Written by: Web4 Research Center

On the morning of March 31, employees of Oracle in several countries around the world opened their emails to find a bluntly worded message: "After careful consideration of Oracle's current business needs, we have decided to remove your position in the organizational restructuring. Today is your last working day..."

There was no prior communication, no HR interviews. After the email was sent, system permissions were instantly revoked, and unassigned restricted stock was immediately rendered invalid. According to investment bank TDCowen, this layoff is expected to affect 20,000 to 30,000 employees, approximately 18% of its total global workforce of 162,000.

This is the largest layoff in Oracle's history.

However, just days before this email was sent, the company had released a financial report that most publicly listed companies envy. As of the end of the second quarter on November 30, 2025, Oracle's GAAP net profit reached $6.1 billion, with earnings per share growing by 91% year on year. With excellent performance, it suddenly decided to lay off nearly one-fifth of its workforce. What exactly happened behind this?

The answer quickly emerged: according to TD Cowen's analysis, the total investment in the expansion plan is expected to reach $156 billion—almost a trillion yuan.

In order to raise this large sum, Oracle has already raised $45 billion to $50 billion through debt and equity financing for Oracle Cloud infrastructure development, with an additional $58 billion in new debt occurring within two months. The anticipated $8 billion to $10 billion in cash flow released this time is merely "small change" in the face of this massive bill.

How does a company that earns hundreds of billions a year have to rely on layoffs to "raise money" for an all-in AI strategy—what level has the entry fee into the AI competition risen to?

1. Conflicting Signals:

Why is Oracle laying off employees while making profits?

This layoff was not without warning. According to a report by Bloomberg on March 5, Oracle was then planning a multi-department layoff involving "thousands of people," with some positions believed to be replaced by AI. Even so, when the large-scale layoffs actually occurred, it still surprised the market.

The hardest-hit areas of this layoff include the revenue and health sciences departments (approximately 30% layoffs), the SaaS and virtual operations service departments (also about 30%), and a significant downsizing of the NetSuite India development center. According to Oracle's 10-Q quarterly report submitted in March 2026, the company has set up a $2.1 billion restructuring budget for this purpose.

However, Oracle's financial report numbers present a different picture. Total revenue for the second quarter reached $16.1 billion, a year-on-year increase of 14%. Cloud infrastructure (IaaS) revenue increased by as much as 68% year-on-year, and remaining performance obligations—that is, signed but unrecognized revenue—surged over fivefold from a year ago to $523 billion.

Revenue is rising, profits are rising, cloud business is growing at nearly 70% per year, yet Oracle still needs to lay off nearly 20,000 people. This sounds like a paradox.

But a detail in the financial report explains it all: the free cash flow for the second quarter unusually turned negative by $10 billion. Oracle has significantly raised its estimated capital expenditure for the year to $50 billion, about $15 billion more than initially expected.

This is the answer. Profits are up, but cash flow is negative; orders are soaring, yet all the money is invested in data centers.

In fact, the whole Silicon Valley is playing out the same script. Since 2025, Amazon has cut about 30,000 corporate jobs, and Meta resumed layoffs in March 2026; the entire tech industry is doing two things simultaneously—massively investing in AI while heavily cutting back on non-core businesses. The reason is simple: AI infrastructure is too money-consuming, so much so that even the most profitable tech giants must make choices between manpower and computational power.

Oracle chose the latter.

2. The AI Competition has Changed—From “Competing on Technology” to “Competing on Capital”

If the AI competition three years ago was still a battle of algorithms and model architectures, then the AI competition of 2026 has turned into a complete capital game.

The cost of training large models has reached jaw-dropping levels. According to industry estimates, the training cost of GPT-5 level models has exceeded $1 billion. The construction of data centers is astronomically high—establishing a self-built computing center requires tens of billions of dollars for land, electricity, liquid cooling systems, and network bandwidth. The bigger cost lies in operations: with each additional user of AI applications, real cash goes into electricity bills and inference expenses.

Faced with this cost structure, tech giants reacted surprisingly uniformly: spend money, keep spending money, and continue to spend money.

According to a report from market research agency Futurum Group, the five largest U.S. cloud and AI infrastructure providers—Microsoft, Google, Amazon, Meta, and Oracle—are expected to have a total capital expenditure of $660 billion to $690 billion in 2026, nearly double the approximately $380 billion in 2025.

Specifically, the arms race among the giants has reached an astonishing level.

Amazon leads the pack with a capital expenditure plan of $200 billion. This figure even exceeds the most optimistic analysts' expectations—the market had previously estimated around $147 billion. Amazon CEO Andy Jassy argued that AI capacity is being absorbed by the market at an installation-to-revenue speed, with annual revenue for AWS accelerated to $142 billion. Yet even so, Amazon's stock price still fell about 8% to 10% after the announcement, as investors were worried about the investment payback period.

Alphabet, Google's parent company, follows closely, expecting capital expenditures to be as high as $175 billion to $185 billion in 2026, nearly double the $91.4 billion in 2025. This figure alone is indicative—one company’s annual hardware investment has already surpassed the GDP of most countries.

Meta plans to invest $115 billion to $135 billion in capital expenditures in 2026, primarily for the construction of the "Meta Super Intelligence Laboratory" and expansion of data center capacity. Meta's capital expenditure was $72.2 billion in 2025, marking a significant increase.

