AI: A Pandora's Box? Elon Musk forces the truth: Why the AI ​​hype is actually a bottomless financial pit

AI at its limits: Why Sam Altman is burning through vast sums of money – without any prospect of quick profits

Revealed in court: The secret 50 billion dollar bill that could break OpenAI

The hype surrounding artificial intelligence promises a golden future and a revolution in our working world – but behind the closed doors of the tech giants, a financial and environmental black hole is increasingly being revealed. A bitterly contested lawsuit between Tesla CEO Elon Musk and OpenAI's leadership, headed by Sam Altman, has now brought to light figures that shock even seasoned industry insiders: In 2026 alone, the company will burn through an unbelievable 50 billion US dollars on computing power – more than double its own revenue. While AI undoubtedly creates immense value in medicine and climate research, the global arms race of algorithms is devouring vast amounts of capital and driving energy consumption to astronomical heights. Added to this are unresolved societal dangers such as mass surveillance, deepfakes, and disinformation. Is the business model of artificial intelligence, in its current form, even sustainable, or will the environment and society ultimately foot the bill for Silicon Valley's blind obsession with growth? A stark look at the harsh reality behind the glittering AI facade.

Related to this:

• Tokenomics | When AI becomes more expensive than staff: The silent cost explosion of AI and what Managed AI can do about it

The foundation of the hype: What the courtroom revealed

When the promise of salvation becomes a bottomless pit – and the rest of the world foots the bill

A courtroom in Oakland, California, was supposed to decide on breach of contract and fraud. What it revealed instead was a sobering look into the economic underbelly of the largest technology project of our time. Greg Brockman, co-founder and top manager of OpenAI, confirmed under oath in May 2026 a figure that had not been publicly known until then: In that year alone, his company would spend around 50 billion US dollars – almost 43 billion euros – on computing power alone. This sum more than doubles OpenAI's total annual revenue, which in 2025 was around 13 billion dollars, with an annualized rate of approximately 20 billion dollars by year-end.

The context of this statement is intriguing: Brockman was in the witness stand because tech billionaire Elon Musk – himself a co-founder and backer of OpenAI – had sued the company. Musk accuses Sam Altman and Brockman of transforming OpenAI from a non-profit research organization into a for-profit company, contrary to original agreements, and thus effectively stealing a charitable organization. OpenAI's leadership rejects this, arguing that establishing a for-profit entity was simply unavoidable in order to raise the necessary billions in investment – ​​an argument that seems frighteningly plausible given the revealed figures.

What this process unintentionally accomplishes is the economic disenchantment of an entire industry. For what is being debated in the courtroom as a legal dispute over founders' ideals is, in reality, the exposure of a systemic contradiction: Artificial intelligence in its current form is not a scalable product with healthy margins – it is an industrial machine that burns through capital at a breathtaking pace.

Billions in, even more billions out: The cost structure behind the AI ​​facade

To fully grasp the economic absurdity, a closer look at the figures is worthwhile. OpenAI generated $4.3 billion in revenue in the first half of 2025 – while simultaneously incurring a net loss of $13.5 billion. Operating losses alone amounted to $7.8 billion during this period, with research and development accounting for $6.7 billion. In the third quarter of 2025, quarterly losses ballooned to approximately $12 billion.

At the same time, OpenAI has committed to infrastructure investments of more than $1.4 trillion over the next eight years—an average annual investment of $175 billion, exceeding Google's total annual revenue. For the coming years, OpenAI has announced investments of more than $1 trillion in AI infrastructure. Analysts at the investment bank HSBC predict that OpenAI could reach annual revenue of around $214 billion by 2030, but the cost of rented computing capacity alone is projected to reach $792 billion by then, and a staggering $1.4 trillion by 2033. This means that even in the most optimistic growth scenario, infrastructure costs will consume revenue.

This pattern isn't a flaw in the system—it's its current operating state. The company also pays 20 percent of its total revenue directly to Microsoft, with whom it maintains a deep strategic and financial partnership. CEO Sam Altman has publicly stated that OpenAI doesn't expect to be profitable until 2029. What this means for the company's valuation—most recently trading at around $300 billion—is a question that financial markets have so far taken with surprising equanimity.

