Big tech giants are utterly fed up with AI.

Recently, major tech giants have all gone frantic, borrowing money and raising capital at an unprecedented pace.

These are top-tier industry behemoths we are talking about, each holding hundreds of billions to trillions of yuan in cash on their books, yet they are acting in perfect unison, taking out loans and launching financing rounds.

On September 3, ByteDance was exposed to have finalized a USD 29.6 billion syndicated loan.

On August 23, Alibaba announced a private placement on the Hong Kong Stock Exchange, claiming to raise HK$800 billion for AI-related investments.

Tencent has not launched external financing yet, but its capital expenditure in the first quarter stood at around 31.9 billion yuan, and reached 2.784 billion yuan in the second quarter, up 176% year on year. Its total capital expenditure in the first half of the year hit 84.7 billion yuan, directly pushing its free cash flow into negative territory.

……

Why is this happening? Have all these giants lost their minds? Are they not afraid of suffering huge losses by pouring such massive resources into AI? Aren't they worried that this is an enormous bubble?

01

The Inevitability of Capital Burning

A funny thing has recently taken place in the capital market: Microsoft declared that it will scale down its AI investment in 2027, cutting its capital expenditure to 175 billion US dollars. What was the result? Its stock price surged 15.5% in a single trading day, adding 480 billion US dollars to its market value overnight.

The richest company in the world announced that it will no longer burn money on AI so recklessly, and the market celebrated as if it was a grand festival.

This is extremely surreal. Because the market has already started to worry that the capital expenditure of these tech giants is far too large.

Google, the world's top internet company, recorded a negative free cash flow of 5.855 billion US dollars in the second quarter of 2026. This marks the first time its cash flow has turned negative since the company went public.

Where did all the money go? It was all poured into AI.

Meta's free cash flow in the second quarter was only 784 million US dollars, plummeting 91% year on year and hitting a new low in four years.

Both Google and Meta are borrowing money frantically in the capital market, with each loan reaching hundreds of billions of US dollars.

The reason why these giants are acting so aggressively is that they are collectively gripped by anxiety, believing that they can never go back to the old days when they could rake in easy profits without much effort.

In 2010, Nokia's profit was at its historical peak. Back then, Nokia and Apple together took more than 80% of the profits in the global mobile phone market. Nokia had been the profit-making machine of the mobile phone industry for over a decade.

And then? It vanished from the market.

In 2011, it placed a big bet on Windows Phone. Less than a year later, its smartphone sales collapsed. Two years later, its mobile phone business was sold to Microsoft at a low price of 5.44 billion euros.

It took only three years for an industry empire to go from its peak to being sold off at a bargain price.

What about an empire like Microsoft? The same thing happened: it repeatedly missed historic opportunities.

Microsoft released Windows Mobile back in 2000, with the idea of stuffing PC software into mobile phones. It did not wake up to the reality until the iPhone was launched in 2007.

It launched Windows Phone in 2010, but its global market share was only 3.7% in 2013, while Android took 79% and iOS took 13%. It spent 7.2 billion US dollars to acquire Nokia's mobile phone business, but still failed to turn the tide.

Steve Ballmer himself admitted: "We failed to regard the mobile internet as a new core growth driver, and we missed this golden opportunity."

Do these giants lack money? Do they lack technology? Do they lack talents? None of them. What they lack is the insight into the new era, and the courage to abandon the old world.

This is the cruelty of paradigm shift: the certain death brought by inaction is far more unacceptable than the uncertainty brought by taking action.

This is exactly why today's tech giants are trapped in such a painful situation.

They are terrified of missing this era.

Death caused by inaction is certain, while the uncertainty of taking action at least leaves a way to survive.

02

The Distorted Business of Computing Power

The reason why tech giants are investing heavily in AI is also that the computing power business has turned into a game where you get eliminated if you do not buy in.

In the CPU era, there was an iron rule called Moore's Law: chip performance kept improving while prices kept declining. With the same amount of money, you could buy better and better computers. The larger the scale, the lower the cost.

But in the AI era, this rule has been completely rewritten by geopolitics.

Since October 2022, the United States has begun to impose export controls on high-end chips such as A100 and H100.

The entire supply of AI computing power resources has been artificially tightened for political reasons.

There is only one result: prices have skyrocketed.

Once the supply is tightened, prices go completely wild. Take the H100, a chip released four years ago, for example: its one-year lease price rose from 1.70 USD per GPU per hour in October 2025 to 2.35 USD in March 2026, up 40%. By August 2026, the price has reached nearly 3 USD, up another around 50% in half a year. The launch of new chips did not bring down the price of old chips, instead the price went up, which completely broke the GPU depreciation rule.

The domestic spot market is even more exaggerated.

The monthly rent of H200 rose from around 60,000 yuan to over 100,000 yuan; the spot price of B300 complete machines soared from around 4.5 million yuan to more than 13 million yuan, nearly