Nvidia's $68 billion quarter: Triumph or illusion? Why Nvidia's incredible numbers remind experts of the dot-com crash

When a chip manufacturer earns more than entire economies, the question arises: Who ultimately pays the bill?

On February 25, 2026, the financial markets received the most important earnings report of the year. Nvidia reported quarterly revenue of $68.1 billion, a 73 percent increase year-over-year and a 20 percent increase compared to the previous quarter. Net income climbed to $43 billion, with a remarkable gross margin of 75 percent. At first glance, this appears to be a flawless balance sheet. However, a closer look reveals a web of circular financing flows, questionable investment chains, and an AI infrastructure boom built on bets on the future. This analysis contextualizes the quarterly figures, sheds light on the dark side of Nvidia's economics, and poses the central question: Is this growth sustainable, or is it the most expensive gamble in technological history?

Related to this:

• The most important stock market event of the year: How investors can protect themselves now from the Nvidia stress test – fear of the DeepSeek shock?

The record quarterly report in detail

Nvidia once again exceeded Wall Street expectations in the fourth quarter of fiscal year 2026. Revenue of $68.1 billion was approximately $1.9 billion above the analyst consensus of $66.2 billion. Adjusted earnings per share reached $1.62, an 82 percent increase year-over-year. For the entire fiscal year 2026, which ended in January 2026, revenue totaled $215.9 billion, with net income of $120 billion. Nvidia reported operating income of approximately $130 billion, while paying about $21.4 billion in taxes.

The difference of approximately $11 billion between operating profit and net income is striking, stemming primarily from unrealized gains on investments, including one in Intel. These book profits exist essentially only on paper and contribute nothing to the company's core business. Nvidia spent $18.5 billion on research and development for the full year and $4.6 billion on sales and administration. The company poured $41.1 billion into share buybacks and dividends.

The outlook for the first quarter of fiscal year 2027 sets the target at $78 billion in revenue, plus or minus two percent. This would correspond to sequential growth of approximately 15 percent. Crucially, this forecast does not include a single dollar of revenue from the Chinese market, where Nvidia is currently prohibited from selling AI accelerators due to US export restrictions.

The data center as an insatiable money machine

The data center division has long since become the dominant engine of the Nvidia empire. In the fourth quarter, it achieved record revenue of $62.3 billion, a 75 percent increase year-over-year and a 22 percent increase compared to the previous quarter. This means that more than 91 percent of total revenue comes from this single segment. The networking subgroup, which includes network processors and switches, grew particularly dynamically, increasing sequentially by 34 percent to almost $11 billion. Year-over-year, Nvidia recorded 263 percent growth in this segment. AI accelerators alone accounted for $51.3 billion, 19 percent more than in the previous quarter and 58 percent more than in the same period last year.

Hyperscalers like Amazon, Microsoft, Google, and Meta accounted for slightly more than 50 percent of data center revenue, but recent growth has been led by other customer groups, including AI model developers, enterprises, supercomputer operators, and sovereign states. Chief Financial Officer Colette Kress emphasized that Nvidia will surpass its total chip production estimate of $500 billion in 2026, which it projected last fall, and expects sequential revenue growth throughout the fiscal year.

The remaining business segments, however, play a subordinate role. Gaming revenue fell seasonally by 13 percent to $3.7 billion, now accounting for only 5.5 percent of total revenue. Kress also warned of supply bottlenecks that will impact gaming in the current and following quarters. While the professional visualization segment grew to $1.3 billion, and the automotive segment remained stable at $604 million, both remain marginal in Nvidia's overall financial results.

The stock market's reaction: Euphoria with a bitter aftertaste

The market reaction to Nvidia's figures revealed a remarkable ambivalence. Immediately after the release, the stock rose by up to 3.5 percent in after-hours trading. However, during the subsequent analyst conference, the gains evaporated and at times even turned negative. At the close of after-hours trading, the stock was up by approximately 0.7 percent.

For a company that had exceeded all expectations, this was a remarkably subdued reaction. Gene Munster of Deepwater Management noted that the analyst conference lacked the usual enthusiastic drumbeat of the message that Nvidia was only at the beginning of a long upswing. Instead, Jensen Huang relied more on data than his typical barrage of optimistic adjectives. CNBC's Jim Cramer, in turn, blamed the comments about the Chinese market for the share price decline. Nvidia's explicit statement that it currently generates no revenue in China raised questions among investors about the size of the addressable market.

