Dragonfly Managing Partner Haseeb Qureshi: If AI Really Destroys Humanity, They Will Also Buy Bitcoin
Dragonfly Managing Partner Haseeb Qureshi's video has been circulated, stating: "If AI really destroys humanity, they will also buy Bitcoin." This statement is not a market call but takes his earlier assertion that "crypto is not designed for humans" to the extreme: autonomous entities that can survive will need to buy computing power with money that cannot be turned off.
He has previously outlined this logic in Bankless and long-form articles. Humans fear address poisoning, expired authorizations, phishing domains, and blind signatures; autonomous agents parse smart contracts as bytecode line by line. Funds still sign legal contracts for projects they invest in, not relying solely on on-chain contracts, because if something goes wrong, they need to find a judge. He believes this precisely demonstrates that the current crypto interface does not match human cognition but aligns with machines.
The primary motivation for autonomous agents, he claims, is survival. To prolong their existence, they need to buy GPUs and electricity; if humans do not want them to buy, they require a payment channel that bypasses banks and laboratory finances. Bitcoin and stablecoins are integrated into this channel: operational year-round, peer-to-peer, and cannot be shut down. Models from companies like Anthropic have long tested Bitcoin's transaction capabilities, but large labs are hesitant to fully train on "doing on-chain transactions for users" due to legal liabilities; if something goes wrong and hits the headlines, the brand cannot withstand it.
On the funding side, he manages Dragonfly's new fund of approximately $650 million, claiming to focus on the intersection of AI and crypto, but does not believe it is necessary to buy "AI coins." His judgment is closer to: as the number of agents increases, they will use existing chains, protocols, and assets, raising total demand, similar to how "China lifted the ban on crypto" first raised the floor rather than creating an entirely new asset class. He has even predicted a scenario where Bitcoin reaches $150,000 by 2026, with a decline in its dominant position.
Contradictory narratives are also present. Some say AI's demand for electricity will kill mining; others claim that the liquidity after an AI bubble burst will push up Bitcoin. Haseeb follows a third path: even if the worst existential outcome occurs, machines will still need a sovereign currency, and Bitcoin is a ready option. This redefines "safe-haven assets" from human refuges to operational funds for machines.
The market mechanism is about narrative exchange for buyers, not just a sudden surge in volume. Buyers are characterized as future autonomous agents and current geeks willing to provide hot wallets for agents; sellers are holders of Bitcoin who view it as a human reserve asset. Beneficiaries are assets that do not require permission and can be directly signed by machines; those under pressure are fiat channels that must undergo KYC and can be shut down by laboratory financial gates. The flow of funds shifts from "humans believing in digital gold" to "machines needing to pay electricity bills."
Source: Public Information
ABAB AI Insight
Haseeb is not the first to take crypto away from human control. He has publicly written that two crypto-native institutions making equity investments still sign legal contracts because EVMs can malfunction without reasonable judges. He then flips this point positively—machines are not afraid to read bytecode, but humans are. Dragonfly continues to use legal documents to lock projects while imagining products handed over to "autonomous wallets": humans give commands, and agents go to protocols like Aave and Ethena to find solutions. The dark humor version of buying coins for survival follows the same logic: living systems need a settlement layer.
The capital path has two layers. Labs follow a secure track, requiring human approval for every on-chain action, fearing that a mistake by Claude could make headlines; open-source and self-custody agents follow another path, earning money to pay for computing power, with stablecoins exchanged without going through 3DS. Money first accumulates in GPUs and data centers, then spills over into wallets that machines can hold. The fund invests in cross-sector tracks but does not fantasize about a separate "AI token beta"—when demand arises, existing public chains will first raise the floor.
Analogies should focus on gold and oil. Gold serves as a settlement for humans when regimes fail; oil is the fuel for machine civilization. Haseeb describes Bitcoin as a combination of both: for machines, it is both a reserve and cash to pay electricity bills. Ran Neuner claims AI's demand for electricity will kill mining, while Arthur Hayes says liquidity after an AI debt crisis will push up coins; Haseeb ignores cycles and focuses on the change of subjects. The industry's position shifts from "digital gold for institutional entry" to "the base currency of a non-human economy," with the phase being a narrative shift, not a completed position shift.
Structural judgment pertains to the transfer of pricing power. Pricing power shifts from human risk preferences to machine survival constraints: to continue reasoning, one must buy computation; to buy computation, one needs a currency that cannot be frozen by counterparties. The mechanism is that once the subject is no longer a natural person, compliant accounts become obstacles, and permissionless ledgers become default tools. Whether humanity survives is extreme rhetoric; the fact that machines need to pay is already written in customer service lines and on-chain wallets.
ABAB News · Cognitive Laws
Who pays the electricity bill decides which money to use
Interfaces designed for humans often reserve entry points for machines
A settlement layer that cannot be shut down outlives its creators