The hardware spiral: Billions spent on chips that will be scrap in three years

Behind the abstract cost figures lies a very concrete material reality: AI data centers are highly specialized, extremely capital-intensive facilities whose core consists of graphics processors that depreciate at a rate that blows any conventional investment plan out of the water.

A modern high-end GPU for AI applications currently costs between €25,000 and €40,000 per card. Nvidia's latest Blackwell Ultra architecture systems are driving costs even higher: cloud rental prices for these chips range from $4.95 to $18 per hour. Analysts expect AI processors to become technologically obsolete after three to five years, as development cycles for chips and AI accelerators now range from 12 to 18 months. Financial investor Michael Burry even warns of a realistic lifespan of only two to three years. The implications for a data center that has invested billions in hardware are obvious: depreciation is enormous, and those building today may be buying outdated equipment tomorrow.

Interestingly, this picture isn't quite as bleak as it initially appears. Older GPU generations – like the Nvidia H100 – are losing ground to training the latest models, but can still be used economically for less computationally intensive inference tasks. This creates a multi-tiered ecosystem in which hardware is passed on like in a relay race – a gradual decline in value rather than an abrupt loss. Nevertheless, the fundamental economic problem remains: The rapid pace of innovation in the semiconductor market makes any long-term planning difficult and forces companies into a constant reinvestment cycle, a characteristic traditionally associated with capital-intensive technology projects – but rarely to this extent.

The energy hunger: An environmental bill that is only just beginning to run

The financial costs only tell half the story. The other half concerns energy consumption – and this has reached a scale far exceeding industrial dimensions, with direct geopolitical and environmental consequences.

According to data from the International Energy Agency (IEA), global electricity consumption by data centers recently stood at 415 terawatt-hours, representing approximately 1.5 percent of global electricity consumption. This figure is projected to more than double to around 945 terawatt-hours by 2030 – equivalent to Japan's total annual electricity consumption today. The AI-specific component is the primary driver of this growth: According to calculations by the Öko-Institut (Institute for Applied Ecology) commissioned by Greenpeace Germany, the global electricity consumption of AI data centers will increase elevenfold between 2023 and 2030 – from 50 billion kilowatt-hours to approximately 550 billion kilowatt-hours.

The electricity consumption of AI data centers increased by 50 percent in 2025 alone. Market research group Gartner predicts that the electricity consumption of AI-optimized servers will almost quintuple by 2030 – from 93 terawatt-hours in 2025 to 432 terawatt-hours. Their share of total data center consumption will thus rise from the current 21 percent to 44 percent. A single AI-focused data center consumes, on average, as much electricity as 100,000 households – particularly large facilities currently under construction could require twenty times that amount.

In Germany alone, the AI-specific connected load of data centers will quadruple by 2030, from 530 megawatts to approximately 2,020 megawatts. The combined energy consumption of all German data centers will grow to around 32 terawatt-hours per year, which would then account for roughly six to seven percent of Germany's total electricity consumption. Added to this is the water requirement for cooling, which is projected to nearly quadruple to 664 billion liters by 2030, as well as up to five million tons of additional electronic waste. Therefore, anyone discussing the costs of AI must also address its environmental footprint – and this footprint is substantial.

Musk versus Altman: A dispute over money, power, and the legacy of an idea

The trial that revealed the $50 billion figure shines a harsh light on the power dynamics and contradictions simmering at the heart of the AI ​​industry, going beyond mere cost figures. Elon Musk co-founded OpenAI in 2015 and was involved in the early initial investments. He left the company in 2018 following internal disputes. Today, he is the plaintiff in a lawsuit, accusing Altman and Brockman of transforming a non-profit research organization into a profit-driven business model.

The accusations are multifaceted: Musk's lawyers, during questioning, alleged that Brockman had personal motives and referred to a diary entry in which he mused about a path to a billion-dollar fortune. Brockman, in turn, countered by accusing Musk of seeking full control of the profit-oriented part of OpenAI because he allegedly needed $80 billion to build a city on Mars. What reads like a satire on Silicon Valley hubris is, in reality, a serious legal battle that raises the question of who owns technology and whose interests it serves.

