NVIDIA is becoming the central bank of the AI era

Of course, NVIDIA can’t print money or set the federal funds rate. But it is influencing another, even more important form of capital allocation—the scarcest capital in the AI era.
The entire AI industry is entering a stage of financialization (Financialization). Capital is becoming part of AI infrastructure, not just a tool for fundraising. In this system, NVIDIA is increasingly like the central bank of the AI era—not creating money, but to a certain extent determining where capital flows.
NVIDIA is the most active driver of this trend. With nvda leading the way, other GPU makers, cloud providers, NeoCloud, packaging firms, HBM suppliers, banks, Private Credit funds, infrastructure funds, insurance capital, and sovereign wealth funds are jointly building a new capital system. The AI industry is gradually evolving from a semiconductor supply chain into a system of infrastructure assets.
The first step is a financialized supply chain. Long-term procurement agreements (LTA), advance payments, Take-or-Pay agreements, and supply-chain financing are increasing continuously. In essence, they all lock in future demand in advance, giving suppliers the confidence to expand capacity. For companies like TSMC, Amkor, SK hynix, Micron, and others, financing capability is becoming as important as manufacturing capability. Supply-chain investment is relying more on long-term orders rather than short-term market judgments.
The second step is a financialized customer base. NVIDIA is no longer just selling GPUs. It is helping customers obtain financing. By investing in customers such as neocloud, it makes it easier for them to secure bank loans and Private Credit, and easier to achieve asset securitization (GPU lease contracts, power PPAs). This releases procurement volumes from future years into today’s ordering. Future revenue is converted into current orders early on, and demand volatility is absorbed by the financial system instead of being fully passed through to the supply chain.
Financial instruments also change the demand curve. In the traditional semiconductor industry, the biggest risk comes from demand cycles. When demand is strong, the supply is entirely sold out; when demand falls, orders are canceled quickly, causing violent fluctuations across the entire supply chain. After financialization, even if customers don’t have enough cash in the short term, they can continue buying GPUs. Future demand is already “paid out” in advance, smoothing the capital expenditure rhythm across the supply chain, and making suppliers more willing to continuously expand capacity.
There is also a capital loop. Financing is used to buy GPUs. Renting GPUs generates cash flow. That cash flow supports new financing, and the new financing continues to buy GPUs, forming a self-reinforcing capital expansion flywheel. GPUs start to take on the attributes of infrastructure assets—not just electronic equipment, but assets that generate long-term cash flows, can be used to secure financing, and can be securitized. GPUs are evolving from a compute resource into a financial asset.
That’s also why NVIDIA keeps locking in HBM, advanced packaging, and supply-chain capacity.
If capital becomes one of the bottlenecks of the future,
then solving customers’ financing problems is itself the most effective way to expand GPU sales.
Helping suppliers solve their financing problems is itself the most effective way to expand capacity.
Not only nvda— the entire industry has already started evolving in this direction. Microsoft, Google, Meta, and Amazon continuously expand capital expenditure (CapEx). The orders, advance payments, and cooperation agreements they provide to customers or suppliers can similarly help upstream and downstream companies in the industry access funding more easily. NeoCloud heavily uses debt financing; HBM makers keep expanding capacity; advanced packaging companies sign multi-year agreements; Private Credit funds, infrastructure funds, and sovereign wealth funds begin investing in AI data centers. The AI industry supply chain is shifting from traditional manufacturing financing toward an infrastructure financing model.
In the future, Hyperscalers may spend tens of billions of dollars in CapEx every year. If a large portion relies on debt financing, they will continue to absorb long-term capital. New funds won’t just go to GPUs—they will also flow into data centers, power, fiber, cooling systems, HBM, and advanced packaging. The AI industry will become the world’s largest capital-absorbing industry (possibly the largest of all).
The expansion of capital demand and its impact on financial markets cannot be ignored. With limited long-term capital supply, increasing AI financing implies that the price of long-term funds may remain at relatively high levels. Even if short-term policy interest rates fall, long-term financing rates may still stay elevated due to capital demand. At the same time, capital allocation becomes more segmented: the financing cost for AI companies with stable cash flows and clear growth may continue to decline, while financing costs for traditional industries may rise relatively, and credit spreads may widen.
This model has clear similarities to real estate. Both are capital-intensive industries. Both depend on long-term financing. Both can generate stable cash flows. Both have asset collateral and securitization capabilities. And both attract long-term allocations from pension funds, insurance capital, and sovereign wealth funds.
But rather than saying AI is like real estate, it’s more accurate to say that AI infrastructure is closer to railway, power grids, and highway construction. Capital first builds infrastructure. Infrastructure improves production efficiency. New cash flows support further financing, which then drives the next round of construction. Railways once reshaped capital markets, and AI infrastructure may go through a similar process.
But risk and returns always coexist. If GPUs and data centers continue to create returns above the cost of capital, capital will keep flowing in and the industry will continue to expand. If, in the future, GPU lease yields fall below the cost of financing, the capital loop will slow down rapidly. GPU procurement, HBM demand, and advanced packaging investments will all cool at the same time, and the entire supply chain will enter a new adjustment cycle. Anyone who has experienced a real estate de-inventory cycle should know how painful such adjustments can be.
The most important variable for the future AI industry is no longer just technological progress, compute demand, or chip performance—it also includes whether global long-term capital can continue to support the expansion of AI infrastructure. In that sense, NVIDIA is playing an unprecedented new role. It is increasingly like the central bank of the AI era: it doesn’t create money, but it determines what assets capital flows toward; it doesn’t set interest rates, but it is reshaping the direction of global capital allocation.