Holding massive orders related to AI infrastructure is no longer sufficient to "protect" a company.

Oracle holds $500 billion in orders, yet its stock price has plummeted 40% from its peak in September. Broadcom currently has a backlog of approximately $73 billion in AI product orders, and its stock price reversed from gains to losses after the release of its latest earnings report.
CoreWeave, known as "Nvidia's protégé," achieves a quarterly revenue of over $1 billion but secured more than $36 billion in orders from OpenAI and Meta within a single week. Over the past month, the company's stock price has dropped by 17% cumulatively.
The outside world is certainly concerned about whether these companies have the financial capacity to meet customer demands, and at the same time, they worry about whether the customers themselves are truly reliable.
Peeling back the layers of the AI infrastructure onion, we ultimately find just a handful of key players: tech giants including Meta, Alphabet (Google's parent company), Microsoft, Amazon, Apple, and Nvidia, along with leading AI startups like OpenAI and Anthropic.
These prominent startups are still in their early stages and rely almost entirely on external financing for infrastructure development, resulting in obvious risks.
The tech giants were supposed to act as stabilizing forces. With sound financial standing and abundant cash reserves, they have lined up trillions of dollars in aggressive infrastructure plans for the coming years.
However, AI, which takes center stage in terms of expenditures, has yielded only meager returns so far. Whether funding this new vision with their "accumulated profits" will drag the giants down hinges entirely on how quickly this vision can deliver tangible results.
Success will benefit all parties, while failure could lead to total collapse.
I
Armed with the "future" as its trump card, Oracle experienced extreme highs and lows within just a few months.
At the peak of its success, Oracle's stock price soared 40% in a single day, and its founder and CEO Larry Ellison briefly surpassed Elon Musk to become the world's richest person.
At that time, Ellison declared triumphantly: "Artificial intelligence is everything!"
For Oracle, AI truly was the driving force behind this extraordinary boom. OpenAI had just signed a five-year, $300 billion computing power procurement agreement with Oracle, which served as the spark that ignited Oracle's stock price surge.
Yet merely three months later, even though Oracle held more orders, the "magic" had faded away.
Oracle recently released its earnings report for the second quarter of fiscal year 2026 (covering September to November 2025), showing a 14% year-on-year increase in revenue, and the company stated that its backlog of orders had reached an astonishing $523 billion.
This figure represented a $68 billion increase from the previous quarter.
Following the earnings announcement, the stock price plummeted 11% that day, marking the company's largest single-day drop since January. From its September peak, Oracle's stock has now declined by a cumulative 40%.
Amid growing doubts about an "AI bubble," future orders have transformed from promising prospects into an overwhelming burden.
Oracle is clearly struggling. The earnings report revealed a negative cash flow of $10 billion and quarterly capital expenditures (CapEx) of $12 billion, nearly $3.7 billion higher than analysts' forecasts.
Furthermore, Oracle's CFO disclosed that the company has raised its full-year expenditure by as much as $15 billion, reaching $50 billion.
The market's greatest fear is: Does Oracle have the financial resources to support such large-scale AI infrastructure development?
Some analysts predict that Oracle will need to borrow $100 billion to complete the projects. In the second quarter, the company raised $18 billion in debt, one of the largest bond issuances ever recorded by a tech enterprise.
During the earnings call, Oracle vigorously defended itself, explicitly refuting the $100 billion borrowing forecast and claiming that the actual financing amount would be significantly lower. The key lies in Oracle's innovative "customer-supplied chips" cooperation model.
In other words, instead of Oracle purchasing chips to lease to customers, clients bring their own chips—a groundbreaking approach in the cloud services industry.
Additionally, Oracle emphasized that some suppliers are willing to lease rather than sell chips to the company, enabling it to synchronize payments and collections.
If Oracle's claims hold true, it can significantly reduce its upfront investment and substantially boost its rate of return.
However, for the market, the risk has not disappeared but merely shifted—from Oracle to its customers. Clients such as Meta and OpenAI must purchase expensive GPUs and install them in Oracle's data centers.
