What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
OpenAI’s reported plan to burn about $115 billion in cash from 2025 through 2029 is a warning about capital intensity and financing dependence—not proof that artificial intelligence is worthless or that a sector-wide bubble is certain. The figure comes from internal projections reported by The Information, not an audited public forecast. It describes expected cash outflows exceeding inflows, not necessarily $115 billion of accounting losses or infrastructure spending paid entirely by OpenAI.
What the $115 billion figure actually means
Cash burn is the net amount of cash a company expects to consume over a period. It differs from several related measures:
- Net loss: an accounting result that can include non-cash expenses such as stock compensation.
- Capital expenditure: spending on data centers, servers, networking and other long-lived assets.
- Operating expense: recurring research, compensation, cloud usage, sales and administrative costs.
- Committed spending: contractual or announced obligations that may be paid later and may involve partners, leases or debt.
The reported $115 billion is therefore best read as a cumulative internal cash-use projection. It should not be casually described as “$115 billion in losses.” The Information also reported separate loss and cost projections, including non-cash items, in a later article.
The reported burn trajectory
These are rounded projections reported in 2025, not confirmed results. Because the reports use terms such as “more than” and “approximately,” they should not be added together to manufacture a precise 2029 number.
Recommended Free Tools
#1 Best Overall
| Year | Reported projected cash burn |
|---|---|
| 2025 | More than $8 billion |
| 2026 | More than $17 billion |
| 2027 | Approximately $35 billion |
| 2028 | Approximately $45 billion |
| 2029 | Remaining amount needed to reach roughly $115 billion cumulatively; no precise figure was established |
The 2026 estimate was reportedly over $10 billion higher than an earlier projection. That widening forecast illustrates how quickly assumptions about models, capacity and demand can change.
Why the costs are so large
Training and inference
Training new models requires large clusters of accelerators, networking and storage. Inference—serving answers to users and applications—can become the larger recurring bill when millions of requests, long context windows or reasoning and agent workloads are involved. The Information reported that internal documents contemplated computing costs of approximately $9.5 billion annually in 2026; this is an unverified projection, not an audited expense.
Data centers, chips and power
OpenAI needs access to GPUs or custom accelerators, high-speed networking, cooling, electricity and physical facilities. Capacity can be obtained through owned facilities, cloud contracts, leases, joint ventures or supplier financing. A headline commitment therefore does not establish how much cash OpenAI has already spent.
Rank #2
Research, products and partnerships
Highly paid technical staff, safety and research programs, consumer and enterprise products, coding and agent tools, sales and support all add cost. Payments to cloud and infrastructure partners can also rise with usage. Long-term minimum-capacity contracts may create obligations even before capacity is fully utilized.
The revenue bet behind the spending
The same reports describe an exceptionally aggressive growth case: roughly $12 billion or slightly more in projected 2025 revenue and approximately $100 billion in annual revenue by 2029. Another forecast described about $110 billion of services revenue across 2026–2030. Those figures cover different periods or forecast versions and cannot be combined into a single audited model.
Potential revenue sources include ChatGPT subscriptions, business plans, API usage, coding and agent products, licensing and distribution partnerships, and possible advertising or commerce. The Information reported that one revision lowered the five-year API revenue projection by about $5 billion while increasing expectations for other services. That is a reminder that top-line growth is not guaranteed in every product category.
Questions that matter more than user counts
- What proportion of users pay, and what is revenue per paying account?
- How much enterprise revenue recurs and expands at renewal?
- What is gross margin after the compute required to serve each task?
- Can prices rise without reducing usage or switching to cheaper models?
- Do agents create genuinely new spending or deliver existing software at higher compute cost?
Revenue growth alone cannot establish a viable business. A company can report rapid sales while subsidizing every interaction.
Stargate is not a $500 billion OpenAI bill
In January 2025, OpenAI, SoftBank, Oracle and MGX announced Stargate as a company intended to invest up to $500 billion over four years, with $100 billion described as immediately deployed. OpenAI said SoftBank would have financial responsibility and OpenAI operational responsibility in the announcement: OpenAI’s Stargate release.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →That ambition is different from OpenAI’s $115 billion cash-burn projection. It may involve partner equity, debt, leases, cloud contracts and supplier financing. OpenAI later described a goal of securing 10 gigawatts of U.S. AI infrastructure by 2029: its April 29, 2026 infrastructure update. “Up to,” planned capacity and announced investment are not the same as cash spent or operating facilities.
Who bears the risk?
