Month: September 2026

Unhedgeable: EMP and lessons for living with risk

Custodying cryptocurrency invites the contemplation of tail risk: low-probability, high-impact events. These are the feverish Hollywood disaster-movie plots that feed into enterprise disaster recovery playbooks: Category 5 hurricane takes out the company headquarters. Flooding renders parts of the country inaccessible by road, making it impossible to operate offline systems that require physical access. Terrorists sabotage the data-center housing production servers. (That one is less improbable given Iran’s recent statement around targeting US tech companies in the Middle East.) Novel pandemic forces social distancing, stressing dual-control policies that require 2 or more employees physically present at the same time to execute critical operations. By definition force majeure events are unpredictable: before COVID-19 few organizations had contingency plans for a world when employees are forced to work from home and no one wants to be in a room with their colleagues when that could increase their chances of catching a mysterious infection.

One of the more misunderstood items in these sum-of-all-fears scenarios concerns EMP or electromagnetic pulse. An EMP is a short (often nanoseconds to milliseconds) and highly concentrated release of electromagnetic energy. These intense bursts of radiation can damage electronic equipment or even bring down entire power grids depending on the scale. As with most risks better understood by contemporary science, major EMP events were observed long before humanity understood how electromagnetic waves operate. In the first recorded major incident known as the Carrington Event, a major cloud of plasma ejected from the sun reached the Earth’s atmosphere in 1859. According to NOAA:

“Powerful electrical currents flowed through telegraph machine wires, causing them to shock their operators and igniting small fires in some offices because of the surge in electrical currents. Some even sent messages without being plugged in or received nonsensical messages that no one had sent.”

It’s not that major solar events never happened before the Carrington Event. But it was industrial age dependence on electrified systems— specifically the telegraph— that manifested these disruptive effects. In a way the timing was fortunate for team humanity: 1859 predates the widespread adoption of power grids in the late 19th and early 20th centuries. In modern terminology a CME or “coronal mass ejection” carries massive magnetic fields spanning hundreds or thousands of miles that can induce electrical currents on anything vaguely resembling a wire made out of conducting material. One of those objects of course are the high-voltage power transmission lines designed to carry electricity in the first place. A more modern demonstration arrived in 1989 when a geomagnetic storm knocked out the Hydro-Quebec grid for several hours, taking the industry by surprise.

Starfish Prime

Chunks of plasma sent our way by the daystar is one type of EMP, but not the only one modern doomsday planners must contend with. The United States accidentally stumbled into an understanding of what EMP can do to modern grids during a 1962 nuclear weapon test. Codenamed Starfish Prime, the 1.4 megaton thermonuclear weapon exploded at a very high-altitude, 250 miles above the surface of the planet. To put that into perspective, the Kármán line marking the unofficial boundary of outer space sits at 62 miles. At such altitudes a nuclear explosion looks nothing like the classic mushroom cloud. Because the fireball never touches the ground, it does not pick up surface material to mix with fission products and precipitate as fallout.

But 900 miles away in Hawaii, residents of Oahu did notice something else: street lights went out and a microwave repeater was knocked out. It turned out even more extensive carnage occurred in the other direction: high up in orbit, where sending a repairman is tricky. Roughly a third of orbiting satellites were damaged by the artificial radiation belt created by the massive explosion, including the very first commercial satellite launched by AT&T after the initial event, because it stumbled into the residual field.

That was the first observed case of HEMP, for High-altitude EMP caused by a nuclear detonation. HEMP briefly occupied the popular imagination with the 1995 Bond movie “Golden Eye,” as remarkable for special-effects as its complete disregard for actual physics. Our fearless 007 faces off against a villain colluding with a rogue Russian general to take over a secret Soviet satellite designed to trigger HEMP events. Their grand plan is to aim this powerful contraption at London and cause havoc in financial markets. Plot construction aside, the portrayal of HEMP leaves much to be desired. Judging by the comical depiction of a narrow beam of light aimed from the satellite down to earth— is this where MTG got her idea of “space lasers” igniting wildfires?— the screen-writers believe HEMP can be precisely aimed at a specific target: the ground station in Russia operating the satellites or even a city on the scale of London. As the Hawaii incident demonstrates, the effective range is hundreds or even thousands of miles. (It depends both on the altitude where the weapon is detonated and the geographic latitude because the earth’s own magnetic field warps the propagation of the one generated by the blast.) And contrary to the widespread destruction inflicted on the Russian base in the opening sequence, HEMP by itself can not cause kinetic damage to anything on the ground. The explosion happens too far away for destructive shock waves to register.