Whether Oracle can realize its multi-billion-dollar future depends not only on its ability to deliver services but also on its customers' ability to make payments. Approximately two-thirds of Oracle's nearly $500 billion in outstanding orders come from the unprofitable OpenAI, and a known $20 billion stems from a new agreement with Meta.
Broadcom, another company sitting on a huge order backlog, has also received a negative market response.
Broadcom also released its latest earnings report. For the fourth quarter of fiscal year 2025, ending November 2, its core revenue and profits both exceeded expectations, with revenue from AI semiconductors surging 74% year-on-year.
During the earnings call, Hock Tan, CEO of Broadcom, stated that the company currently has a backlog of approximately $73 billion in AI product orders, which will be fulfilled within the next six quarters. He emphasized that this figure is a "minimum," and the backlog is expected to expand further as new orders continue to pour in.
Nevertheless, Broadcom declined to provide clear guidance on its full-year AI revenue for 2026, citing uncertainties in customer deployment schedules that may lead to quarterly fluctuations.
After the earnings report was released, Broadcom's stock price initially rose by about 3% but later reversed course, falling more than 4% in after-hours trading.
Compared to Oracle's dramatic ups and downs, Broadcom's experience was just a minor setback. However, the underlying market sentiment remains the same—optimism about the future of large-scale AI infrastructure development has faded.
Broadcom's customer base is similarly concentrated, with its AI-related orders primarily coming from OpenAI, Anthropic, Alphabet, and Meta.
II
Peeling back the layers of the AI infrastructure onion always reveals the same familiar names: the Magnificent Seven, OpenAI, and Anthropic.
CoreWeave, an AI cloud infrastructure startup that has garnered much attention this year, went public in March. As the largest tech startup IPO since 2021, its stock price more than doubled afterward, even outperforming the Magnificent Seven.
CoreWeave also has an extremely concentrated customer base, relying almost entirely on orders from Microsoft, OpenAI, Nvidia, and Meta for its operations.
This Monday (December 9), CoreWeave issued an additional $2 billion in convertible bonds. By the end of September, its total outstanding debt had already reached $14 billion. Market concerns have intensified, and the company's stock price has dropped 17% over the past month.
Once again, the market holds deep-seated doubts about the entire AI industry—not only whether these AI infrastructure firms can deliver services as planned but also whether major clients engaging in these large transactions can settle their bills reliably.
The complex circular transactions among all stakeholders have formed a dense, opaque web, making the entire sector increasingly difficult to assess.
Among different types of clients, startups like OpenAI and Anthropic were the first to raise concerns.
The reason is straightforward: neither has established a stable profit model sufficient to support their expanding infrastructure plans. They rely heavily on external financing, resulting in significant uncertainties.
In contrast, the tech giants were supposed to serve as bellwethers and safety nets.
These giants allocate hundreds of billions of dollars annually to capital expenditures, a substantial portion of which goes toward expanding data centers. Their combined capital spending in 2026 will exceed four times the total expenditure of publicly traded U.S. energy companies on activities such as drilling exploration wells, oil and gas extraction, gasoline transportation to gas stations, and operation of large chemical plants. Amazon alone spends more on capital projects than the entire U.S. energy industry combined.
Compared to fledgling startups, these giants are clearly financially robust, with sound balance sheets and ample cash flow. For now, their expenditures remain within manageable limits.
For instance, Microsoft, Google, and Amazon will collectively spend over $600 billion between 2023 and this year, with projected revenues of $750 billion.
Their recent earnings reports show strong performance, with results exceeding expectations becoming the norm. On the surface, there is little cause for concern—in other words, they can afford the massive investment in AI infrastructure.
However, a closer look reveals that none of them have achieved a fundamental transformation in their revenue structures. While AI has started generating returns, it still plays a secondary role in overall revenue despite being the top priority in terms of spending.
Take Microsoft as an example. In late July, TheCUBE Research estimated based on its quarterly earnings report that AI services accounted for approximately 19% of Azure Cloud's growth, exceeding $3 billion. Yet this constitutes less than one-tenth of Microsoft's total revenue.
Over half of Google's revenue still comes from advertising and search, while e-commerce and advertising make up more than 70% of Amazon's total revenue.
In essence, the tech giants are nurturing AI's future using profits from their mature businesses.
