AI Revenue Is Booming. But How Much of the Demand Stands on Its Own?

AI growth does not require capital to disappear for its assumptions to be tested. The real question is whether companies can keep executing when financing becomes more expensive, selective, or conditional—and whether each dollar raised makes future growth easier to finance.

By Tom Krutilek
September 15, 2026

AI Revenue Is Booming. But How Much of the Demand Stands on Its Own?

AI revenue is growing at a pace few markets have ever produced. But “revenue” and “demand” are not the same thing, and not all demand is equally durable. Some is genuinely independent — it would exist on comparable terms even without overlapping relationships among suppliers, investors, lenders, and guarantors. Other demand is increasingly supported by the same ecosystem that benefits from it. The question that increasingly separates a durable business from a fragile one is not how fast revenue is growing, but how much of the demand behind it would still stand on its own.

In brief

Recognized revenue and economically independent demand are not the same thing. Revenue can be entirely real, properly recognized, and still depend on financial support from the ecosystem that benefits from it.
AI demand sits on a spectrum — from independent, to assisted, to ecosystem-financed — and investors and credit analysts are increasingly scrutinizing where a company’s demand falls on it.
For leadership, the distinction has become strategic. Companies that can demonstrate independent demand can strengthen investor confidence; those that cannot may face greater scrutiny even when headline growth looks similar.

When the AI ecosystem helps finance its own demand

In August 2026, Nvidia announced plans with six of the world’s largest asset managers — Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR — to mobilize more than $500 billion for AI infrastructure, with Nvidia reportedly prepared to provide support for roughly $125 billion of that financing. It is one of the largest expressions of a broader pattern emerging across AI: suppliers are helping finance the ecosystem that ultimately buys their products. These arrangements can include vendor financing — where a supplier, directly or indirectly, helps provide the money or credit its customer uses to make a purchase — as well as strategic investments, guarantees, and warrants. Suppliers invest in customers. Customers hold stakes or warrants in suppliers. Chipmakers help fund companies that may later become major customers — and in some cases have already committed to buy their hardware.

The issue is now drawing attention well beyond AI skeptics. In late August, S&P Global Ratings projected that AI infrastructure investment will exceed $1.3 trillion by 2027. Days later, it warned that credit quality among hyperscalers — the largest cloud and computing providers — is gradually weakening as capital spending rises, financing becomes more complex, and returns may take longer to materialize.

Nvidia’s position is that its initiative brings independent, long-duration institutional capital into the market, with individual projects underwritten on their own merits. Skeptics, including investor Michael Burry, see echoes of earlier episodes of circular financing and financial engineering that accompanied market excesses and corrections. Both readings can be partly right. Nvidia can help bring genuinely independent capital into AI infrastructure while its guarantees and other financial relationships can still reduce the independence of some downstream demand. The harder investor question is how much of that demand would exist without the ecosystem’s financial support.

That question is a counterfactual: Would this customer make substantially the same purchase, on roughly the same timetable, if the supplier, investor, and guarantor were not sometimes the same parties? Economic independence, in other words, is not an accounting designation. It is an investor test.

Real is not the same as independent

Accounting already addresses many direct customer incentives. Under current revenue standards, certain credits, discounts, and even warrants provided to customers can, in some circumstances, reduce the amount reported as revenue. But the investor question goes beyond whether the revenue is properly recognized: even properly recognized revenue may not reveal whether the customer’s underlying demand would exist on comparable terms without the surrounding financing or support.

Revenue tells us what has been recognized as earned. It does not, on its own, reveal how much financial support was required to make the purchase possible.

The revenue can be real. The harder question is whether customer demand can stand on its own.

A spectrum of demand independence

One useful way to analyze AI demand is along a spectrum with three reference points.

