I haven't really found too much of value in what Rochard has been saying lately, and maybe this is just more coping. But it is an interesting way to tell the story of Bitcoin as it fits into macro markets.

It's a variation of the "AI sucked up all the money" story, but with a few twists:

Bitcoin’s next major move is likely to come not because AI fails, but because AI succeeds just enough to create a historic overbuild. When the capex boom turns into excess capacity, when excess capacity turns into margin pressure, and when margin pressure turns into a crash in the most crowded AI names, capital will look for something that is not another depreciating claim on compute. It will look for savings.

As seems to be the case with most bitcoiners these days, Rochard may be hoping for a crash:

AI is especially vulnerable to this pattern because the asset lives are long and the technology cycles are short. A data center may be financed on a decade-long demand assumption, while model efficiency, chip architecture, inference optimization, open-source competition, regulation, and customer willingness to pay can change in quarters. The risk is not that AI disappears. The risk is that the industry builds for a revenue curve that arrives later, smaller, or in a different part of the stack than investors expect.

Once people realize the AI bros aren't going to provide the returns they wanted, they'll come crawling back to us Bitcoiners...right?

I don't know about all that, but I do know that Bitcoin is still the best solution to a growing problem: capital controls and the disaster that is state controlled money.

here's the whole article in case you don't want to click in to twitter

From Software Surplus to AI Absorption, and Back to BitcoinFrom Software Surplus to AI Absorption, and Back to Bitcoin

The last great technology boom was a deflationary miracle disguised as a venture-capital frenzy. The next one is an industrial buildout disguised as software. That difference explains a great deal about why bitcoin thrived in the 2010s, why its returns have been more muted in the higher-rate, AI-capex-heavy 2020s, and why the next major capital rotation may be ferocious.

In 2011, Marc Andreessen captured the spirit of the coming decade with the line “software is eating the world.” The core idea was not merely that software companies would become valuable. It was that software would invade every incumbent industry because distribution, computing, and startup formation had become radically cheaper. Andreessen noted that a basic Internet application that cost roughly $150,000 per month to run in 2000 could cost about $1,500 per month in Amazon’s cloud by 2011. That was the economic engine of the 2010s: a global market, near-zero marginal distribution, scalable code, and dramatically lower infrastructure requirements for startups. (Andreessen Horowitz)

This did not mean the 2010s were capital-free. The cloud itself required data centers, fiber, servers, and semiconductor supply chains. But for the marginal software company, the business model was relatively capital-light. A small team could rent infrastructure, write code, acquire users globally, and scale revenue faster than fixed assets. Venture capital funded payroll, sales, product, and customer acquisition more than steel, turbines, substations, or gigawatt-scale power commitments. The result was a world in which software firms could absorb attention and equity risk without absorbing the whole pool of physical savings.

That mattered because the 2010s were also a low-rate decade. After the global financial crisis, short-term rates were pushed to the effective lower bound and risk capital was forced outward. The federal funds rate is the central overnight rate in the U.S. financial system, influencing other rates across the economy, and the post-crisis regime made cash and bonds unattractive stores of value for many investors. (FRED) When the opportunity cost of holding non-yielding assets is low, investors become more willing to own long-duration growth assets, venture funds, private tech equity, gold, and eventually bitcoin.

Bitcoin was one of the great beneficiaries of that environment. It was born in the wake of the financial crisis, but it monetized during the software decade. It required no management team, no cap table, no quarterly earnings, and no discounted-cash-flow model. It was a monetary protocol competing for savings in a world where fiat savings earned little, where QE had blurred the boundary between money and policy, and where the cultural center of gravity had shifted toward networks, open-source software, and digital scarcity. Bitcoin was not the same trade as SaaS, but it lived in the same monetary weather.

The 2020s changed the weather. After the pandemic stimulus boom, inflation returned, and the Federal Reserve moved rates sharply higher. Even after subsequent cuts, the current target range remains far above the zero-rate world that characterized much of the prior cycle; FRED shows the upper bound of the federal funds target range at 3.75% as of June 2026, with the effective federal funds rate at 3.63% in May 2026. (FRED) (FRED) This is not a small detail. Bitcoin has no coupon. When Treasury bills yield something meaningful, the hurdle rate for all non-yielding assets rises. A bitcoin holder may still prefer absolute scarcity to fiat yield, but the marginal institutional allocator now has a real alternative.

At the same time, the leading technology narrative changed from software leverage to AI infrastructure. The AI boom is not merely another app-store cycle. It is a race for chips, data centers, energy contracts, cooling systems, transmission capacity, and specialized human talent. Goldman Sachs Research estimated in 2024 that data-center power demand would grow 160% by 2030, that data centers could rise from 1–2% of global power consumption to 3–4% by decade-end, and that U.S. data centers could consume 8% of U.S. power by 2030 versus 3% in 2022. (Goldman Sachs) The U.S. Department of Energy reported that data centers consumed about 4.4% of total U.S. electricity in 2023 and could consume roughly 6.7% to 12% by 2028. (The Department of Energy's Energy.gov)

That is the crucial contrast. The “software eats the world” era multiplied companies without forcing every investor to finance a new power grid. The AI era is absorbing savings directly into physical bottlenecks. GPUs are expensive. Data centers are expensive. Power is expensive. Interconnection queues are slow. Cooling is nontrivial. Depreciation is real. And unlike pure software, where the marginal cost of serving another user can be tiny, frontier AI inference and training scale into tangible resource consumption.