The critical question is: How long can this nurturing last?
III
The tech giants have already sparked a "borrowing 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 around $3.5 billion in Europe.
Data from Bank of America shows that in September and October alone, large AI-focused tech companies issued $75 billion in U.S. investment-grade bonds—more than double the industry's average annual issuance of $32 billion between 2015 and 2024.
Currently, these companies' revenue growth can support their expenditures, but to keep pace in the AI sector, they will ultimately need to take on more debt.
The Wall Street Journal made a sharp observation in an analysis: AI is weakening the tech giants.
By the end of the third quarter this year, Microsoft's cash and short-term investments accounted for approximately 16% of its total assets, down from around 43% in 2020. Alphabet and Amazon have also seen significant declines in their cash reserves.
Alphabet and Amazon's free cash flow for this year is expected to be lower than that of last year. Although Microsoft's free cash flow over the past four quarters appears to have increased year-on-year, its disclosed capital expenditures do not include long-term leases for data centers and computing equipment. Including these expenses would result in a decline in its free cash flow.
This trend seems destined to continue.
Analysts estimate that with lease expenses included, Microsoft is projected to spend approximately $159 billion next year, Amazon around $145 billion, and Alphabet $112 billion. If these forecasts hold true, these companies will invest a total of $1 trillion over four years, with the majority allocated to the AI sector.
Overall, these changes—declining cash balances, shrinking cash flow, and rising debt—are fundamentally transforming the business models of tech companies.
The tech industry is increasingly resembling sectors like semiconductor manufacturing, where hundreds of billions of dollars are invested in cutting-edge factories that take years to build and even longer to generate returns.
Deploying hundreds of billions of dollars across hundreds of large-scale data centers poses significant challenges for AI infrastructure, even from an operational perspective.
Data centers consume enormous amounts of electricity—GPUs require substantial power for computing—and the current power grid cannot cope with this surge in demand. Additionally, cooling presents a major issue. GPUs operate at extremely high temperatures and require large volumes of fresh water to maintain optimal performance. Some communities have begun opposing data center construction over concerns about water supply impacts.
Earlier this year, Nvidia and OpenAI jointly announced a new $100 billion agreement, under which OpenAI plans to deploy 10 gigawatts of Nvidia systems. However, Nvidia's CFO recently admitted that this plan is still in the letter of intent stage and has not yet been formally signed.
This not only casts doubt on the credibility of the bustling AI infrastructure transactions but also underscores future uncertainties.
The reasons for the delayed signing remain undisclosed, but the "risk factors" section in Nvidia's filing with the SEC offers some insights.
In the filing, Nvidia warns that if customers reduce demand, delay financing, or change strategic directions, the company may face risks such as "excess inventory," "penalties for order cancellations," or "inventory write-downs and impairments."
Furthermore, the availability of "data center capacity, power, and capital" is critical to AI system deployment. The filing notes that power infrastructure development is a "multi-year process" that will encounter "regulatory, technical, and construction challenges."
Even if AI infrastructure progresses smoothly, this will not mark the end of the journey toward success.
AI infrastructure ultimately serves AI demand. If the infrastructure is completed but market demand fails to materialize, underutilization of these facilities will lead to massive losses.
Of course, not everyone is filled with anxiety. Proponents argue that this is a high-stakes gamble worth taking, as AI demand will grow exponentially rather than linearly.
Analyst Azeem Azhar calculates that direct revenue from AI services has nearly increased ninefold over the past two years.
In other words, if this growth rate persists, it will only be a matter of time before AI companies start generating record profits.
"I think people who get bogged down in the specifics of financing these investments have outdated mindsets. Everyone assumes this technology will develop linearly. But artificial intelligence is an exponentially growing technology. It operates on an entirely different paradigm," Azhar stated.
However, the critical question remains: Will the era of explosive AI-driven profits ever arrive, and if so, when?
Ultimately, whether AI infrastructure will drag down the tech giants boils down to a race between AI market demand and infrastructure development. If demand catches up, the investment in AI infrastructure will prove worthwhile. If not, these massive data centers will eventually become like ghost towns. This would serve as irrefutable proof that the giants' bets on AI were misplaced, with catastrophic consequences to follow.

