Inside the AI Bubble: $115B Cash Burn, Hidden Losses, Market Risk

We are in the early stages of a lasting AI revolution. Its capabilities will keep growing, driving unprecedented changes across all sectors and redefining work itself.

Artificial intelligence has become the most powerful investment narrative of the current market cycle. It dominates capital allocation decisions, corporate strategy, government policy, and investor psychology. Trillions of dollars in market value have been created on the assumption that AI will drive a new era of productivity, profitability, and economic growth.

Yet after months of analysis, conversations with market participants, and close examination of financial data, the MyEyze Finance Desk has reached a sobering conclusion: while the transformative potential of AI is undeniable, the AI bubble itself is very real—and it is larger, more deeply embedded in the financial system, and more fragile than headline market indicators imply.

Markets at Extremes, Even Before Counting AI Losses

U.S. equity markets are already trading near historic highs on nearly every valuation metric—price-to-earnings (P/E), price-to-book, buffet indicator, and the cyclically adjusted Shiller P/E (CAPE) ratio. These levels have previously been seen only before major corrections, such as in 1929, 2000, and 2007.

What makes the current situation more concerning is that a significant portion of AI-related excess is not yet visible in public markets. Many of the most aggressively valued AI companies—OpenAI, Anthropic, xAI (Grok), and Perplexity—remain private. These firms command valuations in the tens of billions(and even hundreds of billions) of dollars while generating little or no profit, and in some cases, they report growing multi-billion-dollar losses annually.

It is extremely rare to see multiple startups, all within a single sector, valued in the tens or even hundreds of billions of dollars before reaching meaningful profitability. In the AI space, companies like OpenAI, Anthropic, and xAI now occupy this stratosphere, a level of concentrated pre-profit valuation unprecedented in tech history, and one that amplifies systemic risk if expectations are not met.

If these companies were publicly listed and included in market indices, aggregate earnings would be materially lower, pushing market valuation multiples even higher. In effect, a portion of today’s AI-driven valuation inflation sits off-balance-sheet for public markets, masked by private funding rounds rather than public earnings scrutiny.

Circular Investment: Growth Feeding Itself

A defining feature of the current AI investment cycle—yet one seldom explained in straightforward terms—is the extent to which capital is circulating within a narrow cluster of labs, hyperscalers, and infrastructure vendors, rather than flowing directly into broad, diversified end-customer demand. In plain terms, AI companies are investing in each other, spending capital on technology from a small set of providers, and then those providers are being rewarded with higher valuations and further capital inflows. The result is a highly concentrated and self-reinforcing cycle of spending, revenue recognition, and valuation expansion.

Microsoft and OpenAI: A Strategic and Financial Feedback Loop

Microsoft has publicly disclosed cumulative funding commitments to OpenAI of approximately $13 billion, structured through multiyear investments and a profit-sharing arrangement that allows Microsoft to receive up to 49 % of OpenAI’s profits until a capped return is achieved. As part of these agreements, OpenAI is contractually tied to Microsoft Azure as its primary cloud provider for training and inference workloads. According to industry reporting, by late 2025 OpenAI had already incurred around $12 billion in compute and inference spending with Microsoft, indicating that capital raised from investors and partners is being recycled almost immediately into Azure consumption rather than broad monetisation through end-users. This spending flows back to Microsoft as revenue and helps justify its growing cloud infrastructure footprint.

Oracle’s $300 Billion Cloud Commitment

In 2025, OpenAI and Oracle announced a multi-year cloud deal reported at approximately $300 billion over five years beginning around 2027–2028.

Oracle, in turn, has committed tens of billions of dollars in capital expenditure for data centre build-outs anticipated to service these workloads. Separately, Oracle stated plans of up to $40 billion for Nvidia GPU purchases to power these environments. That means the OpenAI-Oracle contract not only creates future revenue for Oracle’s cloud business but also feeds downstream GPU demand for Nvidia.

Nvidia: The Fulcrum of AI Infrastructure

Nvidia’s financial results offer the clearest window into how concentrated AI monetisation has become. In the quarter ended October 2025, Nvidia reported $57 billion in total revenue, with 88 % coming from data-centre products—the segment most closely tied to AI workloads.

According to Nvidia disclosures and financial analysts:

1. About 50 % of data-centre revenue is attributable to large cloud service providers.
2. The top two customers alone accounted for 39 % of total quarterly revenue.
3. The top three customers together represented roughly 53 % of data-centre sales (approximately $21.9 billion).

These figures are extraordinary for a company of Nvidia’s scale, but they strongly indicate that AI spending is heavily concentrated among a very small set of hyperscalers and labs, rather than distributed broadly across businesses or individual consumers.

