Oracle plunges 40%, will AI over-infrastructure bring down tech giants?
TechFlow Selected TechFlow Selected
Oracle plunges 40%, will AI over-infrastructure bring down tech giants?
Success brings joy to all, failure could result in losing everything.
Navigating Web3 tides with focused insightsHolding massive AI infrastructure-related orders is no longer enough to "protect" a company.
Oracle, with $500 billion in orders, has seen its stock price drop 40% from its September peak. Broadcom currently has around $73 billion in backlog for AI products, yet its stock turned negative after its latest earnings release.
CoreWave, nicknamed "Nvidia's son," generates only hundreds of millions in quarterly revenue, yet secured over $36 billion in orders from OpenAI and Meta within a week. Over the past month, the company’s stock has dropped 17%.
The market certainly worries about whether these companies have sufficient financial capacity to meet customer demands—but it also questions whether the customers themselves are truly reliable.
When peeling back the layers of AI infrastructure, you eventually arrive at just a few key players: tech giants like Meta, Google parent Alphabet, Microsoft, Amazon, Apple, Nvidia, plus star AI startups such as OpenAI and Anthropic.
The star startups are still immature, relying almost entirely on external funding for infrastructure—clearly risky.
The giants should be the stabilizing anchors—they have solid finances and abundant cash, now filling their next several years with multi-hundred-billion-dollar infrastructure plans.
But the returns they’re getting from being the primary spenders on AI remain minimal. Whether using their established profits to fund new dreams will ultimately drag down the giants depends entirely on how quickly those dreams materialize.
Success means everyone wins; failure could mean total collapse.
01
In just months, Oracle experienced both euphoria and despair by holding what seemed like the ultimate “future” card.
At the height of euphoria, Oracle’s stock surged 40% in a single day, briefly making founder and CEO Larry Ellison the world’s richest person, surpassing Musk.
Ellison then proclaimed: “Artificial intelligence is everything!”
Indeed, AI was everything—for Oracle, it was the sole reason behind this surge. At the time, OpenAI had signed a five-year, $300 billion computing power procurement deal with Oracle, which acted as the spark igniting Oracle’s stock.
Yet merely three months later, despite Oracle securing even more orders, the “magic” vanished.
Oracle recently released its Q2 FY2026 results (covering November 2025), showing a 14% year-on-year revenue increase, with its order backlog reaching an all-time high of $523 billion.
This figure represents a $68 billion increase from the previous quarter.
Upon release, the stock dropped 11% that day—the largest single-day decline since January. From its September peak, Oracle’s share price has now fallen 40%.
Future orders, amid growing skepticism about an “AI bubble,” have shifted from a symbol of promise into a heavy burden.
Oracle appears strained—its financials show negative $10 billion in cash flow, with quarterly capital expenditures (CapEx) hitting $12 billion, nearly $3.7 billion above analyst expectations.
Moreover, Oracle’s CFO revealed that annual spending guidance has been raised by as much as $15 billion, reaching $50 billion.
The market’s biggest fear: Does Oracle really have the financial capacity to fund such massive AI infrastructure projects?
Some analysts predict Oracle may need to borrow $100 billion to complete construction. In Q2, the company raised $18 billion in debt—one of the largest bond issuances ever recorded among tech firms.
During its earnings call, Oracle strongly contested the “$100 billion borrowing” forecast, insisting actual financing needs would be far lower. The key lies in Oracle’s adoption of a “customer-brings-chips” partnership model.
In other words, instead of Oracle buying chips and leasing them to customers, clients bring their own chips—an unprecedented approach in the cloud services industry.
Additionally, Oracle emphasized that some suppliers are willing to lease rather than sell chips to them, allowing synchronized payment and collection cycles.
If Oracle’s claims hold true, it could significantly reduce upfront investments and greatly improve return rates.
However, for the market, the risk hasn’t disappeared—it has simply shifted: from Oracle to Oracle’s customers. Clients like Meta or OpenAI must purchase expensive GPUs themselves and install them in Oracle’s data centers.
