OpenAI could face a serious funding squeeze by mid-2027, but that date is an analyst’s scenario—not a company forecast, a verified cash-exhaustion date or a bankruptcy prediction. The warning, attributed to economist Sebastian Mallaby in January 2026, reflects the scale of OpenAI’s projected cash needs as it builds and reserves computing capacity. Reports also describe fast-growing revenue, very large infrastructure plans and forecasts that vary by source and vintage. The key question is whether revenue, new financing and partners can cover cash outflows as they fall due.
Where did the mid-2027 warning come from?
The mid-2027 date was attributed to Sebastian Mallaby, an economist and Council on Foreign Relations fellow, who reportedly expected OpenAI to run out of money within about 18 months. It is an outside analyst’s judgment based on reported financial projections and infrastructure plans, not an official OpenAI cash-flow forecast. The reporting does not establish that OpenAI will have no cash on a particular date. Tom’s Hardware’s account of the warning and WinBuzzer’s report frame it as a risk tied to spending and financing needs.
Several different financial concepts can get compressed into “run out of cash,” but they are not interchangeable:
- Operating losses: expenses exceed revenue over an accounting period.
- Negative free cash flow: cash leaves the business after operating needs and capital investment.
- Cash exhaustion: available liquidity is insufficient to meet obligations without new financing or changes to spending.
- Insolvency or bankruptcy: legal and financial conditions that do not automatically follow from a forecasted funding gap.
A company can be growing and strategically important while still needing to raise capital repeatedly. Public reporting cited here does not establish OpenAI’s exact cash balance as of August 18, 2026, or confirm that the company has a particular number of months of cash remaining.
#1 Best Overall
What the reported figures do—and do not—show
The figures below come from different reports and forecast vintages. Reported revenue and spending figures are not audited accounts in the material cited here; projections are not guarantees. In particular, spending, cash burn and infrastructure commitments measure different things.
| Period or horizon | Reported figure | How to read it |
|---|---|---|
| 2025 | About $13 billion revenue; about $8 billion spending | Reuters reported these figures, citing a source familiar with the matter. The spending figure is not necessarily GAAP expenses or total cash burn. |
| 2025 | Adjusted gross margin of about 33%, down from about 40% in 2024; inference expense reportedly quadrupled | Reuters reported these figures. They illustrate pressure from serving models, but do not by themselves disclose overall profitability or cash flow. |
| 2026 | More than $17 billion projected burn in one forecast | Attributed to The Information’s reported company projections; it is not an audited result. |
| 2027 | About $35 billion projected burn in one forecast | Attributed to The Information. A separate forecast report describes roughly $20 billion of 2027 burn, so the estimates should not be blended; they may reflect different definitions, periods or revisions. |
| 2028 | About $45 billion projected burn | Attributed to The Information’s reported projections. |
| Through 2029 | About $115 billion cumulative burn | Attributed to The Information; this is a cumulative projection, not a current cash balance. |
| Through 2030 | About $600 billion in compute spending; more than $280 billion in cumulative revenue discussed | Reuters reported these targets based on CNBC reporting. They are long-range estimates with different meanings, not proof that spending will exceed revenue by a fixed amount. |
| Through 2033 | About $1.4 trillion in broader infrastructure commitment | Reported in secondary coverage as a longer-horizon infrastructure figure. Its scope differs from the compute-spending estimate through 2030; the figures should not be added together. |
Sources: Reuters reporting on revenue, spending, margins and compute projections; The Information’s reported burn projections; a separate reported forecast with a different 2027 estimate; and secondary coverage of the longer-horizon infrastructure figure.
Why AI growth can consume cash so quickly
AI infrastructure costs are not limited to training a model once. OpenAI must also serve user and business requests, maintain capacity, and support ongoing research and product development. Each answer generated can require computation, so usage growth may raise costs along with revenue.
- Model training: large runs require substantial computing capacity, engineering and experimentation.
