How much capital would two giant AI IPOs require?

Even a small float from OpenAI and Anthropic could approach a year’s worth of global IPO supply. Company valuation, investor payments and cash raised by the issuer are three separate quantities. A $175 billion base-offering scenario shows the distinction and the additional conditions required for issuance pressure to become a credit problem.

A new choice competes for existing risk budgets

Listings by OpenAI and Anthropic would give growth investors direct ownership of model developers and a new benchmark for existing cloud, chip and data-center holdings. The first adjustments would likely be offer-price concessions, postponed deals and portfolio sales. A broader credit shock requires an additional transmission mechanism: impaired refinancing or collateral pressure at leveraged holders.

Nor should concurrent offerings be treated as a settled timetable. In a September 12 report, Axios journalists Ben Berkowitz and Donica Phifer reported that Altman said OpenAI would not list in 2026. Axios report This research has not obtained final offer prices, share counts, or primary-versus-secondary splits for either company. Valuation targets reported in the media are not completed underwriting prices, much less evidence of fully subscribed offerings. The analysis therefore considers staggered issuance as well as a concentrated-issuance stress case.

The implications for markets would nevertheless be substantial. Public-market investors have often gained indirect exposure to model demand through cloud providers, chips, and data centers. Direct ownership opportunities could reduce the scarcity premium attached to those proxies. Conversely, if the new listings are priced at difficult-to-justify valuations, suppliers with more stable earnings could become relatively more attractive. Capital will respond to price and cash flow, rather than move uniformly toward the best-known newcomers. Whether underwriting orders represent additional risk budgets matters more than the headline subscription multiple.

Wealth losses must also be distinguished from funding requirements. Assume a combined equity valuation of $3.5 trillion. A 20% post-listing decline would erase $700 billion of paper value; investors would not have paid another $700 billion. The decline could affect the economy through consumption, asset allocation, and collateral values, but the strength of those effects would depend on who owns the shares, whether they borrowed to buy them, and whether they must sell. An unleveraged long-term shareholder absorbing a price decline faces a different financial consequence from an account subject to daily margin calls.

Who receives the subscription proceeds?

Subscription demand differs from cash received by the issuer. Primary proceeds, after fees, finance the company; secondary proceeds go to selling shareholders. Both require payment from buyers, but only the first directly expands the construction budget. The SEC’s investor guidance distinguishes the two and notes the potential increase in saleable shares when lockups expire.SEC investor bulletin

A greenshoe is an over-allotment arrangement. Its terms determine whether the additional shares come from the company or existing shareholders. The corresponding additional proceeds arise only when the option is exercised and the shares are delivered; an underwriter's open-market purchases to cover a short position must not be counted as another round of financing. This article assumes a greenshoe equal to 15% of the base offering, rather than treating that percentage as a disclosed term for either company. The distinction between an over-allotment option and an underwriting short position is addressed in the SEC staff FAQ.

A lockup expiration removes a selling restriction. It neither means that all eligible shares will be sold nor raises capital for the company. Index inclusion creates a separate round of secondary-market demand. In June 2026, S&P Dow Jones Indices retained earnings, float, and IPO seasoning requirements for the S&P 500 and related indices, while some total-market indices have separate fast-track arrangements. S&P's official decision It is therefore inappropriate to count all index-linked capital as demand for the initial offering.

Suppose $1 trillion tracks an index and a new stock ultimately receives a 0.5% weight. The target holding for fully replicating portfolios would be $5 billion. This is an illustrative assumption; actual incremental purchases would also have to subtract existing holdings. Funds could sell other constituents to finance those purchases, so new buying is not equivalent to new savings. If a lockup expiration and index inclusion occur close together, supply and demand may partly offset. If they are separated, the timing of price pressure will differ.

After subscription, money principally moves from investors' accounts into those of the company or selling shareholders; it does not simply disappear. A company may deposit it, buy short-term securities, or pay for equipment. Sellers may reinvest, pay taxes, or repay debt. Each choice changes subsequent cash flows. Repayment of bank loans can also contract deposit money, so the observation that cash has changed hands does not imply that liquidity is unchanged. Investors' willingness to commit capital to equity risk and the banking system's capacity to make payments are separate questions.

Sizing a $175 billion issuance scenario

Four assumptions determine the issuance calculation: post-offer equity value, the fraction sold, the primary share and the overallotment option. The table uses equity value after the base offer and before any greenshoe, with shares sold measured against the same denominator. The scenarios illustrate supply sizes; they are not company guidance, deal forecasts or probabilities.

All amounts are before fees. The formula is valuation × percentage sold, followed by multiplication by the primary or secondary share. In the base case, OpenAI's $75 billion offering would provide $52.5 billion to the company and $22.5 billion to sellers; Anthropic's $100 billion would split into $70 billion and $30 billion. Of the combined $175 billion subscription requirement, only $122.5 billion would reach the companies. If both greenshoes were fully exercised with the same primary-secondary mix, the total would rise to $201.25 billion: $140.875 billion for the companies and $60.375 billion for sellers.

