What's left after the AI ​​and cryptocurrency bubbles burst?

Today, the supply chain of capital is being distorted by three phenomena. The first two are speculative bubbles, which threaten to spread from the financial sector to the real economy, as they have done so often over the past 400 years. The third is an unprecedented state-level shock that is forcing an unprecedented dominance into our already complex political and economic system.

The Danger of Bubbles

First, the bubbles. Financial markets—both public and private—are currently experiencing two parallel speculative manias: one centered around cryptocurrency assets, the other focused on stocks of artificial intelligence-related companies.

The cryptocurrency market is inherently a bubble. It lacks any fundamental value support. Holders receive no cash flow from it; its current value depends entirely on the expectation of selling it at a higher price in the future. This is similar to the Dutch "tulip mania" of the 1630s—the speculative objects then also had no intrinsic value. The hype cycle in artificial intelligence falls into a more common category of bubble. Another innovative technology has emerged, and its long-term economic impact remains unknown. While the lukewarm response to OpenAI's GPT-5 model this month may signal a peak in the bubble, the definitive answer remains to be seen. A common characteristic of both bubbles is that investors are willing to pay exorbitant premiums for securities with extremely low liquidity and no governance rights. From retail investors to institutions, funds are pouring into highly speculative, illiquid assets on an unprecedented scale. Both bubbles originated from an unusual financial environment: negative real interest rates and negative real returns on risk-free assets. Once inflation began, investors' fear of missing out (FOMO) drove all the familiar bubble dynamics. The first law of financial bubbles is that it's easy to tell when you're in a bubble, but extremely difficult to tell when it will burst. However, researchers have identified three signals that signal the end of a bubble: 1. An inverted demand curve: Higher prices increase demand. José Scheinkman of Columbia University and Hyun Song Shin of the Bank for International Settlements have shown this phenomenon to have occurred during the dot-com bubble of the late 1990s and before the 2008 global financial crisis. 2. New supply emerges: Exponential price increases attract a large influx of new participants. Even in the digital world, creating a new asset takes far longer than price fluctuations. Cryptocurrency prices fluctuate instantly, and while private equity markets react quickly, building a new large language model (LLM) takes a long time. 3. Amateur investor-led demand: At the end of a bubble, demand increasingly comes from uninformed retail investors. Currently, the crypto and AI markets appear to be flashing all three red flags. However, the catalysts for their respective collapses may be very different. Parting Ways Cryptocurrency prices are entirely dependent on demand, which comes from both existing holders increasing their holdings and new buyers entering the market. The current surge in demand is inextricably linked to the Trump administration's radical deregulatory agenda—an agenda inseparable from unprecedented government corruption, such as the issuance of presidential memecoins. Therefore, the continued prosperity of cryptocurrencies appears to depend on whether Trump and his inner circle can maintain political power. Given the crypto industry's massive investments in lobbying and campaign funding, this political support is likely to last at least until the 2026 midterm elections, or even longer. The artificial intelligence bubble is different. Sooner or later, current high valuations will need fundamental support—positive cash flow from massive investments in computing infrastructure (such as data centers). Unlike cryptocurrencies (and tulip bulbs 400 years ago), those betting on AI need an economically sustainable business model. Admittedly, some companies and entrepreneurs have found economically valuable and commercially viable applications for large language models. But for current AI valuations to even be justified, two conditions must be met: first, the market size of AI applications must be large enough to generate cash flow justifying the massive investments; and second, and equally important, the economic model must achieve a stable equilibrium where all competing suppliers can generate positive cash flow. The second condition may be difficult to achieve. Because the fixed costs of providing AI services are far higher than the marginal cost per unit of service, the economic model is daunting. Once prices converge toward marginal costs, all participants lose money. This is why past technological revolutions have often ultimately devolved into stable oligopolies (which are difficult to sustain) or regulated monopolies. The histories of railroads, electrification, and the internet are all relevant here. Each technology required massive investments in physical infrastructure before finding viable and scalable applications. Today, we take these industries for granted, forgetting that their developments were punctuated by serial bankruptcies and by various instances of state intervention to protect competitors from self-destruction. As so often happens, individuals’ rational responses to incentives led to devastating coordination failures. Throughout, vast amounts of speculative capital, punctuated by periods of financial crisis, financed the construction of these transformative networks. For AI, what analyst Nicolas Colin calls the “compute-energy stack” is the modern equivalent of the railroad tracks, power plants and grids, fiber optic cables, and server farms of the previous two centuries. Once again, capital must first flow into assets whose economic value is still unknown. If the AI ​​bubble can produce long-term, stable, and profitable outcomes without triggering a dramatic bust, it would be unprecedented in the history of capitalism. Killing the Goose That Lays the Golden Eggs The supply chain that provides capital for cutting-edge technology underwent institutional transformation in the 20th century. Before the 1980s, industrial research labs, funded by the monopoly profits of large tech monopolies (DuPont, AT&T, General Electric, IBM, and Xerox), played a key role in the American innovation system. But in 1982, the Securities and Exchange Commission (SEC) ruled that stock buybacks did not constitute "market manipulation," opening up alternative uses for monopoly rents. This use of cash has continued to grow ever since. In 2024, US public companies will repurchase $942.5 billion in stock, more than 50% more than their total corporate R&D spending. Of course, by the 1980s, the federal government's system for mobilizing science to win World War II, the Cold War, and the "War on Cancer" had matured, partially offsetting the retreat of the old tech monopolies. But this leads to the third and most damaging development: the Trump administration's unprecedented attack on American scientific research. Upon returning to the White House, Trump immediately targeted strategic funding agencies like the National Institutes of Health (NIH), the National Science Foundation (NSF), the Department of Energy's Advanced Research Projects Agency (ARPA-E), and the Environmental Protection Agency's Office of Research and Development. Worse still, he is launching a frontal assault on the research universities that have long fostered scientific discoveries and technological breakthroughs, underpinning America's competitiveness. These institutions are being systematically weakened. This destructive plan, spearheaded by Russell Vought, director of the Office of Management and Budget, was publicly previewed in the far-right Heritage Foundation's infamous Project 2025 (which Vought himself co-authored). While its ostensible goal is to dismantle diversity, equity, and inclusion (DEI) programs and all efforts to combat climate change, its deeper mission is to dismantle the New Deal's legacy and return the United States to the political and economic model of the 1920s. Only the Department of Defense will be largely unaffected. While it's impossible to quantify the long-term consequences of this plan, its negative impact on the American economy and, indeed, humanity as a whole, is almost undeniable. America's innovation economy, once the driving force behind unprecedented material progress in human history, is now being systematically crippled.