Holding massive orders related to AI infrastructure is no longer sufficient to "protect" a company.

Oracle holds $500 billion in orders, yet its stock price has plummeted 40% from its peak in September. Broadcom currently has a backlog of approximately $73 billion in AI product orders, and its stock price reversed from gains to losses after the release of its latest earnings report.
CoreWeave, known as "Nvidia's protégé," achieves a quarterly revenue of over $1 billion but secured more than $36 billion in orders from OpenAI and Meta within a single week. Over the past month, the company's stock price has dropped by 17% cumulatively.
The outside world is certainly concerned about whether these companies have the financial capacity to meet customer demands, and at the same time, they worry about whether the customers themselves are truly reliable.
Peeling back the layers of the AI infrastructure onion, we ultimately find just a handful of key players: tech giants including Meta, Alphabet (Google's parent company), Microsoft, Amazon, Apple, and Nvidia, along with leading AI startups like OpenAI and Anthropic.
These prominent startups are still in their early stages and rely almost entirely on external financing for infrastructure development, resulting in obvious risks.
The tech giants were supposed to act as stabilizing forces. With sound financial standing and abundant cash reserves, they have lined up trillions of dollars in aggressive infrastructure plans for the coming years.
However, AI, which takes center stage in terms of expenditures, has yielded only meager returns so far. Whether funding this new vision with their "accumulated profits" will drag the giants down hinges entirely on how quickly this vision can deliver tangible results.
Success will benefit all parties, while failure could lead to total collapse.
I
Armed with the "future" as its trump card, Oracle experienced extreme highs and lows within just a few months.
At the peak of its success, Oracle's stock price soared 40% in a single day, and its founder and CEO Larry Ellison briefly surpassed Elon Musk to become the world's richest person.
At that time, Ellison declared triumphantly: "Artificial intelligence is everything!"
For Oracle, AI truly was the driving force behind this extraordinary boom. OpenAI had just signed a five-year, $300 billion computing power procurement agreement with Oracle, which served as the spark that ignited Oracle's stock price surge.
Yet merely three months later, even though Oracle held more orders, the "magic" had faded away.
Oracle recently released its earnings report for the second quarter of fiscal year 2026 (covering September to November 2025), showing a 14% year-on-year increase in revenue, and the company stated that its backlog of orders had reached an astonishing $523 billion.
This figure represented a $68 billion increase from the previous quarter.
Following the earnings announcement, the stock price plummeted 11% that day, marking the company's largest single-day drop since January. From its September peak, Oracle's stock has now declined by a cumulative 40%.
Amid growing doubts about an "AI bubble," future orders have transformed from promising prospects into an overwhelming burden.
Oracle is clearly struggling. The earnings report revealed a negative cash flow of $10 billion and quarterly capital expenditures (CapEx) of $12 billion, nearly $3.7 billion higher than analysts' forecasts.
Furthermore, Oracle's CFO disclosed that the company has raised its full-year expenditure by as much as $15 billion, reaching $50 billion.
The market's greatest fear is: Does Oracle have the financial resources to support such large-scale AI infrastructure development?
Some analysts predict that Oracle will need to borrow $100 billion to complete the projects. In the second quarter, the company raised $18 billion in debt, one of the largest bond issuances ever recorded by a tech enterprise.
During the earnings call, Oracle vigorously defended itself, explicitly refuting the $100 billion borrowing forecast and claiming that the actual financing amount would be significantly lower. The key lies in Oracle's innovative "customer-supplied chips" cooperation model.
In other words, instead of Oracle purchasing chips to lease to customers, clients bring their own chips—a groundbreaking approach in the cloud services industry.
Additionally, Oracle emphasized that some suppliers are willing to lease rather than sell chips to the company, enabling it to synchronize payments and collections.
If Oracle's claims hold true, it can significantly reduce its upfront investment and substantially boost its rate of return.
However, for the market, the risk has not disappeared but merely shifted—from Oracle to its customers. Clients such as Meta and OpenAI must purchase expensive GPUs and install them in Oracle's data centers.