CoreWeave: Debt, Equity, and GPU Demand

CoreWeave, a specialised AI cloud provider, provides a textbook example of recursive capital flow in action. Over the last 12 months, CoreWeave has raised more than $12 billion through a combination of equity and debt specifically to build GPU-heavy data centres anchored on Nvidia hardware

Importantly, Nvidia is an equity investor in CoreWeave. CoreWeave markets GPU clusters built around the latest Nvidia accelerators and networking silicon, meaning Nvidia benefits twice:

1. As an equity holder benefiting from CoreWeave’s growth and valuation, and
2. As a product supplier when CoreWeave purchases accelerators and infrastructure.

Layered on top of this, hyperscalers such as Microsoft are among CoreWeave’s largest customers, meaning money spent by Microsoft on AI compute can pass through CoreWeave and circle back to Nvidia revenue.

xAI and GPU-Driven Capital Deployment

Elon Musk’s xAI, originally launched with significant public visibility, has raised approximately $12 billion through multiple rounds in 2024. Investors include leading venture firms (a16z, Sequoia, Valor) and strategic chipmakers Nvidia and AMD—the same companies supplying the GPUs xAI intends to deploy.

According to public reporting and industry estimates:

1. xAI has deployed a supercomputer with up to ~100,000 Nvidia H100 GPUs.
2. Plans for the latest capital tranche include the purchase of roughly another 100,000 Nvidia GPUs over time.

Here again, capital raised from investors—including Nvidia itself—is immediately converted into Nvidia hardware demand, creating a highly circular feedback pattern: Nvidia invests in xAI’s equity, then recognises revenue when xAI spends that capital on Nvidia accelerators.

The Circuitry of Circular Capital in AI

Taken together, these data points reveal a pattern:

1. AI labs and hyperscalers raise or commit tens to hundreds of billions in capital (e.g., Microsoft’s $13 billion to OpenAI; xAI’s $12 billion; CoreWeave’s $12 billion; OpenAI’s $300 billion cloud purchase commitments to Oracle).
2. That capital is deployed primarily into GPUs, cloud services, and specialised infrastructure provided by a narrow set of vendors (Nvidia, Oracle, Microsoft, Broadcom, and a few hyperscalers).
3. Those vendors report surging AI-linked revenues, which are then rewarded by markets with higher valuations and easier access to capital.
4. With stronger balance sheets, these vendors can reinvest equity or strategic capital back into AI startups and infrastructure partners, creating the next cycle of demand.

This feedback loop differs fundamentally from traditional revenue models where capital is spent on goods and services that generate external demand across a broad base of consumers or enterprises. In the current AI boom, much of the revenue recognition and valuation expansion occurs within tightly connected corporate and investor networks, rather than between AI firms and diversified paying customers.

Market Perception: Suspicion and Systemic Concerns

Unsurprisingly, this structure has drawn growing scrutiny. Some commentators describe parts of the ecosystem as reminiscent of circular financing or *Ponzi-like capital flow—not in the literal fraudulent sense, but in the reliance on continuous new capital to justify existing valuations and revenue growth. Critics argue that if one major participant were to falter—if a cloud commitment were scaled back or capital flows slowed—the interconnected revenue streams would quickly recalibrate, exposing just how narrow the base of “real” external demand has been.

This perception has already impacted sentiment: AI stocks have shown volatility even amid strong headline growth, and some institutional investors have openly questioned whether current valuation levels are justified by underlying cash flows.

The data clearly shows that AI investment today is highly intertwined, capital is being recycled among a small set of industry players, and revenue growth is concentrated rather than broad-based. While this has driven impressive headline figures and justified substantial capital flow in the short term, it also creates concentration risk and dependency on sustained confidence in a small number of firms and funding structures.

This circular capital model may work so long as investors remain confident, funding remains plentiful, and adoption continues to expand. But if any of those conditions weaken, the feedback loop that has amplified growth could just as quickly act in reverse, with outsized effects on valuations, debt structures, and market sentiment.

Profitability Remains Distant

Despite rapid revenue growth, most major AI model developers are far from sustainable profitability:

Current monetization is weak, with just 3% of the user base paying for AI services—a group that also demonstrates low loyalty by maintaining accounts with multiple providers. The introduction of enforced payments is likely to trigger customer churn as users migrate to free competitors.

Publicly traded AI-centric firms show a similar pattern. C3.ai, for instance, generates roughly $300 million in revenue while losing over $100 million annually.

Even projected revenue growth for Claude, OpenAI, and Grok assumes enterprise adoption scales exactly as forecast, compute costs decline, and monetisation converts free users into paying customers—all conditions that are uncertain.

AI Is Not a Monopoly Technology

One important point that is often overlooked is that AI is not a technology any single company can completely control and benefit from as a monopoly. Many U.S. companies—like OpenAI, Anthropic, Google, and Meta—as well as several Chinese firms, have advanced AI capabilities. As the quality of these models becomes similar across providers, it becomes harder for any one company to charge premium prices. While offering free or heavily discounted access has helped AI reach more users, it remains uncertain whether enough of these users will eventually pay for the services to make the massive investments worthwhile. Growing competition, both at home and abroad, could further reduce potential profits.

