[AI Bubble] Oracle’s AI Data Center Investment Hits a Red Light as Profitability Faces Scrutiny Amid Bubble Fears

Oracle’s $10 billion data-center financing disrupted by Blue Owl’s withdrawal
Financial institutions reassessing AI infrastructure as high-risk assets
Optimism around AI moderates as financial strain grows and valuation sorting begins

Warning signs are emerging across the global big-tech landscape as capital-intensive artificial intelligence (AI) infrastructure projects encounter mounting financial friction. Oracle’s flagship AI data-center initiative has been jolted after a key financing partner backed out, clouding a planned $10 billion fund-raising effort. The incident is intensifying market concerns that expectations underpinning AI-related investments—spanning semiconductors to hyperscale data centers—are fracturing under the weight of profitability concerns, financial strain, and mounting bubble warnings. Investors are now shifting from exuberant optimism to a markedly more cautious stance.

Oracle’s deteriorating financial position tightens lender requirements

On the 17th (local time), the Financial Times reported that Blue Owl Capital—Oracle’s long-standing investment partner—is poised to withdraw from a data-center project under construction in Saline Township, Michigan. The 1-gigawatt facility is being built to supply computing power to OpenAI. Blue Owl has been a key backer of Oracle’s previous data-center builds in Texas, New Mexico, and other states, investing via special-purpose vehicles that acquire and lease facilities back to Oracle.

According to the FT, lenders have imposed far stricter leasing and debt-structuring conditions in light of Oracle’s expanding AI-related capital expenditure and rising leverage, diminishing the project’s financial appeal for Blue Owl relative to earlier deals. Concerns over potential construction delays at the site also contributed to the decision. Blue Owl had initially considered arranging as much as $10 billion in financing and injecting significant equity capital. Blackstone is now being discussed as a potential alternative investor.

Oracle has been aggressively expanding its AI data-center footprint, relying heavily on debt financing—prompting rising investor unease over its financial burden. Credit analysts have also flagged the rapid build-up of liabilities. Oracle’s share price has fallen more than 40% from its September peak, with its bonds also weakening. In a statement, the company said, “The site developer selected the optimal equity partner after a competitive process, and Blue Owl was not chosen,” adding that final negotiations on the equity transaction “remain on track.”

AI-sector valuations enter bubble territory

As Oracle’s project falters, market vigilance around an “AI bubble” is intensifying. Goldman Sachs projects that up to $1 trillion will be invested in AI infrastructure over the coming years, yet questions whether this spending can realistically translate into corresponding revenue and profit. The firm warned that failure to monetize such large-scale investment would trigger massive depreciation charges, calling the current boom “a trillion-dollar gamble.” Morgan Stanley similarly noted that companies expanding through debt rather than internally generated cash—such as Oracle—will face increasingly unfavorable financing terms.

The fading of indiscriminate optimism around AI-related equities is accelerating a shift toward selective valuation. Investors are moving from momentum-driven buying to a more rigorous examination of profitability and financial resilience. Banks and lenders have begun reclassifying AI data-center projects from safe to high-risk assets, demanding higher interest rates, accelerated repayment schedules, and additional collateral. Tightening financing conditions are pushing up costs, intensifying profitability pressure across the sector. The withdrawal of major capital providers like Blue Owl reflects these deepening structural stresses.

Some analysts argue that parts of the AI and semiconductor ecosystem have already entered bubble territory. A 2017 framework by Harvard professors Robin Greenwood, Andrei Shleifer, and Yang You defines bubble conditions as: By these metrics, major AI-linked semiconductor names—including Nvidia, Broadcom, KLA, Lam Research, Micron Technology, AMD, and Monolithic Power Systems—meet bubble thresholds. Of the 29 S&P 500 constituents classified as bubble stocks under this methodology, 18 belong to AI-related industries—sharpening market anxiety.

Industry debate over the AI bubble has also intensified. Nvidia CEO Jensen Huang told employees that poor earnings are cited as bubble evidence and strong earnings are blamed for inflating it—calling the situation a “no-win trap.” He argued that today’s AI economy differs fundamentally from the dot-com era, noting Nvidia’s third-quarter revenue of $57 billion—up 62% year-on-year—with a net margin of 53%, in sharp contrast to the largely loss-making tech firms of the early 2000s.

AI’s limited monetization raises concerns of large-scale wealth destruction

A central question remains whether massive AI investment can yield sustainable profit. Market assessments increasingly point to underwhelming financial performance. OpenAI, for instance, has roughly 800 million global users, but only about 2% are paying customers. Meanwhile, model-training costs are soaring: ChatGPT-3 cost about $50 million to develop; GPT-4 about $500 million; and GPT-5 roughly $5 billion. Without a substantial increase in paid-user conversion, covering these escalating expenses may prove difficult.

Enterprise-level productivity gains also remain elusive. According to the Harvard Business Review, 95% of companies adopting AI have not achieved measurable financial returns. In a survey of 2,000 white-collar workers using AI tools, 77% reported decreased productivity. Even those who said AI “saved time” still experienced net productivity losses due to the need to review and correct AI-generated content—often verbose, repetitive, or inaccurate—resulting in the so-called “work slop” effect.

Alongside profitability concerns, warnings of AI-driven “wealth destruction” are growing louder. Howard Marks, chairman of Oaktree Capital—who predicted the dot-com crash—argued that the AI boom could trigger massive wealth erosion. He distinguishes between “inflection-point bubbles,” fueled by transformative technologies, and “mean-reversion bubbles,” inflated regardless of intrinsic value—yet believes both ultimately result in wealth destruction. In winner-take-all technological transitions, only a small minority benefit while most investors absorb losses after overestimating future gains.

Marks also underscored AI’s broader societal repercussions. He predicts AI could trigger sweeping job displacement, disproportionately affecting entry-level and low-skill workers, with losses potentially reaching millions. Should governments respond with universal basic income programs, he warns, taxpayer burdens and public debt could deepen. If society perceives AI as enabling a small elite of highly educated billionaires to eliminate millions of jobs, the resulting inequality could fuel political polarization and strengthen populist movements.