Compared to hyperscalers Amazon, Google, Meta, and Microsoft, whose shares all slumped after their quarterly reports, Nvidia's share price performance was still a relative success. The core problem can be summed up in a simple formula: the good news was already priced in, and the bar for positive surprises at Nvidia is now set so high that even 73 percent revenue growth is considered a given.

Circular buying and round-trip financing: The hidden machinery

Behind these impressive figures lies a financing model that is increasingly coming under fire. Nvidia invests heavily in companies, which in turn use this money to buy Nvidia chips. Critics call this practice round-tripping or circular financing, a method in which capital flows in a circle, generating revenue at each stage.

There are now numerous and well-documented concrete examples. Nvidia plans to invest up to $100 billion in OpenAI. OpenAI will use this money to build AI data centers equipped with Nvidia hardware. The planned data centers could cost between $500 and $600 billion, of which $350 to $450 billion is budgeted for Nvidia equipment. To facilitate the financing, OpenAI will lease the chips from Nvidia instead of purchasing them. OpenAI's CFO, Sarah Friar, has herself admitted that the majority of the invested money will flow back to Nvidia.

A similar pattern can be seen at CoreWeave, a cloud provider in which Nvidia has held a stake since 2023. In September 2025, Nvidia committed to purchasing up to $6.3 billion of CoreWeave's unsold cloud capacity, thus acting as a financial safety net for a company that primarily leases Nvidia hardware. CoreWeave also opened a $2.3 billion credit line secured by Nvidia chips to purchase additional Nvidia chips.

Nvidia leased back 18,000 of its own GPUs from cloud provider Lambda for $1.5 billion—chips that Nvidia had previously sold to Lambda. Nvidia is now generating revenue and will cover the leasing costs in the future. The recent deal with Nokia follows the same pattern: Nvidia is investing $1 billion in the Finnish company, which in return has committed to purchasing Nvidia chips.

These entanglements are disturbingly reminiscent of practices from the dot-com era. Lucent Technologies aggressively loaned its own customers to finance the purchase of Lucent equipment and booked the value of these loans as revenue. Its stock price soared to dizzying heights before Lucent's revenue plummeted from $38 billion in 1999 to $8 billion in 2006. Nortel Networks lent over $7 billion to startups, often interest-free and unsecured, and went bankrupt when those customers failed. The most extreme cases, Global Crossing and Qwest, which sold each other network capacity and booked it as revenue, ended in SEC fraud cases.

Nvidia is not Lucent, and today's situation is not the same as the dot-com bubble. Nvidia does indeed produce by far the most powerful AI accelerators in the world, and the demand for AI computing power is real. But the line between legitimate vendor financing and problematic round-tripping is blurry. Jay Goldberg of Seaport Research Partners, the only Wall Street analyst with a sell recommendation for Nvidia, aptly described the situation: It's not a circle, but rather a dodecahedron—a shape that becomes more and more like a circle the longer the game goes on.

Our US expertise in business development, sales and marketing - Image: Xpert.Digital

Industry focus areas: B2B, digitalization (from AI to XR), mechanical engineering, logistics, renewable energies and industry

More information here:

• Expert Business Hub

A thematic hub offering insights and expertise:

• Knowledge platform covering global and regional economies, innovation and industry-specific trends
• A collection of analyses, insights, and background information from our key areas of focus
• A place for expertise and information on current developments in business and technology
• A hub for companies seeking information on markets, digitalization, and industry innovations

The AI ​​investment boom: Hype or real need?

The central question behind Nvidia's record figures is: Does the massive demand for AI chips correspond to a real economic need, or is it a self-perpetuating speculative investment frenzy? The figures regarding the investment readiness of the major technology companies are certainly breathtaking. The five major hyperscalers (Amazon, Microsoft, Google, Meta, and Oracle) are expected to spend more than $600 billion on infrastructure in 2026, an increase of 36 percent compared to 2025. Approximately 75 percent of that, or around $450 billion, will flow directly into AI infrastructure.

Capital intensity has reached historically unprecedented levels. Each of the four largest hyperscalers now spends more than $100 billion annually on infrastructure, with investment ratios as a percentage of revenue soaring to between 45 and 57 percent. To finance these expenditures, hyperscalers raised approximately $108 billion in debt in 2025 alone, with forecasts projected total debt of $1.5 trillion in the coming years. Big tech companies have already issued $100 billion in bonds for 2026, and investors have demanded record levels of credit default swap protection against bond losses. These figures raise an uncomfortable question: What return on investment are these AI investments actually generating? While cloud providers' AI services are growing at an impressive 150 percent annually, their investment ratios of over 22 percent of revenue are well above the historical average of 11 to 16 percent. BCA Research points out that technology-related investments have reached a record high of 7.2 percent of US GDP, even surpassing the peak of the dot-com bubble at 6.4 percent.