Whether Oracle can realize its multi-billion-dollar future depends not only on its ability to deliver services but also on its customers' ability to make payments. Approximately two-thirds of Oracle's nearly $500 billion in outstanding orders come from the unprofitable OpenAI, and a known $20 billion stems from a new agreement with Meta.
Broadcom, another company sitting on a huge order backlog, has also received a negative market response.
Broadcom also released its latest earnings report. For the fourth quarter of fiscal year 2025, ending November 2, its core revenue and profits both exceeded expectations, with revenue from AI semiconductors surging 74% year-on-year.
During the earnings call, Hock Tan, CEO of Broadcom, stated that the company currently has a backlog of approximately $73 billion in AI product orders, which will be fulfilled within the next six quarters. He emphasized that this figure is a "minimum," and the backlog is expected to expand further as new orders continue to pour in.
Nevertheless, Broadcom declined to provide clear guidance on its full-year AI revenue for 2026, citing uncertainties in customer deployment schedules that may lead to quarterly fluctuations.
After the earnings report was released, Broadcom's stock price initially rose by about 3% but later reversed course, falling more than 4% in after-hours trading.
Compared to Oracle's dramatic ups and downs, Broadcom's experience was just a minor setback. However, the underlying market sentiment remains the same—optimism about the future of large-scale AI infrastructure development has faded.
Broadcom's customer base is similarly concentrated, with its AI-related orders primarily coming from OpenAI, Anthropic, Alphabet, and Meta.
II
Peeling back the layers of the AI infrastructure onion always reveals the same familiar names: the Magnificent Seven, OpenAI, and Anthropic.
CoreWeave, an AI cloud infrastructure startup that has garnered much attention this year, went public in March. As the largest tech startup IPO since 2021, its stock price more than doubled afterward, even outperforming the Magnificent Seven.
CoreWeave also has an extremely concentrated customer base, relying almost entirely on orders from Microsoft, OpenAI, Nvidia, and Meta for its operations.
This Monday (December 9), CoreWeave issued an additional $2 billion in convertible bonds. By the end of September, its total outstanding debt had already reached $14 billion. Market concerns have intensified, and the company's stock price has dropped 17% over the past month.
Once again, the market holds deep-seated doubts about the entire AI industry—not only whether these AI infrastructure firms can deliver services as planned but also whether major clients engaging in these large transactions can settle their bills reliably.
The complex circular transactions among all stakeholders have formed a dense, opaque web, making the entire sector increasingly difficult to assess.
Among different types of clients, startups like OpenAI and Anthropic were the first to raise concerns.
The reason is straightforward: neither has established a stable profit model sufficient to support their expanding infrastructure plans. They rely heavily on external financing, resulting in significant uncertainties.
In contrast, the tech giants were supposed to serve as bellwethers and safety nets.
These giants allocate hundreds of billions of dollars annually to capital expenditures, a substantial portion of which goes toward expanding data centers. Their combined capital spending in 2026 will exceed four times the total expenditure of publicly traded U.S. energy companies on activities such as drilling exploration wells, oil and gas extraction, gasoline transportation to gas stations, and operation of large chemical plants. Amazon alone spends more on capital projects than the entire U.S. energy industry combined.
Compared to fledgling startups, these giants are clearly financially robust, with sound balance sheets and ample cash flow. For now, their expenditures remain within manageable limits.
For instance, Microsoft, Google, and Amazon will collectively spend over $600 billion between 2023 and this year, with projected revenues of $750 billion.
Their recent earnings reports show strong performance, with results exceeding expectations becoming the norm. On the surface, there is little cause for concern—in other words, they can afford the massive investment in AI infrastructure.
However, a closer look reveals that none of them have achieved a fundamental transformation in their revenue structures. While AI has started generating returns, it still plays a secondary role in overall revenue despite being the top priority in terms of spending.
Take Microsoft as an example. In late July, TheCUBE Research estimated based on its quarterly earnings report that AI services accounted for approximately 19% of Azure Cloud's growth, exceeding $3 billion. Yet this constitutes less than one-tenth of Microsoft's total revenue.
Over half of Google's revenue still comes from advertising and search, while e-commerce and advertising make up more than 70% of Amazon's total revenue.
In essence, the tech giants are nurturing AI's future using profits from their mature businesses.
