Nvidia is extending its AI moat from chips into the financial sector, announcing up to $105 billion in financial support for OpenAI’s PORTS-Pike mega data center in Ohio and a $1.5 billion investment in SB Energy. CEO Jensen Huang said the project could generate approximately $600 billion in computing business opportunities for Nvidia by 2030. Nvidia has previously partnered with Wall Street financial institutions including Goldman Sachs, Apollo, Blackstone, and BlackRock to raise $500 billion, positioning GPUs as a new asset class akin to real estate. Nvidia’s most recent quarterly free cash flow reached $48.5 billion, an 18-fold surge from three years ago. The move comes as competitors like AMD and Google narrow the technology gap, with Nvidia leveraging capital deployment to sustain the AI investment boom while addressing investor concerns about balance sheet risk.
Nvidia (NVDA-US) has leveraged its massive lead in artificial intelligence to become the world’s most valuable company. However, as the generative AI boom enters its fourth year, with competitors like AMD (AMD-US) and Google (GOOGL-US) gradually narrowing the technology gap, Nvidia is now deploying another powerful asset—its deep capital reserves—to extend its AI moat from chips into the financial sector.
Last week, Nvidia struck partnerships with Wall Street financial institutions to raise $500 billion (approximately NT$15.9 trillion) in funding for purchases of Nvidia graphics processing units (GPUs). On Monday, Nvidia announced it would provide up to $105 billion (approximately NT$3.3 trillion) in support for OpenAI’s massive data center in Ohio, effectively serving as a financial backstop for infrastructure investments should the ChatGPT developer’s future operations underperform expectations.
Capital Deployment: From Chip Supplier to Infrastructure Financier
Nvidia’s strategy is to sustain the AI investment boom through various means. While demand for critical infrastructure appears insatiable, the high concentration of purchases among a handful of hyperscale cloud service providers has made it necessary for Nvidia to help more enterprises access computing resources.
Nvidia’s most recent quarterly free cash flow reached $48.5 billion (approximately NT$1.5 trillion), an 18-fold surge from three years ago, with 12 consecutive quarters of revenue growth exceeding 55%. Ram Bala, associate professor at Santa Clara University’s Leavey School of Business, said Nvidia remains dominant but is highly vigilant about losing its edge.
Cantor analysts on Monday reiterated their Buy rating on Nvidia, downplaying market concerns that the company is “buying revenue” through financial arrangements. Analysts believe the latest agreement signals that this AI investment cycle will be more durable, with Nvidia not merely engaging in circular financing but facilitating AI buildout while establishing additional competitive moats.
The data center campus in Pike County, Ohio, will be leased by OpenAI and deployed at scale with Nvidia’s AI chips. OpenAI has signed agreements to secure up to approximately 8 gigawatts (GW) of computing capacity at the campus, with the first tranche of roughly 800 megawatts (MW) expected to come online by 2028. OpenAI is not required to pay full lease costs upfront; payments begin only after the corresponding computing capacity is built and available for lease.
Nvidia said it will provide support for “lease and power payment obligations for specific portions of the project,” while emphasizing that this is not circular financing and that ultimate lease costs remain OpenAI’s responsibility.
Jensen Huang: GPUs Are a New Productive Asset
Nvidia CEO Jensen Huang noted that frontier AI labs face enormous demand for training and inference computing, but many have outgrown their balance sheets and long-term credit capacity. Even with strong demand and rapidly growing revenue, these companies may lack the ability to sign decades-long infrastructure contracts and secure investment-grade financing.
Last week, Huang signed memoranda of understanding with Wall Street financial giants including Goldman Sachs, Apollo, Blackstone (BX-US), and BlackRock (BLK-US), seeking to position GPUs as a new asset class akin to real estate, enabling third-party investors to fund the next phase of AI buildout. Huang emphasized that GPUs are now productive assets capable of generating revenue, with long useful lives, replaceability, and flexibility.
Companies seeking financing must commit to adopting Nvidia systems, and Nvidia may choose to provide up to 25% support for each loan. This move could help bring more Nvidia technology to market as Google, AMD, and Cerebras (CBRS-US) intensify competition.