- Inference: serving text, code, image, audio or video requests consumes computing resources repeatedly. Reuters reported that OpenAI’s inference expense quadrupled in 2025.
- Physical infrastructure: GPUs and servers need networking, storage, electricity, cooling and data-center space.
- Capacity access: cloud contracts and reservations can secure future computing supply, potentially with minimum payments.
- People and operations: research, engineering, safety evaluation, security, sales and enterprise support also cost money.
That creates a different economic profile from conventional software. A software service may add customers at low marginal cost; a high-volume AI service must pay to run the computation behind each workload. Lower inference costs, higher utilization and higher-value customer uses can improve the economics, but rising usage alone does not prove that margins are improving.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesInfrastructure plans are not the same as cash already spent
A headline infrastructure total can combine commitments and plans that have very different implications for near-term liquidity. It does not necessarily mean OpenAI has paid that amount, owes it all immediately, or will fund every facility itself. The key details are who owns the asset, who borrows to build it, when payments fall due, and whether capacity agreements include minimum purchases or cancellation rights.
- Cash already paid is money that has left the company.
- Operating expenses include costs incurred to run the business; they do not map one-for-one to cash paid in the same period.
- Capital spending by a partner may put the facility and its financing on the partner’s balance sheet rather than OpenAI’s.
- Capacity reservations and long-term contracts can create future payment obligations without requiring the full contract value upfront.
- Debt raised by an infrastructure provider is that provider’s financing, though the arrangement can still affect OpenAI through pricing and contract terms.
- Leases, minimum-purchase obligations and joint ventures distribute costs and risks differently; the public figures cited do not establish the precise terms of every commitment.
For this reason, the reported $600 billion compute-spending figure through 2030 and the roughly $1.4 trillion broader infrastructure figure through 2033 should be treated as estimates with different scopes and time horizons—not combined into a single bill payable by OpenAI.
How partners can help—and add exposure
OpenAI’s infrastructure strategy involves partners including Microsoft and Azure, Oracle, SoftBank and other cloud or infrastructure providers. Partners can own or finance facilities, provide cloud capacity, or support projects through strategic investment. That can reduce the amount OpenAI must fund upfront, but it does not make capacity free: costs may return through usage charges, leases or long-term commitments.
Oracle’s own financing needs illustrate that capital intensity can shift across a chain of counterparties. Reuters reported that Oracle forecast fiscal-2027 capital expenditures of up to $95 billion and expected to raise nearly $40 billion through debt and equity financing in 2027. The company said a major Texas Stargate data center being built with OpenAI and others would be more than three-quarters complete within 90 days of its June 2026 announcement. Those statements describe Oracle’s plans and project progress, not a guarantee of OpenAI’s financing or access to unlimited computing capacity. Reuters’ report on Oracle’s outlook and Stargate also noted Oracle’s 2026 spending of about $55.66 billion, above its $50 billion target.
Rank #3
What partner financing can change
- Partners can spread construction costs over time or fund assets themselves.
- OpenAI may lease or buy computing capacity instead of owning every server and data center.
- Strategic investors may provide equity or other support alongside commercial agreements.
What it cannot eliminate
- Long-term contracts can become costly if demand, utilization or model economics disappoint.
- Minimum-payment terms may limit the ability to reduce spending quickly.
- Specialized facilities may be difficult to repurpose for other customers.
- Partners can face their own funding pressure, construction delays or financing costs.
- A company seeking emergency capital may have less leverage to negotiate its commitments.
Reported plans and project announcements do not establish that every financing commitment has closed, or that every planned facility will be delivered on schedule.
Why strong revenue and a high valuation do not settle the cash question
Reuters reported approximately $13 billion in 2025 revenue and more than $280 billion in cumulative revenue discussed through 2030. Those numbers show the scale of the opportunity, but revenue is not cash available after costs. If inference, infrastructure and other expenses rise nearly as quickly, revenue growth can coexist with substantial losses and funding needs.