This is why the AI boom has crowded out bitcoin. Capital has rushed toward the companies thought to control the bottlenecks: chip designers, foundries, memory suppliers, cloud platforms, data-center operators, power producers, and infrastructure financiers. The market is not just buying future software margins; it is prepaying for a physical buildout. Oracle’s recent AI infrastructure push is a useful example of the new regime: Reuters reported that the company’s heavy AI spending and debt plans spooked investors, with planned fiscal 2026 net capex of $70 billion and a large financing requirement. (Reuters) This is not a handful of engineers pushing code on rented servers. This is capital formation at industrial scale.

The thesis, then, is not that bitcoin has failed. It is that bitcoin has been temporarily starved by two forces: a higher discount rate and an AI capex supercycle. Higher rates raise the opportunity cost of holding a non-yielding monetary asset. AI capex absorbs the excess savings that might otherwise flow into scarce bearer assets. The 2010s created surplus savings because software scaled faster than it consumed capital. The 2020s are consuming that surplus in chips, concrete, copper, cooling, and electricity.

But capital-intensive booms have a habit. They start with shortage, move to panic-building, and end in overcapacity.

The AI buildout can be simultaneously real and overbuilt. Railroads were real. Fiber optics were real. The Internet was real. Housing demand was real. Shale oil was real. In each case, genuine technology or demand did not prevent a financing cycle from overshooting. The more tangible the buildout, the more dangerous the extrapolation. When everyone believes capacity is scarce, everyone finances capacity. By the time the new supply arrives, the shortage may have turned into glut.

AI is especially vulnerable to this pattern because the asset lives are long and the technology cycles are short. A data center may be financed on a decade-long demand assumption, while model efficiency, chip architecture, inference optimization, open-source competition, regulation, and customer willingness to pay can change in quarters. The risk is not that AI disappears. The risk is that the industry builds for a revenue curve that arrives later, smaller, or in a different part of the stack than investors expect.

That is where bitcoin re-enters the story. When the AI capex cycle turns from boom to overcapacity, the capital now trapped in crowded AI tickers and infrastructure financing will search for an exit. If earnings estimates get revised down, if depreciation overwhelms margins, if power costs rise, if debt-funded data-center vehicles struggle, or if compute prices fall because too much capacity was built, the market will rediscover the difference between a productive technology and a good investment at the wrong price.

Bitcoin is the opposite kind of asset. It has no board promising AI monetization. It has no capex budget. It has no debt maturity wall. Its issuance schedule does not accelerate because Nvidia ships a better chip or because a hyperscaler signs a power contract. It is not a claim on future corporate cash flows; it is a scarce monetary asset competing to be savings technology.

There are other factors, and a serious thesis has to acknowledge them. Bitcoin’s 2020s path has also been shaped by the 2022 crypto credit collapse, leverage liquidations, regulatory uncertainty, the rise and failure of various “crypto” intermediaries, stablecoin liquidity, dollar strength, geopolitics, ETF flows, and the halving cycle. The approval of spot bitcoin ETFs in January 2024 was a major structural change, opening access to investors who wanted exposure without direct custody. (AP News) Institutionalization cuts both ways: it increases access, but it can also make bitcoin trade more like a macro risk asset during liquidity shocks.

Bitcoin also has its own risks. Its volatility remains extreme relative to traditional assets. Political and regulatory treatment can shift. Custody mistakes are permanent. Mining economics matter. Some investors will never accept an asset without cash flows. And in a severe liquidity crunch, bitcoin can sell off with everything else before it resumes its monetary role. None of that should be waved away.

Nor should the AI side be dismissed. AI may produce enormous productivity gains. Some AI companies will be generational winners. The best infrastructure owners may earn durable returns. There will be real cash flows, real products, and real improvements in medicine, education, engineering, software development, logistics, and science. The point is not that AI is fake. The point is that a real revolution can still become an over-owned, over-financed, overvalued trade.

That distinction is the heart of the opportunity. In the 2010s, capital-light software and low rates created the conditions for bitcoin’s monetization. In the 2020s, capital-intensive AI and higher rates have suppressed bitcoin’s relative performance by absorbing the world’s speculative and savings capital. But if the AI cycle follows the normal path of infrastructure manias, today’s crowding will become tomorrow’s disappointment. The market will move from “there is not enough compute” to “there is too much expensive compute,” from “every AI dollar deserves a premium multiple” to “who earns the return on this capex?”

When that happens, bitcoin does not need to promise a new product roadmap. It only needs to still be there: scarce, liquid, global, neutral, and outside the liability structure of the AI boom.

That is why the bear market is the window. The attractive time to accumulate monetary savings is usually when the dominant market narrative is elsewhere. Today, that narrative is AI. The world is captivated by GPUs, agents, data centers, and power deals. Capital is crowded into the tickers that have already won the story. Bitcoin, meanwhile, is being treated as yesterday’s trade, a volatile asset waiting for easier money.

That may be exactly backwards. Bitcoin’s next major move is likely to come not because AI fails, but because AI succeeds just enough to create a historic overbuild. When the capex boom turns into excess capacity, when excess capacity turns into margin pressure, and when margin pressure turns into a crash in the most crowded AI names, capital will look for something that is not another depreciating claim on compute. It will look for savings.

The time to build those savings is before the rotation is obvious.