Nvidia confirmed Monday that it will provide financial guarantees for OpenAI’s PORTS-Pike Technology Campus, a massive data center campus in Ohio. Huang said: “Our support is limited to specifically scoped lease and power payments, as well as clearly defined residual value commitments—not the entire cost of the campus or all obligations of the tenant.”
Nvidia also announced a $1.5 billion (approximately NT$48 billion) investment in infrastructure company SB Energy. SB Energy is an infrastructure company jointly established by OpenAI and Japan’s SoftBank Group (9984-JP), which will build, own, and operate the data center, leasing it to OpenAI under a 20-year agreement.
A $600 Billion Opportunity
Nvidia said that each generation of AI systems deployed at the PORTS-Pike campus could require approximately 1.5 million Nvidia GPUs, representing roughly $150 billion to $200 billion (approximately NT$6.4 trillion) in revenue. The data center could undergo multiple upgrade cycles over its 20-year lifespan.
Huang noted: “OpenAI’s current and planned commitments represent approximately 12 GW of Nvidia computing capacity. If Nvidia expands the PORTS-Pike collaboration from the initial 4.25 GW, the overall scale could increase to approximately 16 GW. At that scale, this opportunity represents roughly $600 billion (approximately NT$19.1 trillion) in Nvidia computing business by 2030.”
Nvidia currently possesses staggering cash generation capability. In May, the company announced an increase in its quarterly dividend from 1 cent to 25 cents per share and unveiled an $80 billion (approximately NT$2.6 trillion) share buyback program, committing to return approximately 50% of free cash flow to shareholders this year.
Beyond shareholder returns, Nvidia is actively investing in AI ecosystem companies, including model developers and new cloud service providers that purchase its chips and systems in large volumes. As of the most recent quarter, Nvidia held marketable equity securities worth $30.2 billion (approximately NT$960 billion), up from $12.9 billion (approximately NT$410 billion) a year earlier.
In February, Nvidia invested $30 billion (approximately NT$960 billion) in OpenAI, whose model training relies on Nvidia’s most advanced Vera Rubin systems.
Competitive Pressure and Market Skepticism
Google began recognizing revenue from TPU system sales in the second quarter, driving cloud business growth of 82%; AMD’s data center business revenue more than doubled, with its first rack-scale system, Helios, expected to ship later this year. Paul Meeks, head of technology research at Freedom Capital Markets, noted that intensifying competition will compress Nvidia’s previously extraordinary profit margins, prompting the company to expand its footprint rather than relying solely on GPUs.
Nvidia shares closed down 0.1% Monday at $225.01 (approximately NT$7,200), after reaching an intraday high of $227.92. Nvidia’s stock has risen 21% this year and nearly 24% over the past 12 months.
Market sources indicate that Nvidia now expects to provide guarantees of less than $120 billion (approximately NT$3.8 trillion) for the project, significantly reduced from the previously planned $250 billion (approximately NT$8 trillion). This adjustment was primarily made in response to investor concerns about excessive risk on Nvidia’s balance sheet.
However, investors supporting the AI industry argue that Nvidia is simply responding to genuine demand—the market currently faces a computing capacity shortage, not oversupply. Anthropic told investors that annualized revenue reached $65 billion (approximately NT$2.1 trillion) in July, a sixfold increase from a year earlier; OpenAI’s annualized revenue has also reached $40 billion (approximately NT$1.3 trillion). Matthew Vegari, head of research at Clearwater Analytics, believes the AI industry may eventually face overcapacity, but that day has not yet arrived.
Nvidia has long supported AI companies in expanding infrastructure through investments and guarantees, previously providing funding to AI cloud computing service provider CoreWeave and directly investing in companies like OpenAI and Anthropic. This agreement with OpenAI represents Nvidia’s largest support project to date.
By providing financial assistance for hyperscale data center projects, Nvidia can help core customers like OpenAI overcome the financing challenges required for large-scale computing capacity buildout, while also expecting to further drive demand growth for its AI chips. As Nvidia plays an increasingly critical role in global AI infrastructure development, it will also indirectly boost orders and technology development across Taiwan’s semiconductor supply chain, particularly for Taiwanese wafer manufacturers serving as primary foundry partners for advanced AI chips.
https://finance.biggo.com/news/122ef016-3122-415a-9950-5ef81691f97d
