The last time capital concentrated in a single sector like this, it built the modern internet. That was the dot-com era. At its 2000 peak, U.S. venture capitalists deployed roughly $112 billion across all sectors combined. In 2025, AI companies alone raised $258.7 billion. More than double that. In a single year.

Before the world could see it, the money already moved

Researchers theorise about the future. Analysts forecast it. A much smaller group bets hundreds of millions on a conviction before there is any proof it will pay off.

That is venture capital, and its logic is unforgiving. Roughly three-quarters of VC-backed companies never return investors' capital. The math only works because the rare wins are extraordinary. A single Google or OpenAI generates returns large enough to justify every other loss. That structure forces a particular kind of rigour: you cannot afford to be vaguely optimistic. You have to be specifically right.

A handful of funds have built a track record that is hard to ignore. Sequoia and Kleiner Perkins co-led $25 million into Google in 1999, before it had a business model and before anyone knew if it would survive the dot-com crash. Sam Altman, then president of Y Combinator, co-founded OpenAI in 2015 as a non-profit with no product and no clear path to revenue. Both looked speculative at the time. Both reshaped how billions of people live and work.

The same investors are now all pointing in the same direction.

The funds that called Google and OpenAI are all pointing the same way

Sequoia, Andreessen Horowitz, and Y Combinator are three of the most consequential early-stage investors in the world. Their collective portfolio spans companies that built the modern internet: Google, Airbnb, Stripe, and OpenAI among them. They have different portfolios, different theses, and no reason to align. And yet in 2026, all three have independently arrived at the same conviction: AI has stopped being a feature built on top of existing systems. The foundation is now being built underneath.

Where they differ is which part of that shift they are backing first.

Sequoia is betting on the move from AI that talks to AI that acts. Their 2026 thesis draws a clean line: AI applications in 2023 and 2024 were talkers. Those coming in 2026 and 2027 will be doers. Agents that read data directly, execute tasks, and log outcomes without a human clicking through a single screen. Enterprise software was designed for humans to navigate. Agents do not need interfaces. The software layer does not get disrupted by a better version of itself. It becomes unnecessary. The next decade belongs to companies that build for a world where interfaces no longer matter.

Andreessen Horowitz is betting on data as the new moat. As AI models grow more accessible, the model itself becomes a commodity. Their answer: the advantage lives in what the AI learns from. Healthcare records, legal case histories, financial transaction patterns. Data that cannot be replicated and that makes the AI harder to displace with every passing month. The model is the commodity. The data is the moat.

Y Combinator is betting that AI has crossed into the physical world. Their Summer 2026 Requests for Startups puts hardware front and centre: agriculture robotics, counter-drone defence, inference chips, lunar manufacturing. AI can now operate reliably outside a screen, hardware costs have dropped enough to make deployment viable, and labour shortages have created demand that software alone cannot meet. The industries that software spent a decade promising to transform are exactly where AI is heading next.

Three different bets. The same underlying premise. AI is the foundation.

In APAC, the same conviction takes a different form

Venture capital is one kind of signal. Sovereign wealth is another.

APAC has venture capital activity in AI. But the region's VC ecosystem is a fraction of Silicon Valley's, which accounts for roughly 75% of global AI investment. What APAC brings to this story is something different: sovereign wealth funds, governments investing national reserves with no fixed fund life, no external investors demanding quarterly returns, and no forced exit timeline. Patient capital, moving on a national mandate.

The reason sovereign wealth leads here is structural. Countries across Asia Pacific cannot afford to depend on US or Chinese infrastructure to power their own digital economies. AI is now foundational enough that compute capacity, energy independence, and data sovereignty are national concerns. Private venture capital is not structured to solve that problem. Sovereign wealth is.

Singapore's GIC invested significantly in Anthropic's $13 billion Series F and co-invested $1.6 billion in AI data centre infrastructure across Asia Pacific, including a hyperscale campus in Johor with Abu Dhabi's ADIA. The motivation is strategic, the kind that answers to national interest rather than a returns committee.

The same logic is playing out across the region. Indonesia's INA partners with Granite Asia to invest over $1.2 billion in the country's technology and AI ecosystem, including a new data centre campus in Batam. Danantara names AI one of its first nine priority sectors. Malaysia's Khazanah moves capital into power grids and semiconductor firms, pricing in the reality that AI is energy-intensive and that energy independence determines who can scale it. South Korea's KIC, managing $206 billion in assets, is expanding its AI bets and deepening its Silicon Valley network through its San Francisco office. It is a sovereign fund deliberately adopting the instincts of a VC.

Read together, these bets point to one conclusion: AI infrastructure is being treated as permanent, sovereign, and strategic. The countries building this layer today are making a longer bet: a future where AI capability is as foundational as electricity, and where depending on someone else to supply it carries real national risk.

The verdict

The investors who called Google in 1999 and OpenAI in 2015 were not working with better information than everyone else. They were asking a different question: not what is happening now, but what becomes inevitable next.

That question has a clearer answer today than at any point in recent memory. In 2025, AI captured 61% of all global VC deal value. Three of the world's most influential funds reached the same thesis independently. Sovereign governments across a region are treating AI infrastructure as a national priority. When capital moves like this, from this many directions at once, it adds up to a verdict.

Not every bet placed today will pay off. That is how this has always worked, from the internet to mobile to cloud. But the debate is over. The window is open. The only remaining question is whether you are moving while it is.