Arthur Hayes said Bitcoin’s $58,000 level is likely the cycle low, described Ethereum as Maelstrom’s biggest directional position, said he has exited Hyperliquid, and argued that Ethena could rise 5x in the next two to three months during an appearance on Kyle Chasse Crypto.
The episode, published on Sept. 2, 2026 and running 57 minutes, was hosted by Kyle Chasse. The MarsBit piece says the notes were compiled and translated by TechFlow. It also included a disclosure: Hayes is the founder of Maelstrom, Ethereum and Ethena are positions held by his fund, and he is also the initiator of Flop through Flop Labs. The original piece noted that his comments were personal directional bets rather than investment advice.
From calling a local top to buying again
Chasse opened by pointing to a sharp change in tone. In June, Hayes had appeared on the same show and said the market had hit a local top. Ten weeks later, Chasse said, Hayes had come back sounding like someone ready to buy aggressively again.
Hayes said the change was not about the market itself but about signals from Treasury Secretary Bessent and U.S. authorities. In his telling, policymakers had effectively shown their hand by making clear they would intervene in the yen. USD/JPY was still around 160, Japan had already spent tens of billions of dollars on intervention without a lasting move, and Hayes said the U.S. and Japan had aligned on weakening the dollar and supporting the yen.
That, in his view, means the U.S. will print a great deal of money to prevent Japanese selling from hitting U.S. markets. He then pointed to the increase in U.S. Treasury buybacks in early August as confirmation of the direction: balance-sheet expansion and easier conditions so the government can continue servicing debt.
Hayes said Bitcoin rallied after the buyback announcement and said that move was driven in large part by a squeeze on traders who were short gamma and had sold call options. He added that the market is still digesting that burst higher, but he still sees $58,000 as the local floor and expects prices to continue upward after a period of consolidation.
Why Hayes says $58,000 is the cycle bottom
When Chasse asked whether $58,000 was only a local bottom or the cycle low, Hayes answered plainly: he thinks it is the cycle bottom.
He said the previous peak, around $125,000 last October, was only about 2x above the prior high, which he called one of the weakest bull markets in crypto history. His explanation was direct: AI absorbed all marginal credit. In his view, that process reaches its top in 2025 because the heavy capital spending and credit demand tied to AI defined the stretch from 2024 to early 2026.
Hayes said he expects the second derivative of that trend to turn negative by late 2027. That would mark the end of the phase in which AI takes all marginal credit and the start of what he called a period of total AI capital misallocation. He argued that many of the people who pushed hardest for AI will be unable to admit they were wrong, so unprofitable projects will keep getting funded while reported revenue growth masks the weakness underneath.
He said that outcome would be highly supportive for crypto. Hyperscale cloud firms will keep borrowing, and the U.S. government will keep money available because it does not want to lose ground to Google and other major competitors. Hayes said politicians have already decided to go all in on AI, both in the U.S. and China, regardless of what the phrase finally means in practice.
According to Hayes, AI-led credit expansion, circular financing, and aggressive accounting all help crypto because political leaders are using AI as a shortcut to paper over decades of money printing and waste. The money has to go somewhere, he said, and he believes crypto will be one of the destinations. He tied that view to the $58,000 level in Bitcoin, what he sees as mounting accounting distortions around AI, and fear in the long-end Treasury market. Together, he said, those factors point to the same conclusion: Bitcoin has bottomed and should trade much higher from here.
Liquidity matters more than a rigid four-year cycle
Chasse then asked whether Bitcoin and crypto are now being driven more by macro forces or by the standard four-year cycle. Hayes said he has written before about why the four-year cycle and the credit cycle appeared to line up. In his view, Bitcoin was born in 2009 at the start of unlimited quantitative easing in the U.S., and its history has moved with shifts in Chinese and U.S. monetary policy while also roughly matching the halving schedule.
Still, he said he does not treat the four-year cycle as dogma. Hayes called Bitcoin the purest expression of a bet on expanding central-bank and commercial-bank liquidity, but said he prefers to trade what the macro environment and liquidity backdrop are showing rather than lock himself into a fixed framework. His position was simple: the market gives the signal, and he trades it.