Independent demand is the clearest signal of underlying market demand: the customer would make substantially the same purchase without material financial support from the supplier or other financially interested parties. This is not about whose dollars are physically spent — companies routinely finance purchases with ordinary debt and leases, which is no sign of weak demand. What matters is whether the purchase would still make economic sense on normal market terms without vendor or ecosystem support.
Assisted demand is demand from a customer with a genuine economic reason to buy, but whose adoption is accelerated by discounts, promotional credits, or a strategic partnership. This is ordinary go-to-market, not a red flag. The question is one of trajectory: does the assistance lead to demand that can stand on its own at normal market terms over time, or does continued purchasing remain dependent on it?
Ecosystem-financed demand is where the customer’s ability to purchase depends materially on financing or support from the supplier or other parties that benefit from the transaction — a structure no longer unique to Nvidia, as AMD’s multibillion-dollar investment-and-supply relationship with Anthropic illustrates. The underlying transaction can still be entirely legitimate, and the resulting revenue properly recognized. But it changes what an investor has to underwrite, because the demand and the financing are no longer cleanly separable. For investors, that distinction affects how growth is evaluated: independent demand can support greater confidence in recurring economics, while ecosystem-financed demand requires greater scrutiny of the financing beneath it.

Demand independence can also become easier to observe as you move from infrastructure through the model layer toward applications. As the research firm IDC has argued, enterprise application revenue can provide a cleaner signal because a customer that renews or expands a subscription or service is making an operating budget decision based on the value it believes the offering delivers, typically with less direct investment overlap between customer and supplier. Cleaner does not mean automatically clean — application vendors also use credits, bundling, and strategic partnerships — but the same revenue growth can carry very different degrees of economic independence depending on where and how it is generated.

Six tests for demand independence

The goal is not to label revenue “real” or “circular.” It is to understand how much of the growth is supported by demand that can stand on its own. Six questions provide a practical framework:

Counterfactual — Would the purchase still happen? — Without the related investment, financing, or guarantee, would the customer still make substantially the same purchase on roughly the same timetable?
Funding source — Who is funding the purchase? — Is it supported by normal operating or capital budgets, or materially dependent on vendor or ecosystem financing?
Post-subsidy — Does demand persist when incentives end? — Do usage and purchasing continue once promotional credits, extraordinary discounts, or other incentives expire?
Customer independence — How financially independent are the customers? — Are significant portions of demand linked through shared investors, guarantees, or sources of financing?
Market terms — Would the economics still work?— Would the purchase still make economic sense at market pricing and ordinary third-party financing terms?
Renewal independence — Does demand become more independent over time? — At renewal or expansion, is customer spending becoming less dependent on ecosystem support?

A sale shows demand existed once. Demand independence asks whether that demand would still exist on normal market terms.

Putting the independence test into practice

SB Energy, the SoftBank-backed data-center developer that filed to go public in September, shows why the test matters. Its filing discloses a contracted backlog of roughly $439 billion — future revenue expected over multi-decade agreements, not current revenue — compared with about $140 million in revenue for the first half of 2026 and no operating data centers yet. OpenAI is simultaneously an investor, a warrant-holder, and the anchor tenant on which much of the build depends. OpenAI's warrants are valued at nearly $5.5 billion, representing the estimated value of the stake rather than $5.5 billion of cash invested. Nvidia, meanwhile, has committed $3 billion in equity — a $1.5 billion prepaid purchase of shares at a discount to the IPO price, plus a $1.5 billion private placement alongside the offering — and has provided a guarantee of up to $105 billion backing OpenAI's lease obligations on the first phase of SB Energy's largest campus. The filing notes that the campus's financeability depends in significant part on that guarantee, because OpenAI is not an investment-grade tenant. It also states that SB Energy is substantially dependent on OpenAI. None of this makes SB Energy a bad business — but viewed through the independence test, the demand and the financing cannot yet be cleanly separated.

The application layer can look very different. ServiceNow reported in the second quarter that its AI annual contract value (ACV) had crossed $1 billion. In the same quarter, the company reported 123 transactions worth more than $1 million in net-new ACV, 658 customers each generating more than $5 million in ACV, and a renewal rate near 98%. Those figures do not make every dollar of AI demand automatically independent. But the breadth of the customer base and the high renewal rate make the underlying demand considerably easier to evaluate than a backlog concentrated in a single, financially intertwined customer.

The market is already recalibrating some of these structures. In August, Nvidia reportedly paused some transactions under a program introduced the previous month that had offered cloud providers credit support in exchange for a share of their revenue, amid antitrust concerns and broader questions about how much influence a supplier should have over the businesses of its customers. Nvidia said the model remains in place and continues to evolve, and its most recent quarterly filing puts commitments under the program at $36 billion.

