Nvidia joins $750 billion bet by linking SK Group and OpenAI! AI "circular financing" warning triggered again

Nvidia is advancing a new round of AI infrastructure deals totaling more than $750 billion. While this accelerates global AI computing power deployment, it has also reignited concerns over "circular financing" from the market.

Nvidia and Korea's SK Group announced a more than $500 billion AI partnership last Friday evening. Meanwhile, the company is negotiating with OpenAI on a guarantee deal as large as $250 billion to help OpenAI rent a data center project being developed in Ohio by a subsidiary of SoftBank. The disclosure of these deals has garnered widespread attention in the market.

The core concern among critics is: The companies in which Nvidia invests or holds shares are often the major buyers of its chips, a model that could distort business decisions, create adverse incentives, and amplify industry losses if AI demand does not meet expectations.

Gary Tan, a portfolio manager at Allspring Global Investments, commented, "Although Nvidia's investments and partnerships reinforce market confidence in long-term AI buildout, investors remain concerned about circular financing. Increasing capital is being used to support future AI clients and infrastructure rollout."

The transaction between Nvidia and SK Group exceeds $500 billion, covering Nvidia's purchase of memory chips from SK's Hynix and SK Group's purchase of Nvidia's supercomputers. Nvidia CEO Jensen Huang said in a Bloomberg Television interview, "Business dealings between the two companies will amount to $500 billion."

Under the agreement, both sides will jointly build AI data centers exceeding 2 gigawatts on the Korean Peninsula—a scale equivalent to the total energy required to power 1.5 million households. The first "AI factory" led by SK Telecom is expected to go into operation next year.

For Nvidia, this partnership has a clear supply chain strategy. High bandwidth memory (HBM) chips are in extreme shortage globally due to the AI data center boom, with SK Hynix and Samsung Electronics as the world’s two main suppliers. Nvidia will help Hynix design the next generation of HBM chips to ensure its critical component supply.

Additionally, Nvidia announced the same day that it would invest $1 billion in Korean internet and cloud service provider Naver Corp. to support its ongoing data center development in Korea. The project will more than triple the facility's footprint, with Naver and US private equity firm Brookfield Corp. co-developing it, utilizing Nvidia's AI compute hardware. After the news, Naver’s share price in Seoul surged more than 8% in one day.

Nvidia's potential deal with OpenAI is equally massive. According to the media citing sources, Nvidia is in discussions to provide OpenAI with a guarantee of up to $250 billion to help it rent a $50 billion, 10-gigawatt data center hub being developed in Ohio by a SoftBank subsidiary. Simultaneously, Nvidia is negotiating $350 billion in chip procurement financing for OpenAI.

Billy Leung, Investment Strategist at Global X Management, noted:

"Nvidia guaranteeing more OpenAI data center debt further deepens the already scrutinized vendor financing model. This may reflect less a signal of demand and more a snapshot of the funding pressure in AI buildout."

If the guarantees materialize, they would help alleviate creditor concerns over providing ongoing credit to this still-unprofitable, non-public company, and support SoftBank founder Masayoshi Son’s ambitions to play a pivotal role in AI. Currently, SoftBank’s investment commitment to OpenAI is near $65 billion and includes a $40 billion bridge loan—one of Asia-Pacific's largest such facilities. However, SoftBank's increasing exposure to a company over which it has limited control has made investors uneasy.

Questions over Nvidia’s dealmaking model are not new, but the pace of such agreements is quickening. Nvidia already holds shares in OpenAI, Marvell Technology, and other AI developers and chip peers, and has invested in data center operators like IREN Ltd., CoreWeave Inc., and Nebius Group NV. Since 2026 alone, Nvidia has announced such deals totaling over $540 billion, not counting the potential new agreement with OpenAI.

Similar phenomena are spreading across the industry. Google has underwritten Anthropic’s lease payments at five data center locations, helping this OpenAI competitor secure up to $35 billion in loan support. Multiple cross-shareholdings and mutual guarantees are deepening ties between AI companies, exposing the entire sector to potential systemic shocks. At the same time, AI firms’ borrowing continues to climb to support data center and chip project capital expenditure.

Jensen Huang has consistently refuted accusations of circular financing. Commenting on the CoreWeave investment in January, he stated, "This is just a small fraction of the total capital they need to raise. To call this circular—that’s just absurd." He also emphasized that Nvidia’s investments in companies like OpenAI and Anthropic drive industry growth and generate returns for Nvidia as well.

However, as Nvidia becomes ever more entangled in the AI ecosystem, acting as a chip vendor, equity investor, and financing guarantor, the market’s concerns over the fragility of this model—especially in a scenario of slowing demand—have yet to dissipate.