When the music stops¶

The Patrician has observed that economic cycles in Ankh-Morpork follow patterns of remarkable predictability. There are periods when money flows freely, when speculative ventures find eager funding, and when everyone is convinced that this time the prosperity will continue indefinitely because the circumstances are different from previous cycles. These periods are inevitably followed by the moment when someone notices that the circumstances are not actually different, that the ventures are not actually profitable, and that the money has stopped flowing. This moment is always surprising to everyone involved, despite being entirely predictable to anyone who has seen a previous cycle.

The technology industry is currently in a phase characterised by abundant capital, minimal concern about profitability, and widespread belief that AI will generate returns sufficient to justify current valuations. Many participants have never experienced different conditions. They have concluded that current arrangements are permanent rather than temporary. The Patrician finds this conclusion familiar. He also notes that in early 2026, Azure growth hit what analysts described as an infrastructure wall and Microsoft’s shares fell accordingly, which is the kind of signal that is only obvious in retrospect to people who were not paying attention before it arrived.

The question is not whether the music stops but when and how abruptly. The people most confident that it will continue indefinitely are typically the people who would be most embarrassed by its stopping, which creates incentive structures that discourage realistic assessment. This makes predicting timing difficult while making the eventual stop straightforward to predict. These are not contradictions.

When interest rates matter again¶

The technology industry operated for over a decade in conditions where interest rates were effectively zero, which made capital cheap and made investors willing to accept minimal returns because alternatives provided even less. This created an environment where growth was valued over profitability and where companies could operate indefinitely at a loss as long as they demonstrated increasing revenue or user engagement.

Interest rates that are no longer zero change this calculation. When safe investments provide reasonable returns, investors require higher returns from risky technology investments to justify the additional risk. Companies that were acceptable investments when alternatives yielded nothing become unattractive when alternatives yield something. The companies have not become worse. The opportunity cost of funding them has become more apparent.

The down rounds where companies raise funding at lower valuations than previous rounds become common when capital becomes selective. Companies that raised at billion-euro valuations discover that current investors value them at hundreds of millions. The companies survive, usually. The early employees discover that their equity is worth what the paperwork said rather than what the optimism implied. This is always a surprise.

When venture capital dries up¶

The companies that survive transitions in funding conditions are those that prepared for change rather than assuming continuation. This preparation involves financial conservatism, which is the practice of spending less money than you have, which sounds straightforward and is not universally practised.

The companies that raised aggressively and spent based on projected future funding find that projected future funding is unavailable and that their current runway was calculated on optimistic assumptions. They pivot to profitability with the urgency of people who have recently discovered what a runway is for. The pivot involves layoffs, which are announced as strategic restructuring. The employees experience them as layoffs.

The Patrician observes that abundant venture capital creates the illusion that all reasonable ventures deserve funding, while scarce venture capital reveals that funding was never about deserving but about supply relative to demand. Many ventures fundable during abundance cannot attract capital during scarcity regardless of their merits. The merits are not the variable that changed.

When the AI hype cycle completes¶

The current AI enthusiasm will complete its hype cycle, transitioning from inflated expectations through disillusionment toward what practitioners call the plateau of productivity. The transition is inevitable. The timing is uncertain. The pattern is reliable enough to have a name.

The first sign is when impressive demonstrations stop translating to business value. The technology continues working but organisations discover that production deployment is harder than pilots suggested, that integration with existing systems is expensive, and that the value generated does not justify the costs. The demonstrations were accurate. They were also not representative of production deployment, which is a distinction that becomes clear at the worst possible time.

The business model failures where companies discover that AI capabilities are impressive but difficult to monetise create disappointment among investors who expected that impressive technology would naturally translate to profitable business. Inference costs that exceed revenue per user. Willingness-to-pay that is lower than delivery costs. Competition that prevents differentiation. These problems are individually solvable and collectively demoralising.

The Patrician’s assessment¶

The music will stop because it always does. The stop will surprise people who should have been prepared. The aftermath will be survivable for prudent participants while being catastrophic for those who treated current conditions as permanent.

Survival requires financial conservatism, unit economics that work at the transaction level, and focus on specific problems where genuine value is provided rather than ambitious visions that require perpetual funding. Companies that developed these qualities during abundance have competitive advantage during scarcity. Companies that did not have a different kind of experience.

The Patrician recommends dancing while the music plays while staying near the door and keeping your coat to hand. When the music stops, you want to exit smoothly rather than be caught in the rush of everyone simultaneously discovering that the party is over and nobody arranged transportation home. He has attended many such parties. The exits are always crowded. The coat retrieval is always chaotic. People who prepared for this outcome fare noticeably better than people who were surprised by it, which should not be possible to be, but consistently is.