Price adjustment has two distinct consequences. With the share count unchanged, an offer price 20% below the base case would reduce total proceeds to $140 billion and company proceeds to $98 billion. Absorption would become easier, but the companies would have to scale back their funding plans. Maintaining the original company fundraising target would instead require 25% more primary shares, imposing greater dilution on existing holders. This calculation holds other terms constant; the original 5% share-sale assumption could no longer be carried over unchanged. The cost of an excessive valuation ultimately has to be allocated among price, share count, existing shareholders' exits, and the pace of construction.

The appropriate scale comparison is with securities supply that has actually reached the market. EY records 1,293 global IPOs raising $171.8 billion in 2025. Renaissance Capital reports $44.0 billion in the United States that year and $142.4 billion in 2021. The latter sample includes IPOs and direct listings with market capitalizations of at least $50 million, excluding SPACs and closed-end funds; it cannot be combined uncritically with the global series. EY annual review, page 4, Renaissance annual review, page 1

The ratios are calculated for this article. Issuance statistics provide a comparison of transaction proceeds, not cash flowing entirely to issuing companies on a net basis. The base assumptions for just two companies approach a full year's global IPO supply. That rules out dismissing their size as immaterial, but does not establish that absorption must fail. Historical supply reflects issuers' willingness to sell, valuations, and available windows; demand also responds to discounts, and deals can be downsized. The stress case would reach $414 billion after full greenshoe exercise. That creates a strong incentive to adjust prices and timing, not a mathematical threshold for a crisis.

Why spacing the deals matters

A large pool of low-risk savings is not an equivalent pool of willing AI equity buyers. ICI reported on October 1 that US money-market funds held $7.89 trillion on September 30, down $45.45 billion that week. The $175 billion scenario is 2.22% of those assets, but operating cash, collateral reserves and household savings cannot all be treated as subscription money. Nor can the weekly fall be attributed to offerings that had not settled.ICI data

Fund flows provide a closer reference for incremental risk-taking capacity. ICI reports combined net inflows of $12.875 billion into equity mutual funds and ETFs in the week ended September 23. The preceding four weeks recorded net outflows of $12.650 billion, $11.701 billion, $5.461 billion, and $16.105 billion, respectively. ICI release, September 30 One positive week does not establish a sustained inflow trend. These figures cover different investment categories and geographies; they are not subscription cash available in its entirety to the two new listings.

For the $175 billion base case, preparing funds over 26 weeks would imply an average requirement of about $6.73 billion a week, equivalent to 52% of the cited single week's equity inflow. Compressing preparation into four weeks would imply $43.75 billion a week, or about 3.40 times that inflow. These weekly figures are arithmetic budget averages. Actual settlement remains concentrated, and an IPO should not be represented as a series of equal weekly deductions. The comparison nevertheless shows how scheduling changes the urgency with which managers may need to sell other securities.

Consider a separate assumption: buyers finance subscriptions with 40% existing cash and new fund subscriptions, and 60% sales of existing securities. The base case would then require $105 billion of portfolio reallocation, averaging $4.04 billion over 26 weeks or $26.25 billion over four weeks. That is not a net outflow from markets, because the old securities still have buyers. It measures the transactions for which existing holders must find counterparties. If those counterparties will buy only at lower prices, the pressure would show up as lower valuations for other growth stocks and larger concessions on new offerings.

Cross-border capital also faces constraints. Non-dollar investors may need currency conversion or hedges, while sovereign wealth funds and pension funds face geographic allocation and position limits. Higher dollar funding and hedging costs raise their required returns. Underwriters can reduce near-term disruption by attracting long-term allocations, spacing settlement dates, and allowing existing shareholders to delay sales. If both deals depend on the same highly leveraged accounts, however, staggering merely postpones the point at which concentrated risk becomes visible.

The use of seller proceeds provides a countervailing force. Venture capital funds can distribute cash from secondary sales to their limited partners, who may subsequently increase public-equity allocations and return some money to the market. Alternatively, those investors may first meet commitments to other funds or retain liquidity reserves. Without account-level data, immediate and complete reinvestment cannot be assumed. Nor can company fundraising, secondary proceeds, and future index purchases simply be added together and called a permanent funding gap. Staggering allows time for redistribution, without guaranteeing that money returns to the same asset class.

How procurement commitments reach lenders

Long-term payment obligations provide the more direct route from AI operating risk to lenders. Reuters reported on September 29, citing confidential materials, that Anthropic had at least $518 billion of ten-year infrastructure commitments, roughly 80% non-cancellable or payable irrespective of usage, including services and equipment leases. This research obtained the Reuters account, not the confidential document. Anthropic separately confirmed a commitment exceeding $100 billion to AWS over ten years.Reuters; Anthropic

The $518 billion is neither an amount due in one year nor entirely interest-bearing debt. Dividing by ten produces $51.8 billion, an average scale indicator that cannot replace an annual payment schedule. Similarly, the approximately 80% share, equivalent to $414.4 billion, is not a debt balance that can simply be inserted into the balance sheet. Service commencement dates, equipment delivery conditions, minimum payments, default remedies, and accounting recognition all affect the timing of cash obligations. Public documentation is insufficient to conclude that a cash shortfall in any particular year is inevitable.