Whether Oracle can deliver on its multi-hundred-billion-dollar future commitments depends not only on its ability to “deliver” but also on customers’ ability to
“pay.” Of Oracle’s nearly $500 billion in undelivered orders, about two-thirds come from unprofitable OpenAI, with another known $20 billion stemming from a new agreement with Meta.
Similarly, Broadcom, despite holding massive orders, received negative market feedback.
Broadcom also released new earnings, reporting core revenue and profits exceeding expectations for its Q4 FY2025 (ended November 2). Its AI semiconductor-related revenue grew 74% year-over-year.
During the call, Broadcom CEO Hock Tan stated the company currently has approximately $73 billion in AI product order backlog, expected to be fulfilled over the next six quarters. He emphasized this is a “floor value,” and the backlog is likely to grow further as new orders pour in.
However, Broadcom declined to provide clear full-year 2026 AI revenue guidance, citing uncertainties in customer deployment timelines and potential quarterly fluctuations.
After the report, Broadcom’s stock initially rose about 3%, then reversed to close down over 4% after hours.
Compared to Oracle’s dramatic highs and lows, Broadcom experienced only minor turbulence, but the underlying market sentiment is similar—a waning optimism toward the grand AI infrastructure “future.”
Broadcom’s customers are similarly concentrated, with its AI-related orders primarily coming from OpenAI, Anthropic, Alphabet, and Meta.
02
Peeling back the layers of AI infrastructure inevitably reveals the same familiar names—America’s “Magnificent Seven” and AI stars OpenAI and Anthropic.
Another AI cloud infrastructure startup, CoreWave, attracted significant attention this year. CoreWeave went public in March, marking the largest tech IPO since 2021. Its stock more than doubled afterward, even outperforming the “Seven Tech Giants.”
Its customer concentration is extremely high, essentially surviving on orders from Microsoft, OpenAI, Nvidia, and Meta.
Just this Monday (December 9), CoreWave issued another $2 billion in convertible bonds, adding to its already substantial debt totaling $14 billion as of end-September. Market concerns intensified—its stock has dropped 17% over the past month.
Again, the point stands: the market now harbors deep skepticism about the entire AI sector—not only questioning whether AI infrastructure providers can deliver as planned, but also whether their big-spending customers can actually pay their bills.
A complex web of interlocking transactions has formed among all parties involved—tightly knit and opaque—making the situation even murkier.
Looking at customer types, startups like OpenAI and Anthropic were the first to raise red flags.
The reason is simple: neither has stable revenue generation, clearly insufficient to support their ambitious infrastructure plans. They rely on external funding, introducing obvious uncertainty.
In contrast, the giants act more like trendsetters and safety nets in this game.
These giants spend hundreds of billions annually on capital expenditure, a large portion going toward expanding data centers. Their combined capital spending in 2026 will exceed four times the total spending of the U.S. listed energy industry on drilling, oil and gas extraction, fuel transportation, and operating large chemical plants. Amazon alone will spend more than the entire U.S. energy sector combined.
Compared to fledgling startups, the giants clearly have deep pockets, solid finances, and strong cash flows. At least for now, their spending remains within manageable limits.
For example, Microsoft, Google, and Amazon together will spend over $600 billion from 2023 through this year, with projected revenues of $750 billion.
Reviewing their recent earnings reports reveals strong performance—“exceeding expectations” has become routine. In short, there seems little to worry about: when it comes to building AI infrastructure at scale, they can afford it.
But upon closer inspection, none have fundamentally transformed their revenue structures. While AI is generating returns, it still plays a minor role in overall income—despite dominating spending.
Take Microsoft: In late July, TheCUBE Research estimated that AI services contributed about 19%—over $3 billion—to Azure’s growth, yet this accounts for less than one-tenth of Microsoft’s total revenue.
More than half of Google’s revenue still comes from advertising and search, while e-commerce and ads make up over 70% of Amazon’s revenue.
In essence, the giants are using mature businesses to fund AI’s future.
