SoftBank's $11B Junk Bond Sale, OpenAI AGI Claim, NYT Trial

SoftBank is tapping global debt markets for roughly $11 billion to finance its next OpenAI investment tranche, a move that would rank among the largest high-yield bond sales by a single issuer on record. Elsewhere, OpenAI’s copyright battle with The New York Times just got more uncomfortable, and the company’s AGI declaration continues to draw skepticism.

Key points

• SoftBank launched a bond sale of approximately $11.15 billion, with Bloomberg reporting the transaction would rank among the largest junk bond sales by a single company, excluding distressed debt exchanges.
• Newly unredacted court filings in the NYT v. OpenAI and Microsoft case describe extensive copying of news content, alleged efforts to bypass paywalls, and internal concerns about AI products undermining publishers.
• OpenAI has publicly claimed that AGI has arrived, pointing to its latest model’s ability to perform across a wide range of intellectual and professional tasks, though the assertion is contested.
• Anthropic and OpenAI explored a binding mutual agreement to stress-test each other’s AI models, according to a new report.
• A former OpenAI researcher’s AI-focused hedge fund, Situational Awareness, nearly collapsed after a sharp AI and chip stock sell-off, drawing regulatory scrutiny from the Fed and Bank of England toward its prime brokers.

How does SoftBank fund an $11 billion bet on OpenAI?

With debt, apparently. SoftBank has launched a bond sale worth approximately $11.15 billion to finance the next stage of its OpenAI commitment, turning to global debt markets rather than equity or internal cash reserves. Bloomberg characterized the transaction as one of the largest junk bond sales by a single company, excluding distressed debt exchanges, a framing that underlines just how much leverage SoftBank is willing to absorb to maintain its position in OpenAI’s capital structure.

For investors tracking OpenAI’s path to a potential public offering, the structure of this deal matters. SoftBank is effectively passing a portion of its OpenAI exposure risk onto bond buyers. If OpenAI’s valuation stalls or its revenue growth disappoints, SoftBank still owes those bondholders. The size of the transaction signals conviction, but the use of high-yield debt rather than, say, Vision Fund equity capital also signals that SoftBank is stretching to get there.

What do the NYT court filings actually reveal?

The copyright case between The New York Times and OpenAI and Microsoft moved into sharper focus this week as newly unredacted court filings described extensive copying of news content, alleged attempts to access paywalled articles, and internal concerns that AI products might undermine the very publishers whose journalism was used to train them. The last detail is particularly damaging from a narrative standpoint: it suggests employees inside these companies understood the tension between their commercial interests and publishers’ economic survival.

The practical legal stakes here are significant for OpenAI investors. An adverse ruling or a costly settlement could set precedent for how OpenAI (and every other major AI lab) licenses training data going forward, potentially adding substantial ongoing costs to a business model that has yet to demonstrate sustainable profitability at scale.

Is AGI actually here?

OpenAI has declared that AGI has arrived, grounding the argument in its latest model’s capacity to operate across a broad range of intellectual and professional tasks rather than any single benchmark. The framing is deliberate: AGI defined by breadth, not superhuman performance on one measure.

The skepticism is equally deliberate. Critics point out that OpenAI has a commercial incentive to define AGI in a way its current products can satisfy. The company’s governance documents and investor agreements have historically tied specific rights and obligations to an AGI threshold, which makes the definition anything but academic. Until there is an agreed external standard, the claim functions more as a marketing and contractual positioning move than a scientific milestone.

A hedge fund footnote with regulatory consequences

A peripheral but notable data point: Situational Awareness, an AI-focused fund run by former OpenAI researcher Leopold Aschenbrenner, came close to collapse after a sharp sell-off in AI and chip stocks. The fund was forced to sell most of its public equities portfolio to Citadel Securities, contributing to Jane Street taking a reported $15 billion hit. The Federal Reserve and Bank of England are now asking global banks about their exposure to large trading firms in the aftermath. The SEC has also subpoenaed Wall Street banks including Goldman Sachs, JPMorgan, Citigroup, and Bank of America regarding the fund’s trading activity and use of leverage. This is not a direct OpenAI story, but it illustrates how broadly the AI investment thesis has spread into leveraged financial structures, and how a sector-wide drawdown can cascade.

Finally, Anthropic and OpenAI explored a binding deal for mutual AI model stress-testing, per a new report. Details on whether any agreement was reached are thin. If formalized, such an arrangement would be an unusual form of competitor cooperation on safety evaluation, worth watching for further developments.

This update is produced independently and is not affiliated with OpenAI. Nothing here constitutes investment advice.