Funding runway should be assessed by starting with available cash plus net primary proceeds and subtracting net cash consumption over successive periods. Revenue cannot simply be netted against purchase commitments. In another purely hypothetical example, $70 billion of additional usable funds would cover two years of annual net cash consumption of $35 billion after customer collections. At $50 billion of annual consumption, that falls to 1.4 years. The example excludes existing cash and subsequent financing and is not an Anthropic cash forecast. If net consumption already includes cloud payments, the same contractual commitments must not be deducted again in full. Supplier financing may defer cash spending while increasing the concentration of future repayments.

Long-term contracts improve suppliers' demand visibility while concentrating risk in the customer's ability to pay. HyperFrame Research practice lead Stephen Sopko told TechTarget that even non-cancelable contracts depend on the financial strength of the signatory. Shane Snider's September 29 report If customer sales growth slows while minimum purchases cannot fall alongside it, operating leverage will consume newly raised equity. Suppliers that then use those contracts to support construction loans transmit operating risk to creditors.

The following stress example does not describe any disclosed borrower. Assume a project has $10 billion of collateral and $7 billion of debt, with an initial collateral haircut of 30%. If asset values fall 30% and lenders raise the haircut to 40%, borrowing capacity falls to $10 billion × 70% × 60% = $4.2 billion, leaving a $2.8 billion funding shortfall. A borrower financing long-lived equipment with short-term debt, unable to supply cash, would have to sell assets or curtail construction. GPU residual values, alternative uses for data centers, and power assets cannot all be assigned the same haircut. The example illustrates only the compounding effect of two forms of tightening.

Assume further that $5 billion of that debt needs refinancing. A 300-basis-point increase in the financing rate adds $150 million to annual interest. If cash available for debt service was $800 million against $600 million of annual principal and interest payments, coverage would fall from 1.33× to 1.07×. A further 20% decline in cash flow would reduce coverage to about 0.85×. An unsuccessful offering could then be a catalyst, but the underlying problem would be inadequate cash-flow buffers and maturity mismatch. The loan amounts in the model must not be presented as actual bank exposures.

The IMF's April 2026 report discusses a retrenchment in AI capital expenditure, tighter credit, and data-center refinancing risk, while also noting that the average private credit fund's AI-related exposure remained relatively limited at the time. The FSB identifies links between leveraged nonbanks and systemically important providers of financing as a key risk. IMF, page 38, FSB, July 9, 2025 Both support examining loans and collateral; neither supports declaring a banking crisis from aggregate fundraising figures alone.

Equity, supplier and credit returns can diverge

An AI financing can support supplier orders and protect creditors while leaving new shareholders with an expensive entry price. Reuters quoted XTB’s Kathleen Brooks on infrastructure demand and Steward Partners’ Newton Jones questioning how large losses and capital needs justify an equity valuation above $2 trillion. They address different claims on the economics.Reuters

Both can be true. Orders for suppliers do not necessarily translate into adequate returns for shareholders in model developers. Primary equity can improve a customer's credit quality while exposing new investors to substantial valuation risk. Conversely, delaying an IPO need not immediately reduce equipment purchases: the customer may still have cash, operating collections, or other long-term capital. Translating offering success or failure directly into an industry-wide change in capex overlooks existing funding and differences in contractual timing.

For creditors, new equity generally adds a loss-absorbing buffer. For shareholders, the same money protecting debt service may soon be spent and expensed. An offering can therefore improve credit quality while reducing prospective equity returns. The use of proceeds, suppliers' cash collections, and the offer price require separate assessment; a favorable industry narrative cannot substitute for analyzing returns across these different securities.

The conditions that would overturn the base case should be specific. If final offerings are substantially smaller, separated by several months, and predominantly primary, this article's estimate of pressure to reallocate existing equity holdings should be reduced. If secondary sales account for a large share, the companies will receive less financial protection than issuance headlines suggest. If public payment schedules reveal concentrated spending in the first two years that available cash, operating inflows, and committed financing cannot cover, the refinancing-risk assessment should be raised.

Only broader evidence—simultaneously rising collateral haircuts, redemption-driven forced selling, concentrated losses disclosed by banks or private credit lenders, and financing cut off even to healthy borrowers—would justify escalating an industry adjustment into a systemic-risk diagnosis. The strongest evidence against that outcome would be increased cash after issuance, long-term financing replacing short-term debt, customer collections covering minimum purchases, and functioning loan markets. For global investors, the most defensible view today is to expect substantial competition for capital and divergent pricing, while refusing to turn undisclosed bank exposures into a prediction of an inevitable crisis.