Some analysts, including the investment firm Man Group, consider the current debt-fueled AI investment cycle one of the most intense in modern economic history. Goldman Sachs, on the other hand, predicts that hyperscalers will invest a total of $1.15 trillion between 2025 and 2027, and even sees $200 billion in upside potential compared to current estimates. The truth likely lies somewhere in between: The demand for AI computing power is undoubtedly real, but the pace of investment has developed a momentum of its own, no longer driven solely by concrete demand figures.

The DeepSeek Factor: Paradigm shift or false alarm?

In January 2025, the Chinese AI startup DeepSeek sent shockwaves through the financial markets. The company had unveiled a powerful AI model, reportedly trained with a fraction of the computing power and at a fraction of the cost of its Western competitors. Nvidia's stock plummeted 17 percent in a single day, wiping out nearly $600 billion in market value.

Nvidia responded immediately, calling DeepSeek's model a remarkable advancement in AI and a prime example of test-time scaling. The company argued that more efficient models would actually increase demand for computing power because they would make AI applications accessible to a wider range of users. This assessment proved accurate. Chinese tech giants like Tencent, Alibaba, and ByteDance increased their orders for Nvidia's H2O chip, which was specifically designed for the Chinese market while complying with US export controls.

A year later, the seven model updates DeepSeek has released since then have not triggered a comparable market reaction. The initial panic was a one-off repricing of cost assumptions for AI training, not a structural shift in chip demand. However, in February 2026, Reuters reported that DeepSeek was withholding its latest model from US chipmakers, instead granting early access to Chinese suppliers like Huawei. This suggests a strategic attempt to mask its reliance on American hardware.

Nvidia CEO Jensen Huang countered at the analyst conference by arguing that the demand for AI computing power is a thousand times greater than for traditional computing power, and that the need far exceeds $700 billion. The world has entered the age of agentic AI, in which computing power is directly translated into revenue for customers.

The chip roadmap: Blackwell delivers, Rubin accelerates

Nvidia's technological dominance is beyond question. The current Blackwell platform generated double-digit billion-dollar revenues in the reporting quarter, driven by demand that CEO Huang described as exceptional. A gross margin of 75 percent indicates that Nvidia continues to sell its hardware at scarcity prices; a server with 72 Blackwell chips costs around three million US dollars, or approximately 40,000 US dollars per GPU.

The next generation of chips, Rubin, surprised the industry with its accelerated timeline. At CES 2026, Jensen Huang announced that the Rubin platform has already entered full production, with volume shipments starting in the second half of 2026. Analysts had only expected availability in early 2027. Nvidia has thus compressed the development cycle from the usual 24 to 30 months to 18 months. The Rubin platform comprises six new chips and promises to reduce the cost per inference token by a factor of ten compared to Blackwell.

The shift to an annual release cycle (Blackwell 2024, Blackwell Ultra 2025, Rubin 2026) is a strategic move that forces competitors like AMD and Intel into a permanent generational lag. However, capacity constraints exist in TSMC's N3 manufacturing process and with HBM4 memory, which could limit Rubin production in 2026 to an estimated 200,000 to 300,000 GPUs.

CFO Kress explained that it was too early to forecast revenue for the Rubin platform, but emphasized that cloud providers such as AWS, Microsoft Azure, Google Cloud, and Oracle Cloud had already secured capacity. Rubin's actual revenue contribution will not materialize until the second half of 2026.

Substance or speculation: The crucial question of AI economics

The question of whether Nvidia's figures are based on solid foundations or future bets cannot be definitively answered because both elements are simultaneously present. On the one hand, the operational performance is impressive. Revenue growth translates almost directly into net profit, gross margins are stable at 75 percent, and cash flow is substantial. Nvidia sells real products to customers with purchasing power who actually use those products.

On the other hand, a significant portion of demand is fueled by circulating investment capital, debt financing on an unprecedented scale, and the expectation that AI applications will generate revenue in the coming years to justify these investments. OpenAI forecasts a cash burn of $115 billion by 2029. So-called neoclouds like CoreWeave and Lambda partially finance their Nvidia hardware through GPU-secured loans, a new debt market of over $10 billion that has only existed since 2024. The notion that GPUs can serve as long-term collateral for loans raises fundamental valuation questions, because the rapid pace of technological development, with new chip generations released annually, means that the resale value of this hardware is declining rapidly.

