Nvidia Q2 Earnings: Burry Warns of an AI Bubble

Nvidia is heading into one of the market’s most closely watched earnings reports with Wall Street expecting roughly $92 billion in second-quarter revenue, while investor Michael Burry is challenging a deeper assumption behind the AI trade: whether the demand supporting the boom is as economically independent as it appears. Nvidia is scheduled to report its fiscal second-quarter results after the U.S. market closes on August 26, 2026.

The strategic question for boards is no longer simply whether Nvidia beats expectations. It is whether executives can distinguish profitable end-market demand from infrastructure spending increasingly supported by financing relationships among technology companies, cloud providers, infrastructure operators and investors.

Boardroom Briefing

Nvidia’s earnings have become a major signal for the AI capital-expenditure cycle because its sales sit at the center of global data-center investment.
Nvidia’s revenue is expected to reach roughly $92 billion for the quarter, representing about 96% year-over-year growth based on current Bloomberg consensus cited by Yahoo Finance.
Michael Burry’s circular-revenue claim applies to chipmakers broadly; he said roughly 100% of announced chipmaker revenue is circular and attributed the claim to the Bank for International Settlements.
AI infrastructure financing is moving toward institutional capital, with Nvidia announcing partnerships intended to mobilize more than $500 billion of third-party capital.
Utilization and cash flow provide a stronger test of infrastructure economics than announced capacity or hardware orders alone.
Nvidia’s financing strategy creates a legitimate strategic opportunity while also making counterparty exposure, guarantees and residual-value assumptions more important for investors and boards.

Nvidia’s Blowout Earnings Test Has Become a Referendum on the AI Economy

Nvidia’s earnings matter because its results provide one of the clearest market signals for the scale and direction of AI infrastructure spending.

Nvidia’s previous quarter produced $81.6 billion in revenue, while Data Center revenue reached $75.2 billion, according to the company’s financial results. Current expectations put second-quarter revenue at roughly $92 billion, meaning investors are preparing for another period of extraordinary growth.

That scale gives Nvidia’s earnings significance well beyond the semiconductor industry. Its results influence expectations for hyperscaler capital spending, AI-cloud expansion, data-center construction, power demand, networking equipment and the broader technology market.

Yahoo Finance reports that Bloomberg consensus calls for $2.09 in adjusted earnings per share on $92 billion of revenue. The same report says Data Center revenue is expected to exceed $85 billion, with Hyperscaler revenue estimated at approximately $43.5 billion and AI, Cloud, Infrastructure and Enterprise-related sales at about $41.7 billion.

The numbers create an unusual market test.

Nvidia can continue reporting extraordinary growth while the customers buying its hardware are still trying to prove that their own AI investments generate sufficient returns.

That distinction matters to CFOs and investment committees. Supplier revenue and customer return on invested capital are related, but they are not the same measurement.

Why one company’s quarterly results now influence AI infrastructure, cloud investment and equity-market expectations

Nvidia’s position at the center of AI capital spending makes its earnings a proxy for the health of a much broader investment cycle.

A strong Nvidia outlook can reinforce expectations for new AI factories, cloud capacity, networking equipment and power infrastructure. A weaker outlook could force investors to reassess the pace at which those investments can produce commercial returns.

The current debate is not about whether companies are spending money on AI. They clearly are.

The harder question is whether the revenue generated by those investments will eventually justify the capital being committed today.

That is precisely where Burry’s argument enters the discussion.

Michael Burry’s Circular Revenue Thesis: Where the Financial Risk Could Be Building

Michael Burry’s circular-revenue thesis is an argument that interconnected investments, financing arrangements and guarantees can make chipmakers’ announced demand appear more independent than it actually is.

In comments reported by Stocktwits, Burry said “roughly 100%” of chipmakers’ announced revenue is circular, adding that the claim was based on the Bank for International Settlements. He also described Nvidia’s upcoming earnings as “lights out,” meaning his bearish position is not dependent on the company missing its immediate quarterly expectations.

That distinction is critical.

Burry’s statement applies to chipmakers broadly. It should not be presented as a verified claim that 100% of Nvidia’s recognized accounting revenue is circular.

His argument is instead a thesis about the financial relationships surrounding AI spending.

The issue can be illustrated through a simplified transaction. A technology company invests in an AI developer. That developer obtains financing to expand compute capacity. The infrastructure operator then purchases chips from the original technology supplier. Revenue is recorded by the supplier, while capital relationships elsewhere in the chain may have helped make the purchase possible.

Such financing is not automatically artificial or fraudulent.

Infrastructure markets have always depended on capital.

The concern begins when financing availability becomes a substitute for independently demonstrated customer economics.

How investments, customer financing and guarantees can create economic interdependence

A vendor-supported financing structure becomes more difficult to evaluate when the same group of companies influences both the supply of AI infrastructure and the capital available to purchase it.

