With so much money going into AI, I keep wondering which companies it will actually help build. I am most interested in businesses bringing AI into manufacturing. Funding a model is one part of that process; getting a customer to use it in a factory takes a different kind of commitment.
At GCCEI, a South Korean startup support organization, my work included helping set up an early-stage venture fund and sourcing startups for possible trials with corporate partners. So I look at this investment boom with both the investor and the prospective customer in mind. A technology can be promising enough to fund before a factory is ready to depend on it. Who pays for the work between those two decisions?
The scale of the money makes that question worth following. Crunchbase reported $211 billion in venture funding for AI companies in 2025. Bloomberg, citing Global SWF, reported $66 billion from sovereign-owned investors into AI and digitalization. I would not add the two figures together, because their categories differ and may overlap. I care more about what each investor is prepared to finance than about producing a larger total.
Financing the first factory installation
To work through the question, I am using a hypothetical startup that wants to install AI-controlled robots in a Korean factory. Before the customer buys, the startup has to show that the robot can handle the customer's parts reliably. During that trial it pays for equipment and engineering without knowing whether a sale will follow.
From an investor's point of view, what I would examine most closely is how much of the installation can be sold again. Venture funding can carry substantial uncertainty in a young company, but the successful investments have to become valuable enough to make up for the failures. If each new factory requires another large engineering project, growing sales may not improve the business as quickly as the investor expects.
That is why I separate money for the company from money for the assets it uses. A data center can raise financing against its assets and expected customer payments, provided lenders can assess delivery and repayment. The robotics startup may still be discovering what its customers will buy. Access to infrastructure capital would not, on its own, resolve the uncertainty around its product.
I also want to leave room for what a policy investor might value. A government could support a trial because better domestic manufacturing has benefits beyond the startup's return. That does not make every sovereign fund a suitable investor, or mean its financial requirements are relaxed. I would need to read a particular fund's mandate before treating it as a source of deployment financing.
The AI Infrastructure Partnership
For a concrete example of private and state-backed investors working together, I am following the AI Infrastructure Partnership. BlackRock, Global Infrastructure Partners, Microsoft and Abu Dhabi-based MGX established it in 2024, and the Kuwait Investment Authority joined in 2025. The partnership's announcement describes investment in data centers and related power infrastructure, with a goal of mobilizing $30 billion in equity and potential investment of $100 billion including debt.
I read those figures as financing goals, not money already spent. What interests me more is that the partnership names the assets it intends to finance. Investors from different kinds of institutions, with different objectives, have found a potential common project.
For the robotics startup in my example, though, a data center would solve only part of the problem. Computing power cannot establish that the robot handles the customer's parts or that the installation earns back its cost. I would still need a separate answer to who funds the trial and who bears the loss if it fails.
Working with manufacturing customers
The cross-border opportunity I am interested in is the work of making that first installation possible. Introducing a startup to a factory is useful, but I would judge a potential partner by how well it helps the two sides agree on what the trial must establish and who pays for it.
Even after a successful demonstration, the factory has practical questions about production. How often does the robot stop, and who gets it running again? Will the improvement justify the cost of integration? The investor has a related concern: once the startup answers those questions at one site, does it have a product it can install elsewhere with less work?
This is where I see a possible meeting point between private and policy investment. If repeated deployments show that the product improves manufacturing and gets easier to install, a private investor can see a business with room to grow, and a policy investor can see a benefit to domestic production. A partner who understands both decisions could help finance the work that gets the company there.
My working thesis is that some of the investment in AI infrastructure will be followed by more joint investment in industrial applications. I treat that as a hypothesis about adoption, not a timetable set by today's funding totals. The evidence I am watching for is a supplier delivering the same product across several factories with less custom work each time. That would change my view of the business being financed far more than another large announcement would.