How leadership can strengthen the independent core

The same framework investors use to evaluate demand can also help leadership strengthen it — and the most important decisions are made inside the business long before diligence begins.

Start with how demand is reported to the board. Rather than relying on a single blended top-line number, track the independent core separately across the commercial chain — pipeline, bookings, recognized revenue, and cash collections — and identify the portion that is assisted or financially linked. That makes it easier to see when the quality of demand is changing even if headline growth remains strong.

Stress-test forecasts against a zero-subsidy baseline: what would growth look like if promotional credits, strategic discounts, or other incentives were reduced or removed? Shift go-to-market incentives away from simply initiating pilots and toward production milestones, expansion, and renewals, where demand independence is easier to assess. A renewal can itself be discounted or supported, so it should be treated as evidence of durability rather than proof.

And where related-party, assisted, or financially linked demand is material, disclose it proactively. Handled well, that transparency can strengthen investor confidence rather than erode it, because it signals a leadership team that understands not only how fast the business is growing, but how durable the demand supporting that growth may be.

Financing support is not the problem. Reliance is.

Assisted and even ecosystem-financed demand can be entirely rational. Infrastructure has long been built ahead of demand, often with debt, long-term commitments, and strategic financing. In a capacity-constrained market, a supplier helping customers finance purchases can be a sensible response to genuine scarcity.

The standard, then, is not whether financing relationships exist at all. Many infrastructure markets depend on them. The sharper distinction is between reciprocity and reliance. “We invest in you, and you buy from us” can be common and economically rational. The concern is material reliance — where demand would weaken substantially if the customer could no longer fund its purchases without continued support from parties that benefit financially from the transaction.

The telecom build-out of 1999–2001 offers a useful precedent. The problem was not vendor financing itself, but the point at which financially weaker customers became dependent on suppliers to finance purchases they could not otherwise sustain. That is the risk investors should watch for in AI as well: financing can accelerate real demand, but it can also make that demand appear more durable than it would be without continued support. At that point, the risk can compound: if demand weakens, a supplier may face declining sales while also becoming more exposed to the guarantees or financial commitments it made to support those customers.

Ecosystem financing can accelerate demand. The risk begins when it becomes a key reason demand exists.

A separate question is capital resilience. Capital dependence describes how much outside financing a company needs; capital resilience describes how well the business can adapt when the conditions attached to that financing change. A business may rationally depend on outside capital to fund expansion. What matters is whether its growth strategy remains viable if that capital becomes more expensive, more selective, more restrictive, or slower to arrive. That is distinct from demand independence, and it is the subject of the next article in this series.

Booming — but built to compound?

The question is no longer whether AI revenue is growing. It plainly is. The harder question — for founders, CEOs, boards, and the investors who back them — is how much of that demand would still stand on its own when incentives fade, capital tightens, and capacity catches up?

For leadership teams, being able to answer that convincingly is becoming an increasingly important part of the investor narrative. For investors, distinguishing assisted growth from economically independent demand may help identify which businesses are built to compound — and which remain more dependent on the financing beneath their growth.

Access to capital is a competitive advantage. Customer demand that stands on its own is the foundation of a durable business model.

#ArtificialIntelligence #VentureCapital #AIInfrastructure #CapitalReadiness #InvestorNarrative

References

About the Author

Tom Krutilek is a Board Advisor and Chief Marketing Officer who works with early-stage and growth-stage companies on capital readiness, investor narrative, go-to-market strategy, and AI strategy and governance. His work focuses on a central question: whether growth is truly investable—not just impressive on paper.

He helps founders, CEOs, and boards connect commercial performance, capital strategy, and execution to the evidence investors need to underwrite growth. That includes strengthening the investor narrative, identifying gaps in capital readiness, building measurable growth systems, and evaluating where AI can create durable business value and competitive advantage.

Tom is also the founder of The Investor Narrative, a publication focused on private capital markets from the perspective of investor readiness. It examines the questions behind the headline numbers—how investors assess growth, capital efficiency, demand quality, competitive advantage, governance, and the assumptions that ultimately determine whether a company is ready for investment capital.

Connect with Tom on LinkedIn