deprecates RSA and ECC after 2030 at every security strength, not just 112-bit.
In diligence, ask who you have to trust. For any quantum-security startup, have the founders explain the adversary model behind their min-entropy claim. A precise answer signals technical depth.
Looking ahead: the next 18 months
Expect three developments. First, HQC, selected by NIST on March 11, 2025 as a code-based backup to ML-KEM, was slated for a draft standard about a year after selection, with the final standard expected in 2027. Every new algorithm adds key-generation demand. Second, January 2027 brings the first hard CNSA 2.0 procurement deadline, and the EU deadline for national roadmaps arrives at end of 2026, so randomness audits will move from security teams into compliance reporting. Third, certified randomness will move from Nature papers to service agreements. The open question is cost: verifying 71,313 bits took exascale computing, so early customers will be those who need public, auditable randomness more than cheap randomness.
The bottom line: classical min-entropy told us how much an attacker could guess. Quantum min-entropy tells us how much an attacker could guess while holding part of the universe our key came from. Security architectures for the 2030s need to be designed against the second number.
FAQs
What is min-entropy and why is it important for security engineers?
Min-entropy measures the probability that an attacker guesses a secret on the first try, using the most likely value. It sets a hard ceiling on how many secure key bits a source can produce, which is why NIST SP 800-90B uses it to validate entropy sources.
How does classical min-entropy differ from quantum min-entropy?
Classical min-entropy assumes the attacker holds classical information about your source. Quantum min-entropy, H_min(X|E), allows the attacker to hold quantum information, including entanglement, and to choose the best quantum measurement later. It can be lower than the classical value and can be negative for quantum states.
Does quantum computing break random number generators?
Quantum computing doesn't break physical randomness. It breaks generators whose unpredictability rests on factoring or discrete-log problems, such as Dual_EC_DRBG. Grover's algorithm also reduces brute-force margins, which is why CNSA 2.0 specifies AES-256.
Are quantum random number generators more secure than classical TRNGs?
Only if their min-entropy is properly bounded and certified. A validated classical TRNG beats an uncertified QRNG. The strongest option is a device-independent QRNG, which certifies min-entropy through a Bell test even if the hardware isn't trusted.
How is min-entropy measured in practice?
For classical and trusted-device sources, NIST SP 800-90B combines a suite of statistical estimators, takes the most conservative result, and adds continuous health tests. For quantum-proof guarantees, min-entropy comes from a security proof or Bell-inequality certification rather than from statistics alone.
What do PQC algorithms need from randomness?
ML-KEM, ML-DSA, SLH-DSA (FIPS 203, 204 and 205, August 2024) and the coming HQC standard all require high-min-entropy seeds from an approved random bit generator. A PQC key generated from a weak source is weak, regardless of the algorithm.
Subscribe to Unbound
Get weekly insights on emerging tech trends, funding rounds, and strategic opportunities, straight from the field. As a Fractional CTO working across AI, blockchain, quantum, IoT, and Web3, I share what I'm seeing in the trenches, not what I'm reading in press releases. Subscribe to Unbound →
Barbara Bickham is Founder & CTO of Trailyn Ventures and writes Unbound, weekly deep-tech intelligence for executives, investors, family offices, and LPs.
Sources
König, Renner and Schaffner, "The operational meaning of min- and max-entropy," IEEE Transactions on Information Theory 55(9), 2009 (arXiv:0807.1338): https://arxiv.org/abs/0807.1338
Renner, "Security of Quantum Key Distribution," PhD thesis, ETH Zurich, 2005 (arXiv:quant-ph/0512258): https://arxiv.org/abs/quant-ph/0512258
Gavinsky, Kempe, Kerenidis, Raz and de Wolf, "Exponential separations for one-way quantum communication complexity, with applications to cryptography," STOC 2007 (arXiv:quant-ph/0611209): https://arxiv.org/abs/quant-ph/0611209
De, Portmann, Vidick and Renner, "Trevisan's extractor in the presence of quantum side information," SIAM Journal on Computing, 2012 (arXiv:0912.5514): https://arxiv.org/abs/0912.5514
NIST, "NIST Publishes SP 800-90C" (Sep 25, 2025): https://csrc.nist.gov/news/2025/nist-publishes-sp-800-90c
NIST SP 800-90B, "Recommendation for the Entropy Sources Used for Random Bit Generation": https://csrc.nist.gov/pubs/sp/800/90/b/final
PostQuantum.com, "NIST IR 8547 Final" (Oct 2026): https://postquantum.com/security-pqc/nist-ir-8547-final/
Digital Policy Alert, NIST IR 8547 adopted (Oct 27, 2026): https://digitalpolicyalert.org/change/22218-transition-to-post-quantum-cryptography-standards-nist-ir-8547
NIST, "NIST Selects HQC as Fifth Algorithm for Post-Quantum Encryption" (Mar 11, 2025): https://www.nist.gov/news-events/news/2025/03/nist-selects-hqc-fifth-algorithm-post-quantum-encryption
SafeLogic, "CNSA 2.0 and the 2027 inflection point": https://www.safelogic.com/blog/cnsa-2.0-and-the-2027-inflection-point
CyberInsider, "EU charts roadmap for transition to quantum-resistant cryptography by 2035" (Jun 24, 2025): https://cyberinsider.com/eu-charts-roadmap-for-transition-to-quantum-resistant-cryptography-by-2035/
JPMorganChase, "Certified randomness" (Mar 2025): https://www.jpmorgan.com/technology/news/certified-randomness
Quantinuum, "Quantum Origin Becomes First Software QRNG to Achieve NIST Validation" (Apr 2, 2025): https://www.quantinuum.com/press-releases/quantinuums-quantum-origin-becomes-first-software-quantum-random-number-generator-to-achieve-nist-validation
IonQ, "IonQ Completes Acquisition of ID Quantique" (May 6, 2025): https://investors.ionq.com/news/news-details/2025/IonQ-Completes-Acquisition-of-ID-Quantique-Cementing-Leadership-in-Quantum-Networking-and-Secure-Communications/default.aspx
The Quantum Insider, "SK Telecom Partners With ID Quantique for Samsung Galaxy Quantum 5" (Sep 2, 2024): https://thequantuminsider.com/2024/09/02/sk-telecom-partners-with-id-quantique-for-samsung-galaxy-quantum-5/
The Quantum Insider, "Quside Extends Series A to Raise More Than €10 Million": https://thequantuminsider.com/?p=2356245
Raising.fi, "QNu Labs Secures $22.80 Million in Series A1 Funding" (Sep 2026): https://raising.fi/news/qnu-labs-series-a1-september-2026
Chinese Academy of Sciences, "USTC Achieves Zero-knowledge Proof Based on Device-independent Quantum Random Number Beacon" (Nov 2023): https://english.cas.cn/newsroom/news--archives/2023/research-news/index_40.shtml
Precedence Research, "Quantum Random Number Generator Market Size, Share, and Trends 2026 to 2035": https://www.precedenceresearch.com/quantum-random-number-generator-market