Where does venture capital go?

Not only did the crypto and AI bubbles predate this shock, but their rise also largely escaped the traditional venture capital cycle documented in the industry classic by Harvard Business School's Paul Gompers and Josh Lerner. While there are some venture-backed crypto projects (like Coinbase) and VC firms like Andreessen Horowitz actively advocating for crypto, the majority of funding still comes from retail investors.

Now, retail speculation in crypto has regained momentum, thanks to the potential for institutional support brought about by deregulation. Without strong state support, it should have been clear long ago that this speculation had reached the self-destructive limits of any Ponzi scheme. Consider the proliferation of the "crypto vault" company model, pioneered by MicroStrategy (recently renamed Strategy). Its enthusiastic founder and executive chairman, Michael Saylor, seemingly created a perpetual motion machine: continuously raising funds at public market valuations multiples of his Bitcoin holdings to purchase more Bitcoin. Dozens of imitators followed suit, but many now trade below the value of their crypto assets; Strategy's stock price fell 15% in August. In other words, they have triggered the second signal mentioned above (price inflation creates new supply). By contrast, the real asset funding supporting the AI ​​bubble primarily comes from large tech platform companies. Some venture capital firms are trying to participate in the AI ​​craze, even if it means abandoning venture capital principles of valuation and governance. In the first half of 2025, OpenAI raised a staggering $40 billion from a mix of sources, led by SoftBank and other "famous bubble investors." In the second quarter, five venture-like deals raised over $1 billion for AI-related companies. Within the venture capital industry, after four consecutive years of minimal exits and distributions (realized profits distributed to investors), the industry itself has become polarized: one faction adheres to its legacy, controls scale, and engages deeply in the governance of early-stage projects; the other transforms into private equity-scale asset aggregators and fee collectors. For the industry as a whole, the increasing focus on AI has depleted the "dry powder" of funds raised during the "unicorn era" of 2021. While the broader financial environment has normalized from the speculative excesses fostered by unconventional monetary policy before 2021, the future structure of venture capital remains uncertain. The problem isn't just that the crypto and AI bubbles still need to unwind. More fundamentally, the industry can no longer rely on the innovation supply chain orchestrated by the US government. Many links in that supply chain are now being weakened or severed by the Trump administration. The venture capital model was born in the US, but its future may belong to China and Europe. As the flow of transformative, science-based technologies dries up, the strategic role it has long played in its homeland will diminish. Rebuilding this supply chain, built over three generations, will be incredibly difficult.