NVIDIA Binds with SK and OpenAI in $750 Billion Gamble? AI 'Circular Financing' Alarm Rings Again

According to a report by Wall Street Horizon, NVIDIA (NVDA-US) is pushing forward a new round of AI infrastructure transactions totaling over $750 billion. While accelerating global AI computing deployment, the move has reignited market concerns over 'circular financing'.

On Friday evening, NVIDIA and South Korea's SK Group announced an AI cooperation agreement exceeding $500 billion. Simultaneously, NVIDIA is negotiating a $250 billion secured agreement with OpenAI to help the latter lease data center projects being developed by a SoftBank subsidiary in Ohio. The news triggered widespread market attention, with NVIDIA's stock falling over 5%.

Critics are primarily concerned that companies NVIDIA invests in or holds equity in are often its main chip buyers. This model may distort business decisions, create perverse incentives, and amplify industry losses if AI demand fails to meet expectations.

Gary Tan, portfolio manager at Allspring Global Investments, said: 'While NVIDIA's investments and partnerships reinforce market confidence in long-term AI infrastructure development, investor concerns over circular financing persist. Capital is increasingly being used to support future AI customers and infrastructure rollouts.'

The $500 billion bet on South Korean AI infrastructure

The scale of NVIDIA's deal with SK Group exceeds $500 billion, covering NVIDIA's procurement of memory chips from SK Hynix (SKHY-US), a subsidiary of SK Group, as well as SK Group's purchase of NVIDIA supercomputers. In an interview with Bloomberg Television, NVIDIA CEO Jensen Huang stated: 'There will be $500 billion in business between the two companies.'

Under the agreement, the two parties will jointly build over 2GW of AI data centers on the Korean Peninsula—equivalent to the energy required to power 1.5 million households. The first 'AI factory,' led by SK Telecom, is expected to begin operations next year.

For NVIDIA, this partnership holds clear strategic supply chain significance. High Bandwidth Memory (HBM) is currently in severe shortage due to the global AI data center construction boom, with SK Hynix and Samsung Electronics being the two major global suppliers. NVIDIA will assist in designing the next-generation HBM chips, ensuring a stable supply of critical components.

Additionally, on the same day, NVIDIA announced a $1 billion investment in Naver, a South Korean web and cloud services provider, to support its under-construction AI data center in Korea. The project will expand the facility's scale by over three times and is jointly developed by Naver and U.S. private equity firm Brookfield, using NVIDIA's AI computing hardware. Following the announcement, Naver's stock surged over 8% in the Seoul market.

The potential transaction between NVIDIA and OpenAI is equally massive. According to media citing insiders, NVIDIA is discussing providing up to $250 billion in guarantees to OpenAI to help it lease a $50 billion, 10-gigawatt data center hub in Ohio developed by a SoftBank subsidiary. At the same time, NVIDIA is also negotiating $350 billion in chip procurement financing for OpenAI.

Billy Leung, investment strategist at Global X Management, pointed out: 'NVIDIA providing guarantees for more OpenAI data center debt further deepens an already scrutinized supplier financing model. This is less a signal of demand and more a reflection of the financial pressure behind AI infrastructure development.'

If the guarantee is finalized, it would help alleviate lenders' concerns about continuously extending credit to a yet-to-be-profitable private company. It would also support SoftBank founder Masayoshi Son's strategic ambition to play a central role in the AI domain. Currently, SoftBank's investment commitment to OpenAI is nearing $65 billion, along with a $40 billion bridge loan—one of the largest in the Asia-Pacific region. However, SoftBank's growing bets on a company it has limited control over have made investors uneasy.

Cross-shareholding intensifies systemic risk

Criticism of NVIDIA's transaction model is not new, but the pace of such deals is accelerating. NVIDIA has taken equity stakes in AI developers and chip peers such as OpenAI and Marvell Technology, and invested in data center operators including IREN Ltd., CoreWeave Inc., and NEBIUS Group NV. Since 2026 alone, NVIDIA's announced deals of this nature total over $540 billion, excluding the potential new agreement with OpenAI.

Similar patterns are spreading across the industry. Google has backed lease payments for Anthropic across five data center locations, helping the OpenAI competitor secure loan support equivalent to $35 billion. Cross-shareholding and mutual guarantees are deepening the interdependence among AI companies, increasing the risk of systemic shocks across the sector. Meanwhile, AI companies continue to increase their debt levels to sustain capital expenditures on data centers and chip projects.

In response to accusations of circular financing, Jensen Huang has consistently rejected them. In January, when discussing NVIDIA's investment in CoreWeave, he said: 'This is just a small part of the total capital they will eventually need to raise. Calling this circular is absurd.' He emphasized that NVIDIA's investments in OpenAI, Anthropic, and others not only drive industry development but also generate returns for NVIDIA.

However, as NVIDIA becomes deeply embedded in both upstream and downstream segments of the AI ecosystem, its overlapping roles as chip seller, equity investor, and financing guarantor are becoming increasingly intertwined. Market concerns about the fragility of this model under slowing demand conditions have not dissipated.