Lobbying Fitch, Moody's and S&P: OpenAI and Anthropic Chase Investment-Grade Ratings for an $11.7 Trillion Prize

The most consequential AI finance story of the week isn’t a model release or a compute deal — it’s a lobbying campaign aimed at three offices that most AI engineers have never thought about: Fitch, Moody’s, and S&P Global.

According to a Financial Times report published on September 8, 2026, investment bankers and financial advisors hired by OpenAI and Anthropic are actively lobbying the three major rating agencies to grant the two companies investment-grade credit ratings as soon as possible after their IPOs. Morgan Stanley and Goldman Sachs have spent recent weeks engaging with the agencies on behalf of both companies. The core argument, per one senior credit analyst quoted by the FT: Wall Street is trying to minimize concerns about overall debt impact “by emphasizing that these two companies will soon have ample liquidity.”

The prize is enormous. An investment-grade rating is the entry ticket to the roughly $11.7 trillion U.S. corporate bond market at materially lower borrowing costs — capital the two companies need to fund the hundreds of billions of dollars in AI infrastructure they have committed to build.

Why this is happening now

Both companies filed confidential S-1 registration statements with the SEC in June 2026 — Anthropic on June 1, OpenAI a week later on June 8 — and both are eyeing valuations north of $850 billion, with upper-end talk stretching past $1 trillion. Anthropic in particular has accelerated: people familiar with the matter say the company could publicly file its prospectus as early as next week, launch a roadshow by the end of September, and begin trading in New York in late September or early October. Six Anthropic investors have projected a listing valuation of $2 trillion or higher — above SpaceX’s $1.78 trillion at its June listing, which would make it the largest IPO in history.

On underwriting, Morgan Stanley is reportedly the frontrunner for the coveted “lead left” bookrunner role on Anthropic’s IPO and has begun discussing pricing with potential investors. Goldman Sachs is expected to serve as stabilization agent. JPMorgan, Citigroup, and Barclays are also in line for significant roles — all three have previously provided debt financing to Anthropic.

The debt machinery already built around the companies explains the urgency. Anthropic closed a $65 billion Series H round in late May at a $965 billion post-money valuation. Apollo Global Management and Blackstone have finalized a structured notes package worth roughly $35–36 billion to finance AI chip procurement, which received mid-investment-grade ratings with Broadcom providing backing. A consortium of banks including Morgan Stanley, Goldman Sachs, JPMorgan, and Citigroup is assembling a $15 billion pre-IPO revolving credit facility. That is on the order of $50 billion in debt financing for Anthropic alone — before either company has a public rating of its own.

What investment grade actually unlocks

Most institutional investors — pension funds, insurance companies, sovereign wealth funds — operate under mandates that restrict them to investment-grade securities. Without the rating, OpenAI and Anthropic can only reach that capital indirectly, through structures wrapped in partner credit: the Broadcom-backed notes, the Nvidia guarantees, the bank facilities.

With it, several things happen at once:

The Nvidia guarantee unwinds. Regulatory filings show Nvidia has provided $105 billion in credit support for OpenAI’s massive Ohio data center campus — a 4.25-gigawatt, 20-year lease developed by SB Energy, a SoftBank subsidiary in which Nvidia has separately made a $1.5 billion equity investment. Critically, the arrangement terminates once OpenAI obtains a “satisfactory credit rating” — and not just at listing, but at any point during the 20-year lease term. An upgrade would shift that contingent liability off Nvidia’s balance sheet entirely.

Oracle gets breathing room. Oracle has borrowed heavily to fulfill its $300 billion data center construction plan for OpenAI. In July, S&P downgraded Oracle’s long-term rating from BBB to BBB- — one notch above speculative grade — explicitly citing OpenAI as a “key credit risk” in its rationale. If OpenAI can independently secure a higher rating, Oracle’s refinancing pressure eases. Alphabet and Broadcom, which have provided tens of billions in credit support for Anthropic’s chip usage, face similar dynamics. Broadcom CEO Hock Tan said last week that after Anthropic’s listing its “investment credit will change,” describing both companies as “growing into hyperscalers in their own right.”

The buyer base expands. Institutional mandates that bar speculative-grade paper suddenly open. For companies burning cash at frontier-lab scale, that is a structurally cheaper and more durable funding source than venture capital, bank credit, and Big Tech guarantees.

The agencies aren’t sold

The lobbying campaign has a credibility problem: neither company is profitable, and neither has demonstrated the ability to consistently generate positive free cash flow. Analysts note both companies disclose limited financial information and lean on optimistic metrics like annualized recurring revenue, leaving actual profitability insufficiently verified. One credit analyst was blunt: “We currently still view OpenAI and Anthropic as sitting deep in speculative-grade territory… they are still losing money.” The FT notes OpenAI’s most recent disclosed annual net loss approached $42 billion on 2025 figures.

Revenue trajectory is the strongest card in the banks’ deck. Anthropic disclosed to shareholders that July revenue annualized to $65 billion, up from $47 billion in May — though below the $80 billion-plus some investors had expected. Jordan Chalfin, head of technology at CreditSights, argues that if Anthropic raises roughly $100 billion through its IPO, its rapidly growing revenue could help it earn an investment-grade rating.

The SpaceX precedent cuts both ways. In June, SpaceX received investment-grade ratings from all three agencies immediately after listing — Moody’s Baa1, Fitch BBB+, S&P BBB, all stable — and issued $25 billion in bonds days later, upsized from $20 billion on an order book that approached $85 billion across five tranches (2031–2056, coupons 5.35%–6.65%). But those bonds subsequently declined in price, and the stock briefly fell below its $135 IPO price before recovering by mid-August. Note the spread: even for a company with mature launch and Starlink businesses, the three agencies’ judgments diverged by two full notches.

The buyer problem nobody is lobbying away

Even if the ratings arrive, the buyers may not. The long-term capital that investment grade unlocks is concentrated in markets that are currently moving the other way.

In Taiwan, whose life insurance industry holds roughly NT$38.69 trillion (about $1.23 trillion) in total assets with some NT$21.4 trillion (roughly $679 billion) overseas, regulations amended as recently as September 4, 2026 require foreign corporate bonds to be rated BBB+ or equivalent for baseline insurer eligibility — with only insurers meeting specific capital conditions able to buy BBB, BBB-, or even BB+ paper, capped at 10% of policyholder equity per issuer. A BBB- rating — where Oracle sits today — falls below the baseline. Japanese investors, who hold roughly $2.4 trillion in foreign bonds, net sold ¥3 trillion (about $18.7 billion) of foreign bonds by August 22 as 10-year JGB yields broke 3% for the first time in 30 years, and a JPMorgan Asset Management survey of 82 Japanese corporate pension funds released September 2 showed the highest propensity to increase domestic bond holdings since the survey began in 2008.

The direction of the whole chain is what makes this story worth following: partner guarantees underpinned the compute buildout; a rating upgrade would dismantle those guarantees and shift the weight onto the companies’ own credit. Wall Street hopes to convert mega-IPOs into rating upgrades. The rating agencies must decide whether fast-growing, deeply unprofitable AI companies can genuinely make the jump from equity financing and partner credit to standing on their own debt. An IPO solves the liquidity problem. It does not automatically solve the profitability one.