Microsoft is advancing with a quarterly capital expenditure rate of $37.5 billion, expecting to invest $120 billion or more in the 2026 fiscal year.

Oracle's posture in this round of competition is also aggressive. According to insiders, one reason Oracle initiated this layoff is the insufficient return on AI investments, with the technology and financial services departments becoming the hardest hit. Interestingly, analysts from investment bank TD Cowen reported at the beginning of this year that if Oracle laid off 20,000 to 30,000 people, it could generate an additional $8 billion to $10 billion in free cash flow. —This means that Oracle viewed layoffs from the start as part of its plan to raise funds for AI infrastructure.

The combined annual capital expenditure of nearly $700 billion from these five companies means what? To make a comparison: this is higher than the entire GDP of Israel and also exceeds the total revenue of all global cloud infrastructure services.

The AI competition has split into two entirely different tracks.

Track A refers to infrastructure and foundational model layers—requiring hundreds of billions in capital, multi-thousand cluster deployments, and global data centers. This is the battleground for the giants, a capital game of "nuclear deterrence" level. Track B refers to application and scenario layers—having relatively lower barriers, focusing more on understanding vertical industries and insights into specific scenarios.

Is it almost impossible for small and medium players to squeeze into Track A?

3. Implications for Web4 Entrepreneurs:

Don’t compete with giants on who has more money, but on who understands scenarios better.

For entrepreneurs in the AI and Web4 fields, the signal conveyed by Oracle's layoffs couldn't be clearer:

Do not aim to be an "electric power plant" in the AI era, but to be an "appliance company."

Power plants belong to the giants—requiring hundreds of billions in investments, driven by economies of scale, with global deployments. Appliance companies utilize the already constructed "power grid" to create good products that solve specific problems. History has repeatedly verified this rule: in every significant leap of technological infrastructure, the entities that create the most value are not the builders of the infrastructure itself, but the entrepreneurs who construct applications and scenarios on top of it.

The current AI infrastructure frenzy bears an astonishing similarity in underlying logic to previous internet bubbles, mobile internet waves, and cloud computing revolutions. In the internet wave, it was giants and national capital that built the backbone networks and undersea cables, but the companies that truly changed the world were those that established e-commerce, social networks, and search engines on this network. In the era of mobile internet, it was the operators and telecom giants that laid down base stations and 4G networks, while the ones that created trillion-dollar valuations were the entrepreneurial companies that built application scenarios based on smartphones.

The intersection of AI and Web4 is precisely the blind spot that the giants cannot see, do not understand, and temporarily cannot address.

The business logic of the giants dictates their pursuit of "general AI capability"—one model to solve all problems. This logic inherently rejects highly customized, low-frequency, non-standard vertical scenarios. Yet within these overlooked gaps lie the greatest opportunities for startups.

So, what specific directions are worth focusing on?

The first direction is AI Agents and on-chain automation. The smart contract ecosystem inherently requires automated execution strategies, on-chain audits, liquidity management, and other functionalities. Currently, most automation solutions in the Web3 field remain at the level of simple scheduled tasks or trigger-based scripts, lacking true intelligent decision-making capabilities. Meanwhile, the giants’ microscopic understanding and engineering capabilities regarding on-chain complex logic are far less than those of dedicated entrepreneurial teams in this area. Providing intelligent automation services for DeFi strategies, DAO governance, on-chain security, etc., using AI Agents is a typical "niche market overlooked by giants."

The second direction is the combination of private data and AI inference. Users own data ownership, and AI models offer services through zero-knowledge proofs or federated learning, completing inference tasks without touching the original data. This model has huge application potential in fields like healthcare, finance, and law, where data privacy is of utmost importance. While the giants possess powerful model capabilities, their business models inherently rely on data collection and centralized processing, leading to structural contradictions in truly respecting data ownership. Web4 entrepreneurs can leverage the trust layer advantages of blockchain to establish a moat in this gap.

The third direction is AI Copilot for vertical industries. Valuation of Web3 gaming assets, NFT liquidity prediction, cross-chain asset scheduling optimization, on-chain identity credit assessment... these scenarios are narrow enough and vertical enough that giants will not have the motivation to allocate resources to pursue them. Yet each of these niche scenarios could support a small but beautiful entrepreneurial project. The key is to truly understand the core pain points of these scenarios, rather than generically creating a "general Web3 assistant."

A simple but effective evaluation framework can help entrepreneurs assess whether their projects are secure: if your core competitive advantage lies in computational power or model parameters, you will inevitably be crushed by the giants; if your core competitive advantage is industry knowledge, user relationships, or a deep understanding of on-chain data, the giants' entry into this field will instead validate your direction—because what giants need is not your technology, but your scenario understanding.

Ultimately, AI will not eliminate you, but those with AI will. And in the world of Web4, this competitor is often not the giants, but another small team that understands scenarios better than you.

Conclusion

The moment Oracle's layoff email was sent, the 20,000 to 30,000 employees lost not just a job but also a metaphor for an era.

The ticket for entry into the AI competition has risen to levels unimaginable for the average person. But beyond the ticket, there is another path.

Power plants belong to giants, while appliance companies belong to entrepreneurs. Some of the positions that are being eliminated will indeed be permanently replaced by AI, but many more opportunities are emerging from these gaps.

AI is not an opportunity for everyone, but it will forever belong to those who find the right position.

(This article is solely for industry analysis and does not constitute any investment advice)