Illogical threat models

Returning to the question casually asked about so-called high assurance systems in finance: “what if our data-center is hit by an EMP?” That often leads into a misguided wild-goose chase for improvised Faraday cages and heavy-duty surge protection equipment with nanosecond response times. The threat model is muddied because the original question conflates two completely different scenarios:

  • EMP caused by natural sources. In other words, a solar disturbance that happens to be precisely aimed in the direction our planet
  • Golden Eye/HEMP variety, a deliberate act initiated by a hostile nation— and quite possibly the beginning of the end for all involved.

CME resistance at scale comes down to grid resilience. The jury is out on the question of whether North American utilities have done enough to safeguard the national grid against an EMP spike from natural causes. Regardless, these risks are best addressed upstream from the data-center floor, covering all customers.

Resilience against HEMP is a much higher bar than CME. It is also very difficult to achieve with garden-variety IT equipment intended for 24/7 operation in a standard data-center rack. Standard enclosures are likely to have gaps for air circulation and running cables to connect power/networking. Even tiny breaks in a continuous surface can act as an antenna and allow the intense initial pulse (called “E1”) from the HEMP to pass through and fry everything of value inside.

For most threat models, HEMP resistance is also completely beside the point. All bets are off when nuclear weapons are being detonated high up in the atmosphere by a hostile nation to damage US infrastructure. There are only a handful of systems that must be designed to absolutely survive that scenario: they are precisely those necessary for continuity of government and for effective military command & control. Their resilience follows from the logic of nuclear deterrence: the country must retain credible retaliation capability in the event of an unexpected first-strike. (Incidentally HEMP would make for a lethal prelude to a decapitation strike: unlike ICBMs, an overhead satellite detonation will not come with a 20-30 minute warning window.) But the systems that need credible survival story in the face of a surprise attack are few and far between. No one cares about their sports-betting site experiencing an outage when the country is on the verge of DEFCON 1.

Trying to design around apocalyptic scenarios is not only a waste of time: it results in a dangerously misguided mindset that these situations can be planned for or hedged against. Cryptocurrency is no stranger to the doomsday mindset, although the standard villains of the narrative are misguided government bureaucrats debasing the currency or too-big-to-fail banks imploding in a final cataclysmic crisis that ushers in a Greater Depression. It is one thing to seek an insurance policy against runaway inflation or money-printers churning out of control. Pretending that existential risk to civilization is somehow a survivable business disruption— another line item in the corporate disaster recovery & incident response plan to impress SOC2 auditors— is delusional.

That effort is best redirected into public-sphere advocacy to help ensure those tail-risks are either contained or never materialize in the first place. Some of the organizations contemplating EMP protection and wrapping servers in tin-foil are influential players in the policy arena. They can afford to devote significant amounts to lobbying for specific policy outcomes. For anyone concerned about CME, the most effective solution is regulation of grid operators: raise the bar on design and implementation of safeguards for transmission lines. For those worried about HEMP, the equivalent policy objective is nuclear deescalation. Reasonable people may well disagree on how that outcome is best achieved. Scale back the arsenal, renew the expired New START treaty or shift the nuclear triad towards submarines and away from use-it-or-lose-it ICBMs? But that is exactly the policy debate that needs to play out in the open and stands a meaningful chance of reducing risk globally, instead of cosplaying at an IT-themed version of Doomsday Preppers.

