Put the question to any investment committee and the room splits fast. Goldman Sachs and Morgan Stanley have held talks with S&P, Moody’s, and Fitch about granting investment-grade ratings to OpenAI and Anthropic soon after each company goes public, the Financial Times reported Tuesday, September 8, 2026. OpenAI posted a $20.92 billion operating loss on $13.07 billion of revenue in 2025. Anthropic expects to break even in 2028, with OpenAI targeting 2030. Rating analysts, for now, are not persuaded. They continue to describe both companies as speculative-grade and loss-making, citing limited public disclosure on finances and uncertainty over future cash generation.
The banks’ argument runs like this: a public listing would provide substantial liquidity and improve balance-sheet metrics, enabling cheaper financing for rapid and costly infrastructure investments. The goal is access to scale. They want to tap the $11.7 trillion corporate bond market post-IPO. A senior credit analyst told the Financial Times that “Wall Street is trying to minimise their overall debt impact by arguing that these two companies will soon be flush with liquidity.”
The SpaceX precedent is the template Goldman and Morgan Stanley are selling. All three major agencies placed SpaceX in investment grade in ratings actions announced June 18, 2026: S&P at BBB (stable), Moody’s at Baa1 (stable), and Fitch at BBB+ (stable). The outcome should give the OpenAI camp pause. As of early July, SpaceX bonds traded at an average credit spread of 1.62 percentage points over Treasuries, while average BBB corporates traded at 0.92 points and average BB junk bonds at 1.55 points. The label said investment grade. Credit investors priced it as junk.
OpenAI’s situation is structurally weaker than SpaceX’s was at IPO. SpaceX had Starlink’s subscriber base and a satellite business underwriting its rating case. OpenAI has revenue growing fast but a cost base growing faster, and no analogous cash-generating subsidiary. The banks are asking the agencies to underwrite a future earnings story, not a present balance sheet.
Here is where it gets interesting for Nvidia. Nvidia disclosed that in August 2026 it entered into guarantees, capped at a total of $105 billion, to provide credit support for leases tied to affiliates of SB Energy at the PORTS Technology Campus in Pike County, Ohio, on behalf of a customer that it identified as an affiliate of OpenAI Group PBC. Nvidia said the guarantees terminate upon certain events, including OpenAI achieving “a satisfactory credit rating.” The guarantees generally become effective only upon commencement of the applicable leases, with amounts increasing as phases of construction are completed, with the first phase expected in fiscal year 2029.
That termination clause is the sentence investors are underreading. The clause points to the purpose: the guarantee can lapse once OpenAI’s own credit is strong enough to support the leases without Nvidia. An investment-grade rating granted at IPO would, in one step, open the bond market to OpenAI, reduce borrowing costs, and potentially release Nvidia from what it has capped at $105 billion of credit support. Whether that makes Nvidia a beneficiary or a party deeply exposed to the agencies saying no depends entirely on how seriously Moody’s and S&P treat the banks’ lobbying.
Broadcom is running a parallel version of this problem with Anthropic. Broadcom, together with Apollo and Blackstone, launched an AI infrastructure platform with an initial $35 billion financing tranche led by Apollo, in partnership with Blackstone, to facilitate Anthropic’s capacity expansion. The structure has been described publicly as using a special-purpose vehicle that owns equipment and leases it to Anthropic, with Broadcom conditionally supporting portions of the financing if Anthropic stops paying. What is not supported on the record is the claim that S&P downgraded Broadcom because of this arrangement. Investment-grade ratings at IPO would shift much of that exposure from corporate partners onto bond investors. That is the real transaction Goldman and Morgan Stanley are pitching.
Stocks to Watch
- Nvidia (NVDA): The most direct read-through. A successful rating could help terminate the Ohio credit support; a refusal leaves the contingent exposure in place until other termination events occur, including the end of each respective lease term.
- Broadcom (AVGO): Helped launch a $35 billion-financing platform tied to Anthropic’s compute build-out and has been described as conditionally supporting portions of that financing. An Anthropic investment-grade rating at IPO would reduce reliance on partner support and potentially restore balance-sheet flexibility.
- Alphabet/Google (GOOGL): Supplies TPUs and is a key infrastructure counterparty to Anthropic. A successful IPO with investment-grade status could accelerate Anthropic’s path to self-financing and reduce counterparty dependence.
- Goldman Sachs (GS) and Morgan Stanley (MS): Both are underwriting the IPOs and lobbying for the ratings simultaneously. If agencies comply, both banks lock in the full fee chain from listing through bond issuance. If the agencies hold firm, the IPO economics weaken.
