China's wealthiest internet companies have all begun seeking external financing recently. Alibaba is raising capital through a share placement, Tencent has turned to bond markets, and ByteDance has secured a massive $29.6 billion syndicated loan from nearly 30 banks.

But the question is: are they actually short on cash? Based on their balance sheets, absolutely not. As of the end of June this year, Alibaba held roughly 474.5 billion yuan in cash and other liquid investments, while Tencent's official "total cash" figure stood at 511.2 billion yuan. ByteDance, despite not publishing financial reports, is also among China's most profitable internet firms.

So why are these cash-rich companies scouring the market for more money? The answer is simple: AI.

The latest entrant in this funding wave is ByteDance. On September 3rd, media reports indicated ByteDance initially planned to borrow $20 billion. However, after the news broke, nearly 30 banks from China, the US, Europe, and Singapore rushed to participate, with subscriptions exceeding $30 billion. ByteDance subsequently expanded the loan to $29.6 billion (approximately 200 billion yuan), making it Asia's second-largest dollar loan deal this year.

More notably, this substantial sum comes without collateral. One person directly involved in the transaction noted that an unsecured loan of this magnitude is extremely rare. Simply put, the banks are lending based on ByteDance's creditworthiness. In 2024, ByteDance borrowed $10.8 billion from about 20 domestic and international banks and lenders. Now, the loan size has nearly tripled in just two years, with even better financing terms.

What's the money for? ByteDance cites "general corporate purposes," but media reports suggest the funds will primarily support its AI expansion. Alibaba is taking a different route. In late August, it placed new shares in Hong Kong, raising HK$80 billion in one go, with a very explicit purpose: all proceeds will be invested in AI, mainly for expanding computing power, building large-scale AI data centers, and upgrading cloud infrastructure.

Tencent has headed to the bond market. In June, it issued $2.45 billion and 15 billion yuan in bonds, with some yuan-denominated bonds maturing in 2056 — a 30-year term. While Tencent hasn't earmarked the funds specifically for AI, its investment in AI infrastructure is clearly accelerating.

This trend of tech giants seeking external funds isn't limited to China. As of July 7th, Amazon, Alphabet, Meta, and Oracle had collectively issued approximately $194 billion in bonds this year alone, nearly 80% more than their full-year total of $108 billion in 2025. Meta alone issued $25 billion in bonds in late April. Oracle plans to raise $45-50 billion through bonds and equity this year to fund its expanding cloud and AI infrastructure.

From China to the US, more tech behemoths are proactively seeking external financing — not because they're broke, but because AI is exceptionally capital-intensive.

To truly grasp the scale of AI spending, one might visit Guangling County in Datong, Shanxi, where ByteDance has established its sprawling Taihang Computing Center under its Volcano Cloud arm. The second phase alone involves a total investment of 4.5 billion yuan, planning over 15,000 server cabinets, alongside a supporting 220kV power transmission project. What we see as AI — chatbots, instant video generation — translates on the ground into rows of data centers, server racks, and an immense support infrastructure.

While chips like Nvidia's H200 dominate headlines, acquiring them is only the first step. The real challenge lies in converting individual chips into stable, operational computing power. This year, sales of H200 chips to China have faced repeated approval and delivery delays. Even with permits, chips don't arrive instantly. Simultaneously, building data centers, deploying equipment, and getting them powered up takes time.

This means AI infrastructure investment is not only massive but also requires long-term planning and consistent commitment. As computing demand grows, capital expenditures are ballooning. Alibaba announced in early 2025 plans to invest at least 380 billion yuan in AI and cloud infrastructure over three years, spending about 67.7 billion yuan in Q2 alone.

ByteDance is even more aggressive. Media reports suggest internal discussions have proposed raising capital expenditure to as much as $70 billion (approximately 470 billion yuan) in 2026, focusing on data centers and other AI infrastructure. While the final number may be adjusted, it underscores the scale of this race.

This urgency extends to Zhang Yiming, ByteDance's founder. When he stepped down as CEO in 2021, he mentioned wanting to dedicate a decade to learning and research. But this July, he made a rare appearance at a Seed team meeting, explicitly stating not to rely on distillation from competitor models for short-term ranking gains, expressing willingness to sacrifice some short-term profits for long-term goals.

While Zhang preaches long-termism in model development, ByteDance is clearly accelerating on infrastructure. The logic is straightforward: a lagging model today can be iterated upon in six months. But if competitors have already installed GPUs and are running operations while your data center is still under construction or waiting for power, that gap is hard to close. In the AI era, infrastructure isn't just backend support — it's becoming part of the core business. Model capabilities, user capacity, and cost reductions are all directly constrained by computing power. To a degree, business ceilings are defined by infrastructure reach.

So while AI competition may span years, the race for infrastructure positioning is happening right now. What these giants are truly competing for isn't just chips — it's time.

This raises another question: why not simply use their own billions in cash? Because a company's cash serves multiple purposes — share buybacks, acquisitions, new business development, plus reserves for future risks and opportunities. Servers and data centers generate value over many years. Matching long-term investments with long-term funding is a natural choice. Tencent is a clear example. Though it hasn't tied its bond proceeds exclusively to AI, long-term funding clearly opens up more room for future large-scale investment.

Of course, these giants can spend freely because the economics are starting to make sense. During Tencent's Q2 earnings call, an analyst asked a pointed question: with capital expenditure annualized at over 200 billion yuan based on Q2's ~53 billion yuan, how would future depreciation and amortization impact profits, and how long would it take for AI-driven revenue to cover these costs?

Tencent's Chief Strategy Officer James Mitchell's response revealed the logic behind the spending. Under current computing demand and rental prices, if Tencent were to rent out its new computing capacity to third parties like emerging cloud companies, it could almost immediately cover depreciation and achieve decent returns. President Martin Lau added that some computing capacity ordered and prepaid months ago could now be resold for over 30% profit. This provides a "safety cushion" for AI capital expenditure: if internal AI business grows fast, use the capacity internally; if external demand is stronger, monetize it via cloud services. Computing power is transforming from a pure cost into a revenue-generating asset.

Alibaba is doing similar math. In August, CEO Eddie Wu stated that based on current average gross margins of AI products, AI-related capital expenditure could be recovered in roughly three years, potentially shortening to 2.5 or even 2 years with improved margins and operational efficiency. He also noted that A100 chips purchased in 2020 and V100s from 2018 are still running near full capacity, far exceeding their theoretical depreciation schedules.

Revenue is catching up too. Alibaba's AI-related product revenue has grown triple-digits for 12 consecutive quarters, now annualizing at over 49.5 billion yuan. In Q2, Kuaishou's Kling AI generated over 850 million yuan in revenue. These figures don't yet prove that billions in AI investment have fully paid off, but they show that giants can now calculate AI's future beyond speculation.

A new cycle is forming: more computing power enables better models and more customers; revenue growth builds confidence for continued investment and easier access to future funding. Money becomes computing power, computing power drives business, and business supports the next round of investment.

However, market sentiment is shifting. In H1, capital was enthusiastic about AI globally. But entering H2, investors have grown more cautious, questioning lofty valuations and long payback periods. This context makes Alibaba's HK$80 billion placement noteworthy — it's not short on cash yet chose to raise funds now. Similarly, ByteDance's expanded syndicated loan and Tencent's long-dated bonds reflect a similar strategy.

For these giants, securing funding while it's readily available is wiser than waiting until they're desperate. Because these companies won't be the only ones needing massive capital. Platform firms, chip makers, cloud providers, and large-model companies will all require increasing investment. When everyone needs tens or hundreds of billions simultaneously, capital itself may become scarce.

These tech giants increasingly understand that the AI battle differs from the internet era. Technology determines whether you can sit at the table, but the depth of your pockets may determine how long you can stay there.