The artificial intelligence revolution is being sold as the next industrial transformation. Governments celebrate it, investors worship it, and corporate boardrooms market it as an endless gold mine. But beneath the glossy presentations, trillion-dollar valuations, and futuristic slogans, an uncomfortable question is emerging: how much of the AI boom is genuine innovation, and how much is financial engineering disguised as technological progress? What is unfolding in the global tech industry increasingly resembles the dangerous habits that fuelled both the dot-com collapse and the 2008 financial crisis. Only this time, the numbers are bigger, the narratives more polished, and the accounting structures far harder for ordinary investors to decode. At the center of the concern lies a disturbing phenomenon — the “round-trip revenue loop.” The mechanism is deceptively simple. A tech giant invests billions into an AI startup. But instead of functioning as independent growth capital, much of that money is tied to the investor’s own cloud infrastructure. The startup then spends the same capital back on the investor’s servers and computing services. In effect, the money never truly leaves the ecosystem. The relationship between Microsoft and OpenAI is the most discussed example. Microsoft reportedly invested around $13 billion into OpenAI, much of it through Azure cloud credits. OpenAI then used those credits to train its AI models on Microsoft’s own infrastructure. Microsoft subsequently booked that usage as cloud revenue. Economically, the same corporate pocket appears to be paying itself — first as “investment,” then as “customer demand.” It may satisfy accounting frameworks on paper, but it creates a distorted picture of real market consumption. Investors are led to believe there is explosive independent demand for cloud services when a meaningful portion may simply be recycled capital circulating within a closed corporate loop. This is where AI starts looking less like pure technological progress and more like a bubble driven by optics. The numbers themselves raise difficult questions. OpenAI’s cloud spending has reportedly surged far beyond its own revenue-generation capacity. If a company spends vastly more than it earns yet survives through continuous funding injections from strategic partners, one must ask whether the business model is sustainable or merely subsidized hype. The situation involving Anthropic and Amazon appears strikingly similar. Billions flow into the startup, billions flow back into Amazon Web Services, revenue gets booked, valuations soar, and markets celebrate “AI demand.” But demand created by recycled investment capital is not the same as organic market demand. Then comes the second layer of illusion: valuation markups.
When AI startups achieve higher valuations during funding rounds, the investing tech giants can revalue their own stakes upward and record enormous paper gains. Hypothetical future valuations increasingly begin to resemble present-day earnings. This is how companies can report breathtaking quarterly profits even when underlying cash generation remains under pressure. Meanwhile, the real cash burn continues relentlessly. AI infrastructure is not a virtual fantasy. Data centers require land, electricity, cooling systems, semiconductors, and colossal construction spending. While paper profits rise, free cash flow in several cases has weakened sharply because companies are pouring tens of billions into AI infrastructure expansion. That is what makes this moment dangerous: markets are increasingly rewarding perception over fundamentals. The AI ecosystem is beginning to resemble a hall of mirrors. Startups depend on tech giants for survival, while tech giants depend on startups to justify future growth projections, inflate cloud revenues, and sustain investor excitement. Such interdependence creates systemic fragility. If even one major AI startup stumbles, the shockwaves could ripple through the balance sheets of multiple trillion-dollar corporations simultaneously. History has seen this movie before. During the dot-com era, telecom giants engaged in reciprocal transactions involving fibre-optic capacity swaps, booking exchanged assets as “revenue” to manufacture the illusion of explosive growth. Eventually, reality caught up. Investors intoxicated by futuristic narratives were left devastated. Today’s AI accounting structures are not identical, but the psychological similarities are impossible to ignore: circular financial flows, valuation euphoria, inflated expectations, and the widespread belief that traditional business fundamentals no longer matter because “this time is different.” That phrase has destroyed fortunes throughout financial history. None of this means artificial intelligence itself is fake. AI is real, transformative, and capable of reshaping industries, medicine, defence, education, and productivity. But transformative technologies can still exist inside speculative bubbles. The internet was real too — yet the dot-com bubble still burst. The real concern is not innovation. It is manipulation. When corporations begin manufacturing the appearance of demand through internal capital recycling, when paper valuations are treated like realized profits, and when investors celebrate accounting optics more than genuine economic output, the system enters dangerous territory. The greatest threat posed by irresponsible AI commercialization may not be the machines themselves. It may be the humans using them to engineer financial illusions on a scale the world has never seen before.