Executives evaluating these structures should ask:

Would the customer still build the project without vendor-linked financial support?
Does the customer have contracted revenue capable of servicing the investment?
Who absorbs losses if compute prices decline?
Who carries the residual-value risk when new hardware arrives?
How much exposure is concentrated among a small number of counterparties?

Those questions do not assume that Nvidia’s reported revenue is improper.

They test something different: whether the underlying demand can survive independently of the financing structure that supports it.

That is a much more useful question for corporate boards than simply asking whether AI spending is rising.

AI Infrastructure Is Becoming an Asset Class—But Cash Flow Still Determines Its Value

AI compute can attract infrastructure capital, but its economic value ultimately depends on utilization, customer cash flow, financing costs and the residual value of the underlying equipment.

On August 10, Nvidia announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish financing platforms intended to mobilize more than $500 billion of third-party capital for AI infrastructure over time. Nvidia described the initiative as a way to make compute and full-stack AI infrastructure investable for global capital.

Reuters similarly reported that the initiative is intended to broaden access to Nvidia-based infrastructure while creating longer-duration, usage-linked investment opportunities for large asset managers and private-capital firms.

This changes the financial character of the AI buildout.

AI infrastructure is no longer simply a procurement decision for technology departments. Increasingly, it is becoming a capital-allocation question involving debt, equity, project finance, asset values and long-duration contracts.

Nvidia says its financing platforms will involve independent capital providers evaluating projects based on factors including customer demand, utilization, cash flow and residual value.

That independence is important.

If institutional investors can reject projects with weak economics, capital markets can impose discipline on AI expansion.

If financing instead allows marginal projects to continue because investors assume AI demand will remain permanently strong, the same mechanism could amplify risk.

The distinction will become increasingly visible as more AI infrastructure moves from corporate balance sheets into dedicated financing structures.

The Executive Framework: How to Test Whether AI Demand Is Economically Real

Sustainable AI demand is demand supported by productive utilization and credible customer cash flow rather than capacity announcements, vendor order growth or financing availability alone.

Measure utilization, not just booked capacity

AI infrastructure utilization measures whether deployed computing capacity is generating economically valuable workloads rather than simply existing as installed capacity.

A data center can be fully constructed and heavily financed without generating attractive returns.

For a CFO, the relevant question is not simply how many GPUs have been deployed. It is how intensively those GPUs are being used and what revenue or cost savings that utilization produces.

An AI developer might reserve large amounts of compute for training and inference. If demand falls below expectations, the infrastructure owner still faces power costs, depreciation, staffing, financing expenses and maintenance.

AI infrastructure economics dictate that announced capacity is not a sufficient benchmark because long-term value depends on utilization and the cash flow generated from productive compute demand.

Boards should monitor:

Actual utilization versus contracted capacity
Revenue per unit of compute
Customer concentration
Gross margin after power and financing expenses
Project payback periods
Firm contracts versus speculative demand

Separate contracted demand from independently financed demand

Contracted demand is stronger when customers have the financial capacity to honor commitments without relying on continuous fundraising or extraordinary vendor support.

This does not make vendor investment inherently problematic.

Strategic investment can accelerate useful infrastructure. The issue is whether the customer could continue funding the project if capital markets became less accommodating.

That distinction becomes especially important for AI startups and specialized cloud providers whose business models depend heavily on continued access to capital.

A project that works only while valuations rise presents a very different risk profile from a project supported by recurring customer revenue.

Independent AI infrastructure underwriting functions by evaluating the customer, utilization, cash flow and residual value of a compute project before capital is committed.

Stress-test counterparty concentration, cash flow and residual value

Residual-value risk matters because AI hardware can lose economic competitiveness as newer generations of processors arrive.

A project can generate strong cash flow today and still face uncertainty about the value of its equipment several years from now.

For boards, stress testing should include scenarios involving:

Lower customer growth
Lower compute prices
Reduced utilization
Higher financing costs
Faster hardware obsolescence
Customer defaults
Delayed project completion

The objective is not to predict which scenario will happen.

The objective is to determine whether the investment remains viable when several assumptions deteriorate simultaneously.

The Contrarian Risk: Strong Nvidia Revenue Does Not Automatically Prove the Entire AI Economy Is Healthy

Strong Nvidia revenue demonstrates extraordinary demand for Nvidia products, but it does not by itself prove that every downstream AI infrastructure investment will generate attractive returns.

This is the distinction executives should preserve.

Nvidia can record exceptional revenue because customers are making enormous infrastructure commitments today. Those customers may still require years to demonstrate that their investments generate sufficient revenue, productivity gains or cost reductions.

The two statements can be true at the same time:

Nvidia’s demand can be genuine.

