Over the past two years, AI has upended the technology industry. AI startups have raised capital on an unprecedented scale: OpenAI has secured US$110 billion in funding, four times the size of the largest IPO in history; meanwhile, many entrepreneurs who were not originally working in AI have either pivoted into the field or done everything they can to embed AI into their own companies.
But AI is not only changing entrepreneurs and the technology industry. It is also reshaping the capital markets upstream and rewriting the once-clear division of roles among venture capital, private equity, sovereign wealth funds, and corporate capital.
A recent report by the World Economic Forum provides the broader macro context for this phenomenon. According to the report, more than half of global venture capital funding went into AI in 2025; by the first quarter of this year, that share had surged to an astonishing 80%. OpenAI, Anthropic, xAI, and Waymo alone absorbed nearly two-thirds of global venture capital funding.
These figures reflect more than just the hype around AI. In my view, they reveal a much deeper shift: AI has changed the “physical attributes” of technology companies.
In the past, SaaS was the type of company venture capitalists understood best. For a typical SaaS company, the main costs were talent and cloud expenses, while its asset-light nature allowed venture investors at different stages to align with the company’s growth path from seed round to IPO.
But AI has rewritten that rule. Behind a leading AI company are tens of thousands of GPUs, dedicated data centers, and power contracts spanning decades. Their funding requirements, payback periods, and risk profiles have moved beyond what traditional venture capital can bear on its own.
The deeper shift is that their cost structure has also diverged from that of pure software companies: in software, marginal costs approach zero, but every AI inference consumes computing power and electricity. AI’s demand for infrastructure is so immense that even tech giants are moving into territory once associated with utilities: Microsoft is seeking to restart the Three Mile Island nuclear power plant through long-term contracts, while Google is making large-scale purchases of small modular reactor (SMR) technology to secure stable, low-carbon baseload power. Today’s most advanced “software companies” are no longer what we once understood them to be.
When a single funding round requires tens of billions of dollars, no single type of investor can complete it alone. The market’s response has been the emergence of a “hybrid” capital structure: sovereign wealth funds provide scale, corporate capital provides strategic value, private growth funds provide execution capabilities, and traditional venture capital provides early-stage judgment, all sitting together on the same cap table. This combination, which would have been unthinkable five years ago, has now become the standard configuration for frontier AI companies. Notably, a significant portion of these massive investments is not pure cash, but commitments of computing power. This is a playbook that venture capital firms holding only cash simply cannot replicate.
Under this hybrid capital structure, the role of venture capital itself is also changing, as venture firms must learn to co-invest alongside sovereign wealth funds and corporate capital. But from another perspective, I personally believe the importance of early-stage investing has actually become even greater. AI companies are scaling from zero at an extremely rapid pace. Some have crossed US$100 million in ARR in less than a year. Regardless of how much substance there is behind that figure, it at least shows that the market’s patience in valuing AI companies has been compressed into an extremely short window. By the time the metrics become clear, valuations may already be ten times higher than they were at the start. The only investors who can place bets before that point, and make a judgment while the startup idea is still on a whiteboard, are the earliest-stage investors.
Every technological revolution redraws the boundaries of the capital markets. The railway era of the nineteenth century gave rise to the modern bond market, a process that took nearly half a century; the internet revolution at the end of the twentieth century transformed venture capital from a small-scale experiment into the core source of capital driving the technology industry, and that also took nearly twenty years. AI is reshaping the capital markets on a similar scale, but this time, the timeline appears to be only three to five years.