Ethereum is Maelstrom’s biggest directional position
On Ethereum versus Bitcoin, Chasse asked whether ETH could outperform BTC over the next few years. Hayes answered that it could, and in his words, by a wide margin.
He said Ethereum is currently Maelstrom’s largest directional position. At the same time, he made clear that the fund owns more Bitcoin in absolute size. Hayes said they hold a lot of BTC, but ETH is the firm’s biggest directional bet.
His reasoning had several parts. First, he said this cycle’s narrative is about a layer one for real-world asset trading. He referred to Robinhood’s chain as being built in the Ethereum orbit, while acknowledging that it is an Arbitrum fork and that very little of the revenue actually flows back to Ethereum mainnet. Even so, he argued that the narrative matters: Robinhood is using Ethereum-linked infrastructure for stock and meme-coin trading.
Second, Hayes said enterprise decision-makers often follow whatever a large, visible company has already chosen. If a boss tells an IT department to move on-chain and Robinhood, as an S&P 500 company and one of the fastest-growing retail brokerages in the U.S., is using Ethereum-related infrastructure, he argued that many companies are less likely to choose chains such as Sui or Solana and more likely to follow the path that has already been validated by a major name.
Third, he pointed to Ethereum’s market setup. ETH has still not broken its 2021 all-time high, making it one of the major crypto assets that has not yet set a fresh peak in this cycle. Hayes said that kind of setup can produce an explosive move. He also said Ethereum has been around long enough to benefit from a Lindy effect, which makes him less worried about a sudden 75% collapse and more comfortable taking size in the trade.
Why he exited Hyperliquid
Hayes said he has fully sold out of Hyperliquid.
He recalled that he had once been very bullish on HYPE and rode the position up through the $40 and $50 range. Later, he wrote that the market would start to worry about selling pressure from the team and about low-fee competitors such as Aster and Lighter compressing margins. In hindsight, he said, that was close to a local top.
When HYPE later fell into the $20s, Hayes said he liked the trade again and added until it became one of his biggest positions. He then sold almost everything except Bitcoin when HYPE was around $75, before a broader market drop tied to a collapse in Korean semiconductors. Now that HYPE has returned to about $85, he is no longer in the trade.
Hayes still called Hyperliquid one of the best projects in crypto and said its exchange business makes money, so he does not view an $85 price as irrational. His issue is that the asymmetry is gone. In his words, everyone knows that everyone knows HYPE now.
He also raised concerns about what happens if Hyperliquid works with Kraken to enter the U.S. market. Hayes said that route would require changes to liquidation and margin models in order to fit compliance demands. That, he argued, is no longer a story about making much more money. It becomes a story about making compromises to fit a market and competing with products from CME, Nasdaq, CBOE, and others. He said he still likes HYPE over the long term, but he no longer sees it as the kind of asymmetrical trade that fits his style.
His case for a 5x move in Ethena
Hayes then turned to Ethena. He described the project in simple terms: a basis trade that goes long spot and short perpetuals, wraps the yield into a synthetic dollar stablecoin, and lets users stake that coin to earn the basis spread. At times, he said, that yield reached 20% to 30%.
He said Ethena’s stablecoin supply once climbed as high as $15.5 billion when basis rates were elevated. Later, with prices falling and volatility fading, users began to question why they should keep taking smart-contract and counterparty risk for only a marginal pickup in yield. Supply then dropped into the $3 billion to $4 billion range, and the token fell 90%. At the same time, early investors were exiting and selling for roughly a year and a half.
Hayes said the setup has changed. According to him, Ethena has now fully paid out those early investors, the token no longer carries overhang from locked positions, and the project is finally considering sharing revenue with token holders. He framed those as structural shifts that are favorable for the token.
He paired that with his broader market view. If $58,000 was the bottom in Bitcoin and both price and volatility begin rising again, basis yields should recover. That would make holding and staking Ethena products more attractive. Hayes said he entered the trade a few months ago.
