Add preferred source
The Economist reports that NVIDIA has become the “central bank of AI,” dominating capital flows across the AI industry. The company has invested over $70 billion in startups over the past three years and provided $300 billion in financial support to customers. As its largest customers—Amazon, Google, Meta, and Microsoft—push forward with developing their own custom semiconductors, NVIDIA is cultivating an independent AI ecosystem through strategic investments. Its involvement extends from direct investments to complex financial schemes, including guaranteeing power purchase agreements for a data center in Ohio. Experts warn that the proliferation of speculative projects and interconnected financing structures could amplify losses if the AI bubble bursts.
Key Elements
NVIDIA, which holds an overwhelming share of the AI semiconductor market, has evolved beyond being a mere chip supplier into an entity that shapes capital flows across the entire industry. The Economist has described the company as the “central bank of AI,” detailing the scope of its investment and financial support activities.
Over the past three years, NVIDIA has invested more than $70 billion (approximately ¥10.9 trillion) in startups and provided $300 billion (approximately ¥46.8 trillion) in financial support to customers. In 2025, the company made roughly 90 investments—nearly double the number in 2024. As of September 2026, it had already agreed to 60 additional investments.
In October 2025, NVIDIA became the first company in the world to reach a market capitalization of $5 trillion (approximately ¥780.2 trillion). Riding the AI boom, the company has come to play a central role in the industry on both the semiconductor supply and capital provision fronts.
The risk of “defection” by Big Tech
Behind NVIDIA’s aggressive financial strategy lies a structural problem: its largest customers are increasingly becoming competitors. Amazon, Google, Meta, and Microsoft account for roughly half of NVIDIA’s revenue, yet all four are developing their own custom AI semiconductors.
In-house chips cost roughly one-fifth to one-third the price of NVIDIA products and can be optimized for each company’s own software, giving them a performance edge in certain workloads. Microsoft is reportedly preparing to unveil its next-generation AI chip, “Maia 300,” as early as September 2026, and is in talks with TSMC to manufacture more than 300,000 units by 2027.
Moreover, Big Tech companies are moving beyond internal use to external sales. Google has signed a deal to deploy its AI-specialized processor, the “TPU,” at scale for Anthropic. Amazon has also begun discussions with customers about selling its custom “Trainium” chips externally. Bloomberg Intelligence projects that custom semiconductors’ share of the AI processor market will grow from roughly 40% in 2026 to approximately 50% by the end of 2030.
A “financial intermediary” bridging the credit gap
The core of NVIDIA’s investment strategy is to bridge the funding cost gap between Big Tech and emerging AI companies. Highly rated companies like Alphabet can raise capital at low interest rates, while startups like CoreWeave face borrowing costs roughly double that level.
NVIDIA fills this gap by investing in startups that may become future customers, as well as companies that indirectly generate demand for its semiconductors. Some investments aim to promote open-weight AI models, cultivating an independent ecosystem to counter the closed models offered by Anthropic, OpenAI, and Google.
A prime example is the acquisition of Hugging Face. The purchase price was $12.9303 billion (approximately ¥2 trillion), with the stated goal of increasing the number of AI products and companies independent of Big Tech, thereby stimulating demand for NVIDIA semiconductors.
Increasingly complex capital circulation mechanisms
NVIDIA’s involvement extends well beyond direct investment. A massive data center in Ohio that the company supports is owned by SB Energy, a SoftBank subsidiary, with OpenAI as the tenant. NVIDIA has guaranteed not only the land and building lease agreements but also the power purchase agreements. In return, the project is expected to use 1.5 million NVIDIA processors, with multiple upgrades planned over its operational life.
Even Anthropic cannot escape this network. The company has signed a $35 billion (approximately ¥5.5 trillion) contract to lease cloud computing capacity from Lambda, a company in which NVIDIA has invested. Lambda, in turn, uses facilities developed by Hut 8—and since NVIDIA leases all of Hut 8’s computing capacity, The Economist notes that much of that computing power may be subleased back to Lambda by NVIDIA.
NVIDIA is also exploring ways to attract outside capital. The model involves investors providing funds to an independent investment vehicle, which then purchases NVIDIA hardware to build infrastructure and sells computing capacity. NVIDIA neither commits cash upfront nor assumes debt in these arrangements. However, in some deals, if the hardware’s value falls below expectations after a certain period, NVIDIA provides a “residual value guarantee” covering up to one-quarter of the total transaction value to make up the shortfall.
Competitor AMD is making similar moves. In exchange for securing large-scale purchase agreements from OpenAI and Meta for its semiconductors, AMD has proposed selling both companies substantial equity stakes.
Warnings about mounting risk
Regarding this expansion of capital circulation, The Economist warns that “there is a danger that speculative projects that would otherwise struggle to secure financing end up getting built.”
No other company in the AI market is as deeply involved in financing related businesses as NVIDIA. The magazine cautions that while NVIDIA’s investments have performed exceptionally well so far, the longer the boom continues, the greater the losses the company may have to absorb if the AI bubble bursts.
The Bank for International Settlements (BIS) has expressed similar concerns. Frank Smets, head of economic analysis at BIS, said, “What concerns us most is the rapid increase in debt and leverage,” noting that a substantial portion of AI-related financing is off-balance-sheet and that transaction structures are interconnected. Pablo Hernández de Cos, General Manager of the BIS, also observed that AI investment funding increasingly relies on debt and private loans rather than corporate profits, describing the financial structure as “opaque and interconnected.”
The financing structure of AI investment creates a circular flow in which semiconductor companies invest in AI companies, and AI companies purchase those semiconductors. Hyperscalers provide guarantees to data center companies, which use those guarantees to borrow from private lenders and provide computing reinterconnected.”
NVIDIA’s investment strategy accelerates the growth of the entire AI ecosystem while simultaneously contributing to the accumulation of risks that could ripple through the broader financial system. The more the company expands its role as the “central bank of AI,” the greater its influence—and its responsibility—becomes.
Once added, BigGo Finance appears first in Google Search Top Stories, so you get the broadest, most up-to-the-minute, and most comprehensive global financial news first.