Some customers can still destroy capital.

That is why Burry’s thesis is worth examining without treating it as a verdict on Nvidia.

Current market reporting shows how high the expectations have become. Yahoo Finance reports that Nvidia’s upcoming quarter is expected to produce roughly $92 billion of revenue, while Wall Street remains focused on whether the company can maintain its extraordinary growth trajectory.

The principal risk is a timing mismatch between infrastructure spending and monetization.

Capital can be committed immediately.

Data centers can be built over several years.

Customer adoption can take longer.

Profitable AI applications may take longer still.

For investment committees, that sequence matters more than a single quarterly revenue number.

Real-World Application: Nvidia’s $500 Billion Financing Push and the New Economics of AI Factories

Nvidia’s $500 billion financing initiative represents a shift from selling hardware toward helping create the capital structures required to deploy AI compute at industrial scale.

Nvidia’s August 10 announcement named Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR as partners in financing platforms designed to mobilize more than $500 billion of third-party capital.

The optimistic interpretation is straightforward.

AI infrastructure requires enormous upfront investment. Institutional capital can provide longer-duration funding and allow infrastructure operators to build capacity without carrying every asset directly on their own balance sheets.

The more skeptical interpretation focuses on incentives.

Nvidia benefits when more AI infrastructure is built because that infrastructure requires computing equipment and software. If Nvidia also participates in financing structures or provides financial support around infrastructure projects, the boundary between supplier and financial participant becomes more important to analyze.

That does not make the model inherently circular.

It makes the risk-transfer mechanism worth examining.

The distinction is particularly relevant given Nvidia’s recent Ohio data-center arrangements. Reuters reported that Nvidia agreed to provide up to $105 billion in guarantees connected to an OpenAI data-center project being developed by SB Energy, while Nvidia also planned to invest $1.5 billion in SB Energy. The facility is expected to reach 8 gigawatts of capacity, with 800 megawatts targeted for operation by 2028.

For investors, the important question is not simply how much infrastructure will be built.

It is who finances it, who owns it, who guarantees it and who bears the loss if utilization falls short.

That is the financial architecture behind the AI boom.

The Strategic Playbook for Managing AI Infrastructure Exposure

Leadership teams should treat AI infrastructure as a capital-allocation decision requiring the same financial discipline applied to acquisitions, factories and other long-duration investments.

1. Validate end-user cash generation

Measure whether AI customers are producing measurable revenue, cost reductions or productivity gains.

User growth and pilot activity are useful operating indicators. They are not substitutes for cash generation.

2. Measure utilization against commitments

Compare contracted capacity with actual workload demand.

A project can appear fully committed while still generating weak economic returns if customers are not consuming capacity at profitable rates.

3. Map counterparty dependencies

Create a financial map showing relationships among suppliers, customers, infrastructure operators, lenders and investors.

Counterparty concentration becomes more important when several entities depend on the same source of AI spending.

4. Stress-test financing assumptions

Model lower utilization, slower customer growth, higher interest costs, falling compute prices and delayed project completion.

The investment should remain viable without requiring every assumption to remain optimistic.

5. Monitor residual-value exposure

Determine who absorbs losses if hardware becomes economically obsolete faster than expected.

A financing structure can look attractive under optimistic residual-value assumptions and materially weaker under conservative ones.

6. Establish independent investment gates

Require major AI infrastructure projects to meet predetermined return thresholds.

Investment committees should be able to reject projects even when strategic partners strongly support them.

This framework also applies to enterprise AI implementation strategy. Companies should not measure AI success by how much computing capacity they acquire. They should measure whether that capacity produces measurable business outcomes.

Executive Outlook: The Next AI Test Is Not Demand—It Is the Quality of Demand

The next phase of the AI investment cycle will be judged less by whether demand exists and more by whether that demand produces durable cash flow at the scale required to justify today’s infrastructure spending.

Nvidia enters its August 26 earnings report with expectations that would have been extraordinary for almost any other technology company. Current Bloomberg consensus cited by Yahoo Finance calls for approximately $92 billion in revenue and $2.09 in adjusted EPS.

Burry’s position creates an unusual contradiction.

He expects the quarter to be “lights out,” yet he remains concerned about the financial architecture supporting the broader AI boom.

That contradiction is more informative than a simple bullish-versus-bearish debate.

A company can deliver exceptional quarterly results while the market around it is still taking increasingly aggressive financial risks.

The next question is not whether Nvidia can sell more chips.

It is whether the companies buying those chips can generate enough economic value to justify the infrastructure being financed around them.

That is why boards should look beyond headline capacity, purchase orders and supplier revenue.

They should examine utilization, customer cash flow, counterparty concentration, financing structure, residual value and return on invested capital.

The AI boom may not have a demand problem.

It may have a demand-quality problem.