The quality of growth matters as much as its headline rate. Subscription revenue, usage-based API sales and enterprise contracts can have different compute requirements, support costs and partner revenue shares. The reported figures cited here do not provide enough detail to determine profitability by customer type or product. Relevant questions include whether enterprise workloads generate more revenue per unit of compute, whether prices are falling faster than serving costs, and whether capacity is being used efficiently.
Likewise, valuation is not a cash balance. Reports of a financing process exceeding $100 billion or a valuation in the hundreds of billions describe prospective fundraising or investor pricing—not confirmed, unrestricted cash on hand. Equity can dilute existing owners; debt adds repayment obligations; strategic financing may come with commercial or governance terms. Reuters’ account of the reported financial outlook also underscores the distinction between revenue targets and the capital needed to pursue them.
What could happen before a cash shortfall becomes acute?
Cash exhaustion is not the only possible outcome of a mismatch between spending and available funds. The most plausible paths depend on whether OpenAI can raise capital, grow profitable revenue and adjust infrastructure plans before obligations come due.
Funding and growth keep pace
A large equity round or strategic investment, faster enterprise growth, cheaper inference and higher capacity utilization could make the reported risk less immediate. In this scenario, partner financing carries more of the construction burden while OpenAI earns enough from products to support ongoing costs. A financing announcement would matter most once funds actually close and its terms are known.
A funding gap prompts restructuring
OpenAI could seek to renegotiate capacity commitments, slow expansion, defer projects, rely more on leased compute or prioritize smaller and less costly models. These steps may preserve operations while changing the company’s growth rate, access to computing resources or negotiating position. Additional debt or equity could bridge a gap, but the price could include dilution, interest costs or strategic concessions.
Funding is delayed and costs remain high
If revenue misses internal targets, margins weaken, infrastructure obligations remain inflexible and financing becomes unavailable or too costly, liquidity pressure could intensify. The response could involve emergency financing, payment negotiations, reduced capacity or a broader restructuring. A sale of a stake or strategic merger is possible in principle, but an acquisition by a specific company is not established by the reported forecasts.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Best Value
Why the comparison with Microsoft and Meta matters
Mallaby’s reported argument contrasts a standalone AI company with large technology companies that can draw on established businesses to support AI investment. Microsoft and Meta have mature revenue streams that can help absorb spending, while OpenAI has fewer legacy businesses to offset infrastructure and model costs. That does not mean incumbents’ AI investments will earn an acceptable return, or that OpenAI cannot attract substantial capital because of its strategic value.
Nor is the pressure confined to OpenAI. Reuters has reported that AI spending is weighing on free cash flow at major technology companies, showing that the wider industry faces a capital-intensity challenge. Reuters’ analysis of hyperscaler free-cash-flow pressure provides that broader context. Microsoft’s relationship with OpenAI may matter commercially, but its exact future financial support and obligations should not be assumed.
What to watch to judge whether the warning is becoming more or less likely
The mid-2027 scenario becomes more concerning if cash burn rises toward the more aggressive projections while financing remains uncertain. It becomes less concerning if OpenAI secures funding, improves unit economics or reduces the amount it must pay directly for infrastructure. Readers can track several indicators in future reporting:
- Revenue growth compared with reported targets, alongside adjusted gross margin.
- Inference costs and whether serving costs fall as models and hardware improve.
- Financing that has closed, not merely been announced, including its terms.
- Capacity delivered versus planned, and whether data-center timelines change.
- Evidence of contract minimums, renegotiations or delayed infrastructure commitments.
- Enterprise customer retention and expansion, and whether higher-value workloads support better margins.
- Whether partners continue financing capacity and can manage their own capital needs.
Reuters’ reporting on major technology companies’ cash-flow pressure is a reminder that partners’ balance sheets matter too; shifting infrastructure spending to a cloud or data-center provider relocates risk rather than necessarily removing it.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