The question is: how long can this continue?
03
The giants have begun a “debt spree.”
In September, Meta issued $30 billion in bonds. Alphabet recently announced plans to issue approximately $17.5 billion in bonds in the U.S. market and another $3.5 billion in Europe.
According to Bank of America, large tech companies focused on AI issued $75 billion in U.S. investment-grade bonds in just September and October—more than double the industry’s average annual issuance of $32 billion between 2015 and 2024.
Currently, these companies’ revenue growth should support their spending, but to keep pace with AI advancements, they will ultimately require even more debt.
The Wall Street Journal delivered a sharp analysis: AI is weakening the giants.
As of the end of Q3 this year, Microsoft’s cash and short-term investments accounted for about 16% of total assets, down from roughly 43% in 2020. Alphabet and Amazon have also seen sharp declines in cash reserves.
Alphabet and Amazon are expected to report lower free cash flow this year compared to last. Although Microsoft’s free cash flow over the past four quarters appears higher than the prior year, its disclosed CapEx excludes long-term lease expenses for data centers and computing equipment. Including those, its free cash flow would also decline.
This trend appears set to continue.
Analysts estimate that if lease costs are included, Microsoft will spend approximately $159 billion next year; Amazon, $145 billion; and Alphabet, $112 billion. If forecasts hold, these companies will collectively invest $1 trillion over four years, mostly directed toward AI.
Overall, these changes—declining cash balances, shrinking cash flows, rising debt—are fundamentally transforming the business models of tech companies.
The tech industry is beginning to resemble sectors like semiconductor manufacturing, where hundreds of billions are invested in cutting-edge factories that take years to build—and even longer to generate returns.
Deploying hundreds of billions across vast data centers presents clear, immense challenges even from an execution standpoint.
Data centers consume enormous electricity—GPUs require massive power for computation—and current power grids cannot handle the surge in demand. Cooling is another major issue. GPUs run hot and require large amounts of freshwater to operate. Some communities have started opposing data center construction due to water supply concerns.
Earlier this year, Nvidia and OpenAI jointly announced a massive $100 billion agreement, under which OpenAI plans to deploy 10 gigawatts of Nvidia systems. However, Nvidia’s CFO recently admitted the plan is still at the letter-of-intent stage and hasn’t been formally signed.
This casts doubt on the credibility of high-profile AI infrastructure deals and highlights future uncertainty.
The reasons for the delayed signing haven’t been disclosed, but Nvidia’s SEC filings offer clues under the “risk factors” section.
There, Nvidia warns that if customers reduce demand, delay funding, or change direction, the company could face risks including “excess inventory,” “order cancellation penalties,” or “inventory write-downs and impairments.”
Furthermore, the availability of “data center capacity, power, and capital” is critical for deploying AI systems. The filing describes power infrastructure development as a “multi-year process” facing “regulatory, technical, and construction challenges.”
Even if AI infrastructure progresses smoothly, that wouldn’t mark the end of success.
Ultimately, AI infrastructure exists to serve AI demand. If infrastructure is built but market demand fails to materialize, low utilization rates could lead to massive losses.
Of course, not everyone is pessimistic. Supporters see this as a high-stakes gamble worth taking, believing AI demand will grow exponentially, not linearly.
Analyst Azeem Azhar calculated that direct revenue from AI services has nearly increased ninefold over the past two years.
In other words, if this growth rate continues, it’s only a matter of time before AI companies start generating record-breaking profits.
“I think people obsessing over the specific financing methods of these investments have outdated thinking. Everyone assumes this technology will develop linearly. But AI is exponential. It’s an entirely different paradigm,” said Azhar.
But the real question remains: Will the moment when AI starts explosively generating “profits” ever arrive—and when?
Ultimately, whether AI infrastructure will cripple the giants hinges on whether market demand can catch up with infrastructure buildup. If demand catches up, the investment was worthwhile. If not, the sprawling data centers will become ghost towns—an undeniable sign that the giants bet wrong on AI, leading to catastrophic consequences.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