The economics run through an ecosystem rather than OpenAI alone.
| Participant | Exposure |
|---|---|
| OpenAI investors | Additional equity may be required if cash use stays ahead of internally generated funds. |
| Cloud providers | Capacity commitments and data-center spending depend on OpenAI usage and creditworthiness. |
| Chip and networking suppliers | Orders can be large, but cancellations or slower deployments can affect demand. |
| Infrastructure partners and lenders | Debt, leases and minimum-purchase contracts can shift risk away from OpenAI’s balance sheet without eliminating it. |
| Enterprise and consumer customers | They may face price changes, vendor lock-in or service disruption if economics deteriorate. |
| Taxpayers and communities | Public incentives, power infrastructure and environmental costs can socialize part of a buildout’s cost. |
Reported examples include forecast Microsoft-owned data-center spending rising from about $13 billion in one year to $28 billion in 2028, and an Oracle arrangement involving roughly 4.5 gigawatts of planned capacity. These are forecasts or planned capacity, not verified delivered facilities: The Information’s infrastructure report and the CNA report.
Is this an AI bubble?
The case for bubble risk
- Valuations and infrastructure commitments may assume demand and productivity gains that have not yet been proven.
- Falling model prices or open-source competition could reduce revenue faster than costs decline.
- Cloud and chip suppliers may be financing customers whose own cash generation is inadequate.
- Unused capacity becomes a liability when long-term contracts outlast demand.
- Corporate buyers may find that measurable productivity gains are smaller than promotional claims.
The case against a simple bubble label
- Telecommunications, cloud computing and the early internet also required infrastructure years before mature returns.
- Scale can lower inference costs and improve margins.
- A useful technology can survive even when particular companies or investments fail.
- Partner financing distributes exposure across the ecosystem rather than placing every dollar on OpenAI.
- If forecast revenue approaches $100 billion in 2029, very large absolute spending could be economically rational.
“AI bubble” can mean a valuation bubble, an infrastructure overbuild, a financing bubble, exaggerated product demand or inflated productivity expectations. Those risks can coexist with genuine long-term utility.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Best Value
What would make the strategy work?
- Revenue grows near the forecast path while gross margin improves.
- Model and serving costs fall faster than usage rises.
- Customers pay for high-value enterprise, coding and agent products rather than only discounted access.
- Long-term capacity is highly utilized.
- Financing remains available on acceptable terms.
- OpenAI captures enough value instead of passing most economics to cloud and chip suppliers.
- Competitors do not force prices down faster than efficiency improves.
What could break the thesis?
- Revenue misses while fixed infrastructure commitments remain.
- Models become interchangeable and prices approach commodity levels.
- Advanced reasoning and agent workloads keep inference costs high.
- Construction, power, chip supply or export restrictions delay capacity.
- Financing becomes more expensive or unavailable.
- Partners renegotiate contracts or reduce credit support.
- Enterprise buyers demand privacy, copyright protection, reliability or indemnity that raises costs.
- Consumer willingness to pay stagnates or revenue remains concentrated among a few large customers.
- A new architecture makes existing hardware less valuable.
What to monitor before drawing your own conclusion
- Actual annual revenue: distinguish recognized revenue from annualized revenue, bookings and contract value.
- Gross margin after inference: this shows whether each additional customer creates value.
- Cash and financing: track new equity, debt, leases and the terms attached to them.
- Paid-user growth and retention: free usage is not evidence of monetization.
- Enterprise expansion: renewals and larger deployments matter more than pilot announcements.
- Capacity utilization: compare operating workloads with planned gigawatts and data centers.
- Serving cost per completed task: token prices alone can conceal expensive reasoning workloads.
The Information reported approximately $7.6 billion of cash at the end of a prior year and said projected funding was expected to last into 2027 at the forecast burn rate, while additional fundraising had been contemplated. Those were time-specific reported figures, not a current solvency statement: The Information’s financing report.
What this means for personal-finance and business decisions
Do not treat OpenAI’s spending forecast as a reason to buy an AI subscription, commit to a long-term API contract or make a securities trade. For a real workload, compare OpenAI with alternatives such as Anthropic Claude, Google Gemini, Amazon Bedrock and Microsoft Azure AI Foundry on task quality, total cost, latency, privacy, portability and governance. Test representative tasks, set spending limits and avoid assuming today’s model prices or capabilities will remain stable.
Verdict
The reported $115 billion forecast is best understood as a stress test of whether rapid revenue growth can finance an unusually capital-intensive platform race. It is serious evidence of financing and execution risk, but it does not prove that AI is a bubble, that OpenAI will spend exactly $115 billion, or that the technology lacks durable economic value. The decisive evidence will be recurring revenue, post-inference margins, utilization and the terms on which future capital is raised.
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.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors
