Why Are the Giants, Despite Being Wealthy, Frantically Seeking Funds? Unveiling the Logic Behind the “Arms Race” in the AI Era
Hello everyone, I’m your financial observer. Recently, there’s been a rather interesting phenomenon in the financial world: China’s wealthiest internet giants—Alibaba, Tencent, and ByteDance—each of which holds tens of billions or even trillions in cash—have started to seek additional funds externally. Alibaba has issued stocks, Tencent has issued bonds, and ByteDance has borrowed $29.6 billion (approximately 200 billion RMB) from nearly 30 banks.
Many people’s first reaction is, “Aren’t they already very wealthy? Why do they still need money?”
In reality, this isn’t about a lack of funds but rather a high-stakes gamble involving time, computing power, and the right to survive in the future. Today, we’ll break down the logic behind this with simple language, explaining it in five key aspects, so you can understand this “money-burning battle” in the AI era.
---
1. Why Borrow Money When You Already Have It? Because AI Is Extremely Costly
First, let’s dispel a misconception: Borrowing money doesn’t mean you lack funds; it’s about leveraging and matching the timing of investments. You can think of a company’s cash as “current deposits” that need to be used for paying salaries, daily operations, repurchasing stocks, or dealing with unexpected risks. However, building AI infrastructure (such as data centers and purchasing servers) is a massive, long-term “capital investment.”
- Matching the Investment Timeline: Using short-term cash for a project that will last 10 years means you might have to sell the data center or borrow at a high interest rate to repay the debt, which is not cost-effective.
- Long-Term Funding for Long-Term Investments: Tencent issued 30-year bonds, ByteDance took out long-term loans, and Alibaba conducted equity financing without the pressure of repaying principal. The goal is to use long-term, low-cost funds to support long-term, high-return assets.
It’s like buying a house: you might have 500,000 RMB for daily expenses, but you wouldn’t spend it all; instead, you’d take out a 30-year loan. This way, you can keep the cash for emergencies while the monthly mortgage is covered by the loan.
Core Logic: The giants are not just trying to cover emergencies but making strategic investments. They want to retain their cash flexibility while using external financing to bring future funds into the current high-cost AI infrastructure projects.
---
2. Where Does All This Money Go? Not for Chips, but for Building “Computing Power Factories”
Many think AI is just about writing code and training models, but the bulk of the spending goes into infrastructure.
For example, ByteDance’s Taihang Computing Power Center in Datong, Shanxi, cost 4.5 billion RMB and includes 15,000 server racks. This is just a small example. Alibaba has announced plans to invest 380 billion RMB over the next three years, and ByteDance has discussed capital expenditures of up to 70 billion RMB (about 470 billion RMB) by 2026.
This money is mainly spent on three areas:
1. Data Center Construction: Land, buildings, and power infrastructure (such as 220-kilovolt power transmission projects).
2. Server and Chip Purchases: These are the core components. Although everyone focuses on NVIDIA chips, chips are just parts of the system.
3. Network and Cooling Systems: These ensure that thousands of chips can work together efficiently without overheating.
Why the Urgency? Because the time window is very tight: Models can be updated, but infrastructure cannot wait. If your competitors’ systems are up and running while yours is still under construction, the gap in time is irreparable. In the AI era, computing power is a business asset; the amount of computing power you can provide directly determines how many customers you can serve and the size of the models you can train.
---
3. How Is This Investment Calculated? From “Pure Cost” to “Profitable Assets”
Previously, building data centers was seen as a pure expense. But now, the logic has changed: Computing power is becoming a productive asset that can be rented out and generate revenue.
Tencent and Alibaba have provided specific examples of how this can lead to returns:
- Tencent’s Calculation: If they rent out their new computing power to third parties, they can almost immediately cover the depreciation costs and even make a profit. Some of the computing power reserved months ago can now be sold for a 30% profit.
- Alibaba’s Calculation: AI-related investments will pay off in about 3 years, possibly reducing to 2-2.5 years. Old equipment (like the V100 from 2018) is still in use, and AI product revenue has been growing for 12 consecutive quarters, with annual revenue exceeding 49.5 billion RMB. This means the investment has a clear return period, and old assets continue to generate income.
In Simple Terms: Building a data center used to be like opening a restaurant where you need to make a profit from selling food. Now it’s like opening a gas station with a steady stream of customers and high prices for the service. You can even sell excess fuel to other businesses, accelerating the return on investment.
Conclusion: AI investments are no longer a black hole but become calculable, predictable, and even profitable investments.
---
4. The Whole World Is Doing This: It’s a “Arms Race”
This isn’t a unique behavior of Chinese giants; it’s a collective action by global tech leaders:
- The Four American Giants: Amazon, Alphabet (Google), Meta, and Oracle issued $194 billion in bonds in the first seven months of this year, nearly 80% more than the entire year of 2025.
- Meta: Issued $25 billion in one go at the end of April.
- Oracle: Plans to raise $45-50 billion.
Why the Global Sync? Because AI is a global competition:
- Technology Knows No Borders: No matter how good your model is, if you don’t have enough computing power, your training speed will be slow, and so will your innovation.
- Talent and Ecosystem Competition: Developers and businesses will gravitate to platforms with stronger computing power and more stable services.
If Chinese giants don’t accelerate, American ones will, and global leadership in AI, standard setting, and talent attraction will shift to them.
This is a “Game for the Brave”: The first to stop investing might fall behind in the next technological iteration. Therefore, even if they have funds now, they need to secure future resources to stay in the game.
---
5. Why Now? While Funds Are Still Easy to Obtain
The final key point is the changing financing environment. In the first half of this year, the capital market was extremely enthusiastic about AI, making it easy to borrow money and issue stocks. But in the second half, the market has become more cautious:
- Can high valuations be realized?
- How long will it take to recoup such large investments?
- What if AI commercialization doesn’t meet expectations?
The giants’ strategy is to act while the market is favorable: Alibaba raised 80 billion RMB in shares when its stock price was relatively high and market sentiment was good. ByteDance expanded its loan from 20 billion to 29.6 billion due to strong interest from banks, indicating excellent credit and low financing costs. Tencent issued 30-year bonds to lock in low-interest rates.
Why Can’t They Wait? Many other companies also need funds:
- Chip manufacturers (like NVIDIA and AMD) need to expand production.
- Cloud providers (like AWS and Azure) need to build data centers.
- Large model startups (like OpenAI and Anthropic) need to spend heavily on training.
- Traditional industries (like automotive and finance) need to integrate AI.
When the entire industry requires tens of billions or trillions of RMB, capital will become scarce. By borrowing now, these companies can secure a competitive advantage in the future.
---
Summary: The Survival Rules of the AI Era
This wave of financing is, on the surface, about seeking funds, but in reality, it’s about prepaying for the future.
- Technology determines whether you can participate in the competition.
- Your financial strength determines how long you can stay in the game.
The giants are using today’s debts to secure tomorrow’s dominance in computing power. They understand the math clearly:
- Short term: Burning money may pressure profits.
- Medium term: Renting out computing power to cover costs.
- Long term: AI will become a core business, and computing power will be a key competitive advantage.
For everyone, it’s important to understand this: Don’t just look at the giants’ current profit statements; also examine their balance sheets and cash flow structures. They’re not spending money recklessly but making highly precise, risky, but potentially profitable strategic investments.
This AI race has just begun, and the winner will often be the one who understands both technology and finance and can stockpile resources when costs are low.