Physical and Infrastructure Constraints

Unlike past digital booms, AI is heavily constrained by physical realities:

1. Large data centres consume as much electricity as small cities, with high cooling and water demands.
2. Grid infrastructure upgrades are slow and expensive, delaying deployment in key regions.
3. Land acquisition and permitting challenges limit the rate of expansion.

These constraints introduce non-linear scaling costs, which markets are generally not pricing in, creating another layer of fragility.

Debt, Credit, and Hidden Transmission Channels

The AI boom is also deeply intertwined with credit markets:

1. Infrastructure providers have raised multi-billion-dollar debt to support AI workloads.
2. Pension funds, insurers, and sovereign wealth funds hold exposure through private equity or venture funds.

If AI investment slows, these exposures could transmit stress beyond the tech sector into public financial markets, including corporate credit and debt-dependent asset classes.

Accounting Optics and the Burry Warning

Some investors, including Michael Burry, have raised concerns about changes in depreciation schedules for AI infrastructure. Extending asset lives can temporarily boost reported profits, even though cash outflows remain high.

This accounting maneuver smooths earnings in the short term, masking the true economics of these highly capital-intensive businesses.

Early Cracks and Warning Signs

Some early warning signs are already visible:

1. The ongoing operation of several AI providers is contingent upon securing additional rounds of investment.
2. Oracle’s stock price has fallen sharply after AI-driven cloud expectations were not fully me, so has the stock price of core-weave.
3. AI executives themselves have acknowledged risk: overspending, long timelines to profitability, and uncertain enterprise adoption.
4. Analysts increasingly warn that the AI market is overheating, drawing parallels with the late 1990s before the dot-com crash.

Lessons From Past Bubbles

Historical patterns are instructive:

Opportunity Cost of Concentrated Capital

Capital has poured heavily into AI infrastructure, chips, and compute-heavy projects. Over one third of US gdp growth has come from AI related investments. Had the investment uniformly gone to renewable energy, healthcare innovation, small business, or manufacturing, returns could have been broader and more resilient. Concentration in AI increases the stakes: if expectations are unmet, the impact is magnified.

The Feedback Loop of AI Hype

AI investment today feeds on itself: rising valuations are justified by expected AI-driven growth, which encourages more spending and pushes expectations even higher. This creates a self-reinforcing loop. But when confidence drops, the adjustment can be sudden and uneven—small disappointments can trigger large market swings.

The cycle is likely to end not because AI fails, but because the story loses its momentum. When revenue, adoption, or profits don’t match the high expectations, investors may pull back quickly, causing valuations to shift—even if the technology itself continues to advance.

How This Differs From the Dot-Com Bubble

Unlike 2000, some of today’s market leaders are real, cash-generating companies, such as Nvidia, Microsoft, and Alphabet. That provides resilience.

Yet the risk is systemic: valuation concentration, debt exposure, and private market inflation. If expectations reset, the adjustment may be slower—but potentially broader, affecting markets, credit, and GDP growth.

What About AI Is Not a Bubble

While the investment frenzy may carry bubble-like characteristics, several aspects of AI are undeniably real and sustainable. The transformational impact of AI is tangible, with adoption spreading across everyday life and nearly every type of organization—banks, media firms, healthcare providers, and government agencies are all integrating AI into operations and services. The more a company develops AI capabilities, the more immediate and measurable benefits it gains, from automation to improved decision-making. Adoption is expected to accelerate further as AI becomes embedded in core business processes.

AI companies are also exploring new monetization avenues, such as in-chatbot advertising, premium services, and enterprise solutions, providing potential revenue streams beyond the current hype. The employment created by the AI sector—researchers, engineers, data scientists, and operational staff—is unlikely to vanish, as these skills will remain valuable across industries. Beyond this, AI-driven efficiency gains and cost savings for businesses, as well as innovation in sectors like healthcare, logistics, and climate solutions, indicate that the underlying technology will continue to generate real economic value, independent of market speculation.

Conclusion

The importance of AI is undeniable. As a lasting technology, it will keep improving and continue to transform all aspects of our work and society.

But markets have already priced in extraordinary success, flawless execution, and rapid monetisation. History shows that when belief outpaces earnings, even small disappointments can have outsized consequences. The AI bubble stems not from doubts about the technology's transformative power, but from the unrealistic and uncertain revenue expectations attached to it.

When and how this cycle ultimately turns remains uncertain. The bubble can grow bigger. Markets have a long history of sustaining optimism longer than fundamentals alone would justify.

This analysis is the first in a series. The MyEyze Finance Desk will continue to investigate the AI investment cycle, with upcoming articles focusing on its individual risk areas in greater detail.