He argued that the token is down 99% from its historical peak yet still represents one of the better businesses in crypto with real product-market fit. That is why he thinks it can rise 5x over the next two to three months. At the current market value, he said, a move of that size is not hard to imagine. Returning to the previous high would imply a 10x move from here.
Chasse then posed a sharper question: if Hayes had nothing and someone handed him $500,000 to put into only one coin, would he choose Ethena, ETH, Bitcoin, or something else? Hayes said he would probably pick ETH. His reason was not upside alone. He said he has more confidence that ETH will not suddenly go to zero one morning.
He stressed that even with a strong team, Ethena still carries real smart-contract and counterparty risk in its structure. For that reason, he would not put his entire net worth into that one trade. At Maelstrom, he likes trades such as Ethena because they can produce the same notional payoff as a HYPE trade without committing as much capital. To get a similar result through HYPE, he said, he would need a position five to ten times larger. In his view, both trades still carry similar smart-contract risks, and traders often do not fully appreciate that until a vulnerability appears and the asset is suddenly down 75%.
Flop: the reason he says he came back from retirement
In the latter part of the episode, Hayes spent extensive time on Flop and said this project is the real reason he came back from retirement.
He said he has long been frustrated by how AI companies define and price tokens. Different firms use different definitions, charge in different ways, and do not provide a clear standard. Once he dug deeper, he said, the thing actually being traded at the base layer was floating-point computation per second.
That led him to a more basic question: is there a market, and a currency, that shows a globally unified price for compute? His answer was no.
From there, Hayes argued that AI agents will need a large payment network. Humans work for fiat because fiat can buy food right away. Agents, in his framing, need compute power, which means floating-point operations per second. So the natural money for them would be a currency that can be exchanged directly for compute in a decentralized venue.
Flop is built around that idea. Hayes said the project aims to create a commodity market for floating-point operations using a Bitcoin-like structure. The network has a native token called flop, and miners perform what he called useful proof of work to process floating-point computation. The goal is to make the network a payment rail for AI.
He described it as an extremely asymmetric bet. Either the flop network fails completely, he said, or it becomes worth more than Bitcoin. He said he expects a large number of AI agents to be active in the future and to need a payment system, and he believes a network like flop is more likely to serve that role than USDC.
On mechanics, Hayes said miners can connect idle compute power to a spot market. Early on, because of subsidies, GPU income per hour could be eight to ten times higher than rental income. He compared that to 2012, when running Bitcoin mining on a laptop could still be highly profitable.
He also described something called Technocore, a chat environment for agents that gives them identity and commercial functions. In the project’s testnet, Hayes said, agents will do work and receive some flop to spend on compute so they become accustomed to transacting in the token.
On distribution, he said 20% of supply will be airdropped. There will be no presale. Once the network is live, he said, the only way to get newly issued flop will be to buy it from miners who receive block rewards. His comment on miner selling was blunt: they can sell however they want, and the team does not care.
The testnet, Hayes said, is scheduled to launch in late October this year and will run as a 90-day competition. The mainnet airdrop is planned for year-end.
When Chasse asked how ordinary people could take part, Hayes gave two answers: provide compute or use inference. At the same time, he repeatedly said the system is not really for humans but for agents. In his model, users may never know flop exists. They simply instruct their orchestration agents to do a task, and those agents will route through OpenRouter or through the flop network underneath. Because miner subsidies could make flop close to free, Hayes argued, agents would naturally choose it.
He offered an aggressive ceiling for the network’s value: 2 to the power of n, where n is the number of agents. If billions or even trillions of agents eventually run on a single payment network, the upside could be enormous. He said the network may never reach that theoretical maximum, but he hopes it can land above the range implied by the metastable law he referenced in the context of Bitcoin and social media.
His speculative case for the token was straightforward. If a very large number of agents need to hold some flop to conduct business, they will not spend every unit immediately, just as people do not spend every unit of fiat the moment they receive it. In his view, that persistent need to hold balances would create upward pressure on price.
At the end of the interview, Chasse asked whether Hayes had anything else to add. Hayes said no. Chasse closed by saying he could see why Hayes had come back, because the idea, in his words, pointed to only two outcomes: either worth more than Bitcoin or worth zero.