CP

Optimal heist: strategy for quantum-attacks on blockchains

Earlier essays looked at the problem of detecting attacks against digital assets involving quantum computers, focusing on the difficulty of attribution. There is an unstated assumption there— and more generally, discussions of quantum risk: an adversary in possession of a so-called “cryptographically relevant quantum computer” or QCRC will have no qualms about using it to steal every last penny of vulnerable funds they can locate on-chain. That simplified picture is worth challenging. Even a profit-motivated threat actor realizes there are limits to profiting from stolen goods. Stealing cryptocurrency is one thing, being able to fence/liquidate those assets at anything close to their fair-market value is an entirely different problem. In fact major thefts on blockchains often result in temporary dips in the price; they can even result in dips on completely unrelated chains1 as in the case of the Bybit incident from 2025. An attack using QCRC introduces an entirely new dimension to this tendency of thefts to devaluing the very assets the perpetrator walked away with. Once word gets out that the core principle underlying blockchain assets— person with the private key controls the funds— is no longer true in the presence of quantum-computers, it is no longer a matter of taking a small “haircut” to unload stolen assets: the market value of that asset can go to zero immediately.

To untangle the conflicting incentives and understand how quantum-attacks are likely to play out in the wild, it helps to clarify the threat-actors objectives. Roughly speaking there are three possibilities:

  1. Profit-motivated, effectively going long on the asset. The perpetrator plans to profit by selling the stolen assets. Interestingly there is alignment here between the interests of the attacker and the broader ecosystem— although obviously not with the interests of the specific victim targeted. In order to ensure getting highest price for their ill-gotten goods, the attacker wants to steer clear of driving prices to zero or destroying the entire ecosystem.
  2. Profit-motivated, going short. An entirely different way to profit from theft involves capitalizing on the second-order effects, such as widespread market panic. This is monetizable by betting against the asset, for example by buying put options on one of the bitcoin or ethereum ETFs.2 Flipping the script on the first persona, here the adversary does not care about the overall health of the ecosystem. Shaking investor confidence is the whole point, to the extent that loss of confidence can be exploited in a trading strategy.
  3. The state strikes back: not economically motivated, but ideologically driven to undermine confidence in a particular blockchain. For example a central bank threatened by the rise of non-sovereign monetary systems may want to reassert control by sabotaging bitcoin. The attacker objectives are similar to the previous example but they are no longer constrained by any trading strategy. Suppose a QC attack against Ethereum would cost a billion dollars but the available liquidity for shorting Ethereum ETFs only permits a half billion upside under best-case assumptions—namely, the price of ETH goes to zero. This is not a “useful” attack for the profit motivated adversary; the juice is not worth the squeeze. But it may well be worth it for a central banker nursing a grudge against ethereum or stablecoins.

Given this backgr­ound, consider how each type of threat actor is likely to proceed.

Profit-motivated, long

This is at once the most straightforward motivation— old-fashioned greed— and also the most tricky category for monetization. Cryptocurrency theft is self-defeating at scale: public awareness of a major heist will both tank prices and increase the odds that various players in the ecosystem will overcome inertia to coordinate a response. For example they can reach consensus to treat the stolen funds as “tainted” and refuse to accept them for payment, or seize them if ever transferred into their custody as part of an attempt to cash-out into a different currency. In the worst-case scenario where the amount stolen is a substantial fraction of circulating balance or belongs to a systematically important project— witness the DAO breach of 2016— it may be enough to warrant a hard-fork, the blockchain equivalent of historical revisionism: undoing the theft by rewriting the past. To invert the old saying about credits dynamics for banks:

“If you steal a thousand dollars worth of cryptocurrency, it is the owner’s problem. If you steal a million dollars, it is your problem”

Stolen amounts aside, an attacker armed with a quantum computer faces an extreme version of the price impact: if word gets out that quantum-attacks have left the realm of academic speculation and executed successfully in the wild, the price can go directly to zero. Because it means that the only notion of “ownership” recognized on chain— control over private keys— has become meaningless; property rights have been superseded by a quantum-computing version of might-makes-right.