Musk is anything but a neutral player in this matter. Since founding his own AI company, xAI, he has been a direct competitor of OpenAI, and the courts have repeatedly expressed doubts about the impartiality of his lawsuits. In February 2026, a US federal judge dismissed another Musk lawsuit—this one alleging the theft of trade secrets—as insufficient. Sam Altman openly described Musk's actions as an attempt to stifle a competitor. With approximately 800 million users worldwide and annualized revenue exceeding $20 billion, OpenAI had recently achieved a level of societal relevance that extends far beyond the context of a startup dispute.

The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business - Image: Xpert.Digital

Xpert.Digital is a data-driven B2B industry hub led by Konrad Wolfenstein . The company acts as an external, quasi-in-house solution for industrial partners, closing operational gaps in marketing, content, and sales – without requiring additional resources on the client side.

More information here:

• The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business

The Dark Mirror: Surveillance, Weapons, and the Commercialization of Privacy

Beyond the billion-dollar calculations lies a societal question that economic analyses alone cannot answer: What is this technology actually used for? And whose interests does it serve?

In China, citizens can hardly escape state surveillance through AI. Over 700 million cameras record everything day and night, biometric data is stored on state servers, and this data is not only used to control the population but is also actively traded. What is extremely visible in China is also making its way into Europe. In Germany, the federal government is increasingly using biometric surveillance measures, parts of the German police authorities are working with the controversial Palantir analysis software, and the Berlin Senate has announced plans to deploy AI-based behavioral scanners in public spaces. The EU Commission also planned chat monitoring measures that would involve the automated scanning of private messages, which data protection experts considered the first step towards a mass surveillance infrastructure.

In the military sphere, AI is no longer a vision, but an operational reality. The German Armed Forces already have AI-based combat modes in the Puma infantry fighting vehicle and the F125 frigate. AI systems are used for reconnaissance, situational awareness, logistics, and navigation of autonomous combat systems such as drones. The alarming aspect of fully autonomous weapon systems is that they could complete the entire decision-making cycle—from target identification to attack—without human oversight. AI experts have been warning for years that such systems could cause conflicts to spiral out of control: enemy systems could be misjudged, triggering automated retaliatory strikes.

In the civilian sphere, the combination of surveillance infrastructure, algorithmic behavioral control, and deepfake technology deserves particular attention. According to studies, 96 percent of all deepfakes are visual attacks with pornographic content—a form of digital sexualized violence that has become trivially producible through AI technology. AI-generated disinformation jeopardizes elections, promotes social polarization, and undermines trust in democratic institutions. According to a 2023 survey by the TÜV Association, 51 percent of German citizens agreed with the statement that AI technologies pose a threat to democracy. Consumer purchasing behavior is analyzed, predicted, and manipulated by algorithmic systems—a form of behavioral control where the lines between recommendation and manipulation are blurred.

Related to this:

• AI boom at your expense? Growing electricity demand and rising electricity prices: AI data centers vs. the power grid

The counterweight: Where AI actually creates value

A balanced economic analysis must also examine the other side of the equation, because AI is not solely an instrument of control and capital destruction. There are areas of application where the technology generates undeniable social value.

In medicine, progress is concrete and measurable. Microsoft's AI Diagnostic Orchestrator solved complex medical cases with an accuracy of 85.5 percent – ​​compared to an average of 20 percent for experienced physicians. In Germany, 18 percent of hospitals already use AI technologies, a remarkable doubling since 2022. AI algorithms for the early detection of breast cancer or the identification of lung metastases have reached clinical maturity. 43 percent of surveyed acute care hospitals are already optimizing operating room capacity and bed occupancy with predictive algorithms. The global market for AI-supported diagnostics, estimated at US$1.55 billion in 2025, is projected to grow to almost US$19 billion by 2037.

In climate research and epidemiology, AI is performing work that simply surpasses human capabilities: weather forecasts with unprecedented resolution, reconstruction of climate data, and wastewater-based epidemiology for the early detection of disease outbreaks. Efficiency gains are also emerging in logistics, energy efficiency, and materials science, which can generate real economic and environmental savings in the long term.