The critical question is: How long can this nurturing last?
III
The tech giants have already sparked a "borrowing 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 around $3.5 billion in Europe.
Data from Bank of America shows that in September and October alone, large AI-focused tech companies issued $75 billion in U.S. investment-grade bonds—more than double the industry's average annual issuance of $32 billion between 2015 and 2024.
Currently, these companies' revenue growth can support their expenditures, but to keep pace in the AI sector, they will ultimately need to take on more debt.
The Wall Street Journal made a sharp observation in an analysis: AI is weakening the tech giants.
By the end of the third quarter this year, Microsoft's cash and short-term investments accounted for approximately 16% of its total assets, down from around 43% in 2020. Alphabet and Amazon have also seen significant declines in their cash reserves.
Alphabet and Amazon's free cash flow for this year is expected to be lower than that of last year. Although Microsoft's free cash flow over the past four quarters appears to have increased year-on-year, its disclosed capital expenditures do not include long-term leases for data centers and computing equipment. Including these expenses would result in a decline in its free cash flow.
This trend seems destined to continue.
Analysts estimate that with lease expenses included, Microsoft is projected to spend approximately $159 billion next year, Amazon around $145 billion, and Alphabet $112 billion. If these forecasts hold true, these companies will invest a total of $1 trillion over four years, with the majority allocated to the AI sector.
Overall, these changes—declining cash balances, shrinking cash flow, and rising debt—are fundamentally transforming the business models of tech companies.
The tech industry is increasingly resembling sectors like semiconductor manufacturing, where hundreds of billions of dollars are invested in cutting-edge factories that take years to build and even longer to generate returns.
Deploying hundreds of billions of dollars across hundreds of large-scale data centers poses significant challenges for AI infrastructure, even from an operational perspective.
Data centers consume enormous amounts of electricity—GPUs require substantial power for computing—and the current power grid cannot cope with this surge in demand. Additionally, cooling presents a major issue. GPUs operate at extremely high temperatures and require large volumes of fresh water to maintain optimal performance. Some communities have begun opposing data center construction over concerns about water supply impacts.
Earlier this year, Nvidia and OpenAI jointly announced a new $100 billion agreement, under which OpenAI plans to deploy 10 gigawatts of Nvidia systems. However, Nvidia's CFO recently admitted that this plan is still in the letter of intent stage and has not yet been formally signed.
This not only casts doubt on the credibility of the bustling AI infrastructure transactions but also underscores future uncertainties.
The reasons for the delayed signing remain undisclosed, but the "risk factors" section in Nvidia's filing with the SEC offers some insights.
In the filing, Nvidia warns that if customers reduce demand, delay financing, or change strategic directions, the company may face risks such as "excess inventory," "penalties for order cancellations," or "inventory write-downs and impairments."
Furthermore, the availability of "data center capacity, power, and capital" is critical to AI system deployment. The filing notes that power infrastructure development is a "multi-year process" that will encounter "regulatory, technical, and construction challenges."
Even if AI infrastructure progresses smoothly, this will not mark the end of the journey toward success.
AI infrastructure ultimately serves AI demand. If the infrastructure is completed but market demand fails to materialize, underutilization of these facilities will lead to massive losses.
Of course, not everyone is filled with anxiety. Proponents argue that this is a high-stakes gamble worth taking, as AI demand will grow exponentially rather than linearly.
Analyst Azeem Azhar calculates that direct revenue from AI services has nearly increased ninefold over the past two years.
In other words, if this growth rate persists, it will only be a matter of time before AI companies start generating record profits.
"I think people who get bogged down in the specifics of financing these investments have outdated mindsets. Everyone assumes this technology will develop linearly. But artificial intelligence is an exponentially growing technology. It operates on an entirely different paradigm," Azhar stated.
However, the critical question remains: Will the era of explosive AI-driven profits ever arrive, and if so, when?
Ultimately, whether AI infrastructure will drag down the tech giants boils down to a race between AI market demand and infrastructure development. If demand catches up, the investment in AI infrastructure will prove worthwhile. If not, these massive data centers will eventually become like ghost towns. This would serve as irrefutable proof that the giants' bets on AI were misplaced, with catastrophic consequences to follow.