Here the attacker gets a lucky break: it is very difficult to provide attribution for security breaches, much less prove that a particular avenue of attack was involved. By the public nature of blockchains, everyone can observe that funds moved on-chain and based on protestations from the owner, that this movement was not authorized. Beyond that everything is open to dispute: was the root cause a ground-breaking quantum computer in the hands of the defenders or bush-league hardware wallet with bad entropy in the hands of the defenders? With the exception of addresses deliberately constructed as quantum canaries or tripwires, a quantum attack will not leave a tell-tale signature that can be used as proof beyond doubt that the Rubicon has been crossed. If anything, public opinion and Occam’s Razor will favor the more mundane explanation: garden variety human-error, supply-chain attack, disgruntled employee, infiltration by North Korean IT workers. Any one of these is a priori more plausible than having witnessed a historical first: nation states armed with quantum-computers, putting their exotic machinery to the task of making a quick buck. Also recall that victims have an incentive to exaggerate attacker capabilities: every company putting out a press release in the aftermath of an embarrassing incident refers to “sophisticated attackers” and “advanced techniques;” no one wants to admit their third-rate enterprise IT was owned by a bunch of script-kiddies. All the while more reason for public skepticism against any claim that some attack must have involved a quantum computer. About the only time that possibility would be entertained is if the target happens to be one of a handful of the largest institutional custodians. For example the highest-balance BTC address today belongs to the exchange Binance, at a staggering 250K bitcoin. Suppose those funds vanished tomorrow and the Binance security team swears on a stash of pre-IPO shares that they have found no evidence of a traditional breach of their custody system. In that scenario at least some observers may give Binance benefit of the doubt and classify the episode as the first likely use of quantum attacks in the wild. But this scenario is unlikely for reasons pointed out earlier: for a profit-motivated attacker, going after the highest balance address—and draining 100% of it in one shot— is a Pyrrhic victory.

What does an optimal strategy look like from an attacker perspective?

  • Target small balances for each heist. Large balances are harder to launder and may increase the risk of a coordinated ecosystem response to quarantine or otherwise freeze funds before they can even get to the laundering stage. In the worst case scenario,
  • Avoid wallets belonging to well-known figures or projects associated with security initiatives. Similar to the first concern, with a twist: some of these players may have sufficient clout in the eyes of blockchain users that their claims of the attack originating with a quantum-computer may have a modicum of credibility.
  • Resort to the offensive analog of parallel reconstruction: assuming the wallet owner can be identified, orchestrate a traditional infrastructure breach of the same target as a diversion from the quantum attack. For example, deface the website, dump databases containing customer PII or otherwise create an overt, undeniable security incident to undermine market confidence in the security of that organization. A blockchain theft by quantum computer coinciding with such a traditional breach is more likely to be interpreted publicly as part of a single event— despite protestations to the contrary by the organization insisting their cold-wallets are completely air-gapped from the compromised infrastructure. In that scenario, it would be difficult for the defenders or neutral observers to argue that some shadowy organization with unprecedented quantum capabilities is behind the attack.

Profit-motivated, short

The observation that headline-making blockchain thefts result in panic selling and price dip suggests another intriguing possibility for profiting from a quantum computer. One can carry out the heist but instead of attempting to liquidate the stolen assets, instead profit indirectly from the downward price movement. A “socially-responsible criminal” could even return the funds once their position is closed, alleging that it was all part of an emergency white-hat hacker rescue of vulnerable assets.

This approach runs into several, closely related limitations:

  1. Compared to selling a stolen asset with a haircut, the upside from betting on a marginal price decline is limited. It would require building a substantial position to generate the same returns as laundering what is essentially free money taken from another blockchain address. Unless of course the expectation is that the asset value will go all the way to zero, in which case see point #5.
  2. Short-selling or buying put options on that scale requires locating massive liquidity in thinly-traded markets to take the opposite side. In theory short interest in asset can exceeds its market cap—as the GameStop debacle demonstrated— but this is not the case for most cryptocurrencies.
  3. Building up these positions will require upfront investment of capital. Put options are not free. A short position has collateral requirements proportional to risk. (Perpetual futures may offer one way out of this problem.)
  4. It creates exposure to counter-parties and trading venues. Exchanges that allow margin trading and perpetual futures actively monitor open positions, and may close out even profitable positions to avoid getting into a state where the counter-parties on the other side exceed risk tolerances. Alternatively the venue itself may go under, converting an otherwise winning bearish bet into a loss before the position can be closed.
  5. In the extreme case, if the theft is convincingly attributed to a quantum computer, it would completely destroy confidence in not only the security model of that particular blockchain but virtually every digital currency. Even if the original theft occurred on say Bitcoin, the contagion would spread to wipe out Ethereum, Solana and every other layer-one dependent on the same vulnerable cryptography. In that case it becomes unclear if realized paper profits can be captured. Trading would likely be halted and some option contracts may not be fully settled. There would be a total, correlated crash of the market.
  6. A massive, well-timed short-position right before a major price decline attracts attention. An exchange with a robust market integrity program could investigate further or notify regulators of what looks like potential insider trading. (It is insider trading in one sense— the customer was privy to material non-public information, namely that they were about to commit a heist that would result in a sharp price decline.) Threat actors are more likely to get away with this on smaller, less-scrutinized venues or on-chain decentralized exchanges, but then the liquidity problem is worse.