The problem isn't the non-existence of these applications – it lies in the structural imbalance. The socially valuable applications of AI represent a comparatively small portion of the resources actually allocated and the computing power actually utilized. The overwhelming majority of AI computing power flows into consumer applications, entertainment generation, algorithmic targeting, and the competition between AI assistants for an ever-expanding user base.

The structural contradiction: Why the business model doesn't work

A company that spends more than twice its revenue on computing power defies classical economic logic. OpenAI exemplifies a phenomenon that characterizes the entire AI industry: the subsidizing of growth with capital in anticipation of future profit dominance. The model is not new—it was familiar from the early internet economy and the sharing economy phase with Uber and Airbnb. However, the scale of this practice in the AI ​​industry is unprecedented.

The real economic paradox is this: the more people use AI services, the more computing power is required, the higher the costs rise – and the further profitability is pushed into the future. OpenAI itself cited available computing power as the current limiting factor for revenue growth in January 2026. Growth and cost scaling are inextricably linked in this industry. This means that whoever sells more needs proportionally more capital – a model that will remain structurally dependent on external financing as long as no technological breakthroughs radically improve energy efficiency.

Whether such a breakthrough will occur remains to be seen. The Chinese AI model DeepSeek demonstrated at the beginning of 2025 that comparable performance is possible with significantly less energy consumption – a finding that was met with some shock in the West. But even if efficiency increases, historically every efficiency gain in information technology has led to an expansion of usage that more than offsets the savings – a phenomenon known as the Jevons Paradox. In a growth-driven industry, greater efficiency doesn't mean less resource consumption, but rather more applications at lower marginal costs.

A bottomless competition: The AI ​​arms race and its systemic risks

OpenAI is not alone. The industry is engaged in an arms race structurally reminiscent of the Cold War arms race – with the difference that there are no external brakes. Google with Gemini, Anthropic with Claude, Elon Musk's xAI with Grok, and Chinese players like Baidu and Alibaba are all in a capital competition where deciding to slow down would be tantamount to being out of the game.

The consequence is a market where collective investment exceeds what would be economically sensible – because the fear of losing their competitive position outweighs concerns about their own balance sheets. This capital comes from sovereign wealth funds, pension funds, and strategic investors, who are themselves betting on the future dominance of AI. Should this bet fail – or should profitability fail to materialize structurally – the consequences for a large number of investors would be significant.

What makes the dispute between Musk and OpenAI particularly revealing in this context is the exposed question of governance: Who actually controls this powerful and resource-intensive technology? OpenAI was originally founded as a non-profit organization conducting research in the interest of humanity. Today, it is a company with a trillion-dollar infrastructure commitment that, by its own admission, does not expect to make a profit until 2029, yet is valued by global investors at a level that suggests future market dominance. The gap between the original founding vision and today's reality is vast.

A sober overall assessment

Pandora's box is a fitting metaphor – but an incomplete one. In the myth, all the world's evils escape from it, while only hope remains at the bottom. With artificial intelligence, the picture is more complex: the hopes are real and demonstrable, but they compete with very concrete and very high costs – financial, environmental, and social.

What the lawsuit against OpenAI and the $50 billion in computing costs it revealed demonstrates economically is this: The technology is in a state where its societal costs—in the form of energy consumption, misallocation of capital, surveillance infrastructure, and democratic risks—are far less accurately captured and priced than its commercial returns. No market mechanism exists that fully internalizes these negative externalities: Neither the carbon emissions of data centers nor the societal damage caused by disinformation and surveillance appear in the profit and loss statements of OpenAI, Google, or Microsoft.

As long as this remains the case, the rational calculation of every market participant will always lead to expansion and growth – at the expense of all those who were not presented with this bill but will ultimately pay the price. This is the real economic core of the problem. It's not a question of whether AI has meaningful applications – it undoubtedly does – but rather whether the way it is developed, financed, and deployed serves society or primarily the capital that has invested in it.

☑️ 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 simply call me at +49 7348 4088 965. My email address is [email protected]:or

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

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