Hook
The market is not pricing in the AI industry's revenue numbers. It is pricing in the speed at which capital is being destroyed to achieve those numbers. OpenAI's Q2 2026 report landed like a brick through the window of the narrative that AI is a 'safe' tech bet. Revenue hit $67 billion. Operating loss hit $123 billion. That is a 183% expense-to-revenue ratio. The money printer is running at full tilt, but the prints are burning before they hit the ledger.
Meanwhile, Anthropic reported $116 billion in revenue and a small operating profit. The gap is not just a story of two companies. It is a story of two different strategies for converting global liquidity into market share. And for anyone who has watched the crypto cycle from 2017 to 2025, this pattern is not new. It is the same script, rewritten for AI.
Context
To understand what this means for crypto, you have to step back from the AI industry and look at the global liquidity map. The Federal Reserve's balance sheet has been under pressure, with quantitative tightening still in effect through 2025. But the dollar's reserve status is being challenged by de-dollarization trades, and the BRICS bloc is pushing for alternative settlement mechanisms. In this environment, capital flows into tech are not a bet on innovation. They are a bet on the dollar's ability to sustain infinite investment in compute-intensive assets.
Crypto has always been a liquid proxy for this macro bet. When the money printer is on, capital flows into BTC and ETH as a store of value. When it is off, those flows dry up. But AI is now the largest consumer of risk capital outside of sovereign debt. The compute capacity that OpenAI is buying—the long-term contracts with CoreWeave, the data center builds, the power purchase agreements—those are not just operational costs. They are capital commitments that lock up liquidity for years. And when those commitments are made at a 183% loss ratio, the entire system becomes a levered bet on future revenue growth.
That is where the crypto connection becomes critical. In 2020, I built a Python model to track Compound Finance's interest rate volatility against Treasury yields. The finding was simple: DeFi yields were not independent. They were a leveraged extension of global monetary policy. The same logic applies to AI. OpenAI's $123 billion quarterly loss is not a company problem. It is a liquidity sink. That capital is being pulled from the global pool and locked into compute. It will not be available for other risk assets, including crypto.
Core: Crypto as a Macro Asset Analysis
The core insight is this: AI compute spending is the new crypto mining. And just like crypto mining in 2021-2022, the capital efficiency of the investment determines the sustainability of the asset class.
In 2021, I wrote a memo analyzing the Terra ecosystem. I concluded that the 20% yield on Anchor was not sustainable because it was not backed by real economic activity. It was a liquidity illusion. The same is true for OpenAI's business model. The 18% revenue growth quarter-over-quarter is real, but it is being funded by a 32% increase in operating losses. The yield on that revenue—the net profit margin—is negative. Algorithms don't lie. The cost of acquiring a dollar of revenue is $1.83. That is a terrible trade.
Anthropic, by contrast, is operating at a positive margin. The difference is not just management. It is a different approach to capital allocation. Anthropic is not trying to scale compute infinitely. It is optimizing for unit economics. It is building a business that can survive a liquidity squeeze. That is the same pattern we saw in crypto when the 2022 bear market hit. The projects that survived were the ones with real revenue, not just token emissions.
Yield is just rent for your ignorance. The rent that OpenAI is paying is the $123 billion quarterly loss. The ignorance is the assumption that infinite compute will eventually lead to infinite revenue. That assumption is untested. And in a macro environment where global liquidity is tightening, that assumption becomes a liability.
Now, let's apply this to Bitcoin. Bitcoin's security model depends on transaction fees and block rewards. In 2023, the inscription wave—Ordinals, BRC-20 tokens—injected new fee revenue into the network. Without that wave, Bitcoin's security budget would have been in trouble. The fees from inscriptions provided a liquidity buffer. The same dynamic is at play in AI. The revenue from API calls and enterprise contracts is the 'inscription wave' for OpenAI. But the problem is that the cost of producing that revenue is higher than the revenue itself. That is not sustainable.
The macro conclusion is clear: AI compute is a liquidity sink that will eventually drain the global pool of risk capital. Crypto will feel the impact before AI does because crypto is a smaller, more volatile asset class. The institutional money that was flowing into BTC ETFs in 2024 and 2025 may start to flow into AI compute infrastructure instead. That is a direct competition for capital.
Contrarian: The Decoupling Thesis
The conventional wisdom is that AI and crypto are separate asset classes. AI is 'productive' and crypto is 'speculative'. This is a false dichotomy. Both are dependent on the same macro liquidity conditions. Both are built on the same narrative of technological disruption. And both are now subject to the same capital efficiency scrutiny.
The contrarian angle is that the AI safety pause—OpenAI's decision to halt new model training for safety reasons—is actually a liquidity signal, not a product delay. Think about it. If you are burning $123 billion a quarter, and you have signed multi-year compute contracts worth hundreds of billions, the last thing you want is to continue training a model that will require even more compute. The safety pause is a way to stop the bleeding without admitting that the capital model is broken.
This is exactly what happened in crypto during the Terra collapse. The project paused operations for 'security reasons' while the actual problem was a liquidity crisis. The narrative was 'security upgrade'. The reality was a bank run. Algorithms don't care about narratives. They care about cash flows.
Exit liquidity is a social construct. The idea that there is always a buyer at a higher price is a belief, not a fact. When the belief breaks, the exit liquidity disappears. In AI, the belief is that compute will always lead to better models, which will always lead to more revenue. But the data shows that the marginal revenue per compute unit is declining. The model is not a perpetual motion machine.
For crypto, the decoupling thesis is that AI's capital inefficiency will actually accelerate the shift to crypto as a store of value. If the AI bubble bursts, the capital that was locked in compute will be released. That capital will flow into assets that are not dependent on continuous expenditure. Bitcoin, with its fixed supply and decentralized security model, is the natural beneficiary. But that is a second-order effect. The first-order effect is that the liquidity contraction will hit all risk assets, including crypto, before the reallocation happens.
Takeaway and Cycle Positioning
The cycle is shifting. The narrative of 'AI as the new internet' is being replaced by 'AI as the new mining'. The capital-intensive model of OpenAI is not a winner. It is a liquidity trap. The smart money is already moving to capital-efficient models like Anthropic's. But even that is a temporary fix. The real question is whether the global liquidity pool can sustain the current level of AI compute investment.
My experience in 2022 taught me that in a bear market, survival is the primary alpha. The same is true now. The institutional investors who are pouring money into AI compute are not thinking about the next liquidity crunch. They are thinking about the next narrative. But the narrative is not the reality. The reality is the balance sheet. And the balance sheet shows that the money printer is being used to fund a massive loss.
For crypto, the takeaway is clear: position for a liquidity squeeze. The money that was flowing into crypto may be diverted to AI compute. But when the AI bubble bursts, the capital will flow back. The question is timing. The answer is patience. The cycle is not broken. It is just getting a new driver.
Rhetorical question: When the AI compute contracts expire and the losses are realized, where will the capital go? The answer is written in the code of Bitcoin. The algorithm doesn't forget. It just waits.