TL;DR — Key Takeaways

  • SoftBank secured an $11.87 billion two-year loan facility to support its expanding investment in OpenAI, surpassing its original $10 billion target.
  • The company is expected to invest nearly $65 billion in OpenAI by October, underscoring Masayoshi Son’s aggressive bet on generative AI.
  • SoftBank plans to repay the remaining $25.9 billion on a separate $40 billion bridge facility while exploring a potential $10 billion to $20 billion high-yield bond sale.

SoftBank Group Corp. has secured an $11.87 billion two-year loan facility to finance its deepening investment in OpenAI, exceeding its initial $10 billion target, according to a Bloomberg report, citing people familiar with the matter.

The financing agreement, finalized last week with commitments from roughly 20 financial institutions, underscores founder and CEO Masayoshi Son’s aggressive push into generative artificial intelligence (AI).

SoftBank is slated to invest close to $65 billion in the ChatGPT creator by October, having already raised approximately $37 billion this year through offshore and domestic debt markets, credit facilities, and bond sales.

The new credit line comes as SoftBank reshuffles its balance sheet. The conglomerate announced plans to repay the remaining $25.9 billion balance on a separate $40 billion bridge facility by Sept. 15, well ahead of its original March maturity date. That initial facility was deployed to fund SoftBank’s phased commitments, including a previously disclosed $22.5 billion stake in the AI pioneer.

Despite the successful debt syndicate, investor sentiment took a hit in Tokyo trading. SoftBank shares plummeted by as much as 13% on Monday, marking their sharpest intra-day decline since July 17.

“What stands out to me is what this says about the AI stack. Building frontier AI now requires enormous investment across models, chips, data centers, power, networking, and the software needed to operate it all,” said Stephanie Walter, practice leader for AI Stack & Enterprise Application Development at HyperFRAME Research.

“SoftBank is betting that OpenAI will continue to sit at the center of that stack. The unknown is whether OpenAI can turn its early lead into a business that justifies this level of investment. The financing provides the resources to build. What matters now is what gets built, how widely enterprises use it, and whether the economics work over time,” Walter said.

Market scrutiny around the tech giant’s leverage has mounted since March, when S&P Global Ratings revised SoftBank’s credit outlook to negative.

To manage its capital structure and maintain liquidity, SoftBank executives are set to meet with investors in New York this week to gauge demand for a high-yield U.S. dollar bond sale. Sources indicate the firm is considering raising between $10 billion and $20 billion through the junk-bond offering.

The massive capital injection arrives during a delicate transition period for OpenAI. The AI sector faces growing headwind calls from industry leaders urging a slower pace of development to manage potential technology risks.

Addressing these challenges, OpenAI CEO Sam Altman confirmed in a recent interview with Fortune that the company has shelved plans for an initial public offering this year while it prioritizes safety-related concerns and corporate governance.

SoftBank’s substantial debt expansion highlights the escalating financial arms race in generative AI, as major tech conglomerates leverage traditional credit markets to stake early claims in high-valuation AI ventures.

Frequently Asked Questions

How much did SoftBank borrow to finance its OpenAI investment?
SoftBank secured an $11.87 billion two-year loan facility, exceeding its initial $10 billion fundraising target.
How much does SoftBank plan to invest in OpenAI?
SoftBank is expected to invest close to $65 billion in OpenAI by October.
Why is SoftBank raising so much capital?
The company is financing large commitments to OpenAI and positioning itself to benefit from growth across the generative AI ecosystem, including models, infrastructure and computing.