All of this also assumes that the theft in question will trigger a price crash at the right time. Depending on exactly how the event plays out, it may even have the opposite effect. For example if the stolen funds were “burned” either explicitly by the perpetrator moving them to a blackhole address or by consensus among the community to treat those funds as blacklisted— due to say OFAC sanctioning the address— the net effect is a reduction in the money supply. For a deflationary system like bitcoin where the supply is capped, that should theoretically increase the value of every other bitcoin held.

For all these reasons, betting on a temporary price dip does not look any more profitable than old-fashioned theft.

Undermining cryptocurrencies

While a precision strike to profit from temporary price shocks are tricky to engineer, quantum-armed attackers will have an easier time arrange for what might be called the “nuclear option:” drive the price of almost every digital asset to zero, by proving that the notion of ownership as recognized cryptographically on blockchains is meaningless. It is not even necessary for a single penny to be “stolen” on-chain in the traditional sense. Simply proving that the capability exists— for example, by recovering a private-key for a deterministically generated, so-called “nothing up my sleeve” or NUMS public key that controls exactly zero value on chain— would send the same message just as convincingly.

That nuclear option however carries a strategic cost: the attacker is forced to tip their hand and provide proof that a particular attack was carried out using a quantum computer. This is the inverse problem of the attacker trying to fence stolen assets without inducing market panic. Recall that monetizing a quantum theft requires going to great lengths to disguise an extremely sophisticated attack as a garden-variety security breach. A government hell-bent on destroying bitcoin has the opposite problem: they must prove beyond doubt that a QC capability exists and this capability can target any exposed public-key.

The intelligence downside of such a disclosure would be enormous. As long as everyone believes that quantum computers are hypothetical risks belonging to the distant future, complacency reigns. There is no rush to replace classical algorithms with quantum-safe counterparts. At least some potential targets of interest will continue to use vulnerable systems. That is a win from the intelligence perspective: more encrypted traffic that unbeknownst to the enemy can be intercepted and decrypted, more vulnerable authentication systems that can be bypassed. Disclosure of QC capabilities in the wild would serve as a wake-up call, alerting potential adversaries to the exposure. There is historical precedent for cryptographic breakthroughs being protected at great cost, by deliberately not acting on all intelligence obtained from successful decryption. Occasionally letting the enemy get away with known movements in some minor skirmish was deemed preferable to revealing a capability that can change the course of the conflict.

That creates a dilemma for a nation state threatened by the existence of non-sovereign money. No matter how much disdain central bankers may harbor for bitcoin, states’ willingness to wipe out the entire cryptocurrency ecosystem must be balanced against the intelligence value of keeping the existence of QC capabilities secret. As a weapon, cryptographic capabilities have an inverse logic compared to nuclear weapons: states crossing the threshold of nuclear capability rush to announce their new-found prowess with a mushroom cloud, in order to put their adversaries on notice and reshuffle the balance of power. A state that builds a quantum-computer capable of breaking widely deployed algorithms has great, vested interest in keeping that breakthrough a closely guarded secret, in order to lull adversaries into a false of security about continuing to use those algorithms.

What this suggests is that a nation-state may redirect their QC capability into economic objectives— such as systematically trying to undermine all cryptocurrencies— only when disclosure of that capability is already imminent and can not be prevented.

CP

1 This is not completely irrational investor psychology, even as some critics pile on cryptocurrency investors as exhibit A for irrationality: stolen Ethereum is often converted to other assets by bridging or trading for wrapped-equivalents on decentralized exchanges. So attackers’ end-game may well involve a market-crashing fire-sale of a different asset than what they originally stole.

2 It is also possible to bet against assets using DeFi lending pools but it is unclear how well any of those mechanisms would operate in the event of widespread chaos.