Jensen Huang defended the investment climate by arguing that computing power translates directly into revenue for customers and that hyperscalers can easily finance their investments through growing cash flows. However, this argument is circular: more chips generate more computing power, which enables more AI applications, which in turn generate more demand for chips. As long as the cycle works, everyone benefits. But if it turns out that the AI ​​applications do not generate the expected revenue, an investment volume of several hundred billion US dollars is at stake.

Nvidia's valuation, with a forward P/E ratio of around 26 and expected revenue growth of over 60 percent, doesn't seem overly ambitious at first glance. However, this metric reflects the expectation that the current growth rate will continue for years to come, an assumption that depends on the stability of a highly complex financial structure.

Nvidia's position in the global power game

Nvidia's financial results are far more than just a company balance sheet. They are a barometer for the global AI race and the technological dominance of the USA. The fact that Nvidia is not expecting any revenue from Chinese data centers in the current quarter underscores the geopolitical dimension of the business. US export controls have largely closed Nvidia off the world's second-largest market for AI chips, while China is working intensively on alternatives.

DeepSeek's recent decision to deny US chipmakers access to its latest models and instead partner with Huawei signals a deliberate technological decoupling. At the same time, there is speculation that DeepSeek may have trained its latest model with Nvidia's most advanced Blackwell chip, in violation of export regulations.

Nvidia's strategy of steadily expanding its chip production with contract manufacturer TSMC while simultaneously accelerating its annual innovation cycle aims to maintain its technological lead over competitors like AMD, Intel, and the in-house chip developments of hyperscalers. The growing diversification of its customer base beyond the major cloud providers, encompassing sovereign states, research institutions, and enterprises, reduces concentration risks and broadens its revenue base.

The trillion-dollar bet on tomorrow

Nvidia's fourth quarter of fiscal year 2026 marks a turning point, not because the numbers were disappointing, but because they starkly reveal the inherent tensions of the AI ​​boom. The company is delivering operational results unprecedented in the history of the semiconductor industry. At the same time, it operates at the center of a funding cycle where venture capital, debt financing, and technological euphoria generate a growth dynamic that is, in part, self-reinforcing.

The key variables for the coming quarters are clearly defined. First, will the Rubin platform adhere to the accelerated timeline and deliver the expected performance leap? Second, can the hyperscalers translate their unprecedented investments into measurable returns, or will the capital intensity begin to weigh on cash flows and margins? Third, how will the Chinese market perform under tightened export controls and increasing technological self-sufficiency? And fourth, at what point will investors recognize that circular financing structures are not a sustainable growth model?

Nvidia has silenced all short-term doubts with impressive numbers. But the history of technology markets teaches us that the most dangerous turning points occur precisely when the numbers look their best. With $600 billion in annual hyperscaler investments, GPU-backed credit markets, and chip valuations that influence the fate of entire economies, the AI ​​economy has reached a scale where a stumble would have systemic consequences. Nvidia's $68 billion quarter is undoubtedly a triumph of engineering and market position. Whether it is also a triumph of sustainable value creation will only become clear in the coming years.

☑️ Our business language is English or German

☑️ NEW: Correspondence in your native language!

Konrad Wolfenstein

I and my team are happy to be available to you as your personal advisor.

You can contact me by filling out the contact form here or simply call me at +49 7348 4088 965. My email address is: [email protected]

I'm looking forward to our joint project.

☑️ SME support in strategy, consulting, planning and implementation

☑️ Creation or realignment of the digital strategy and digitization

☑️ Expansion and optimization of international sales processes

☑️ Global & Digital B2B trading platforms

☑️ Pioneer Business Development / Marketing / PR / Trade Fairs

Benefit from Xpert.Digital's extensive, five-fold expertise in a comprehensive service package | R&D, XR, PR & Digital Visibility Optimization - Image: Xpert.Digital

Xpert.Digital possesses in-depth knowledge across various industries. This allows us to develop tailored strategies precisely aligned with the requirements and challenges of your specific market segment. By continuously analyzing market trends and monitoring industry developments, we can act proactively and offer innovative solutions. The combination of experience and expertise generates added value and provides our clients with a decisive competitive advantage.

More information here:

• Benefit from Xpert.Digital's 5 areas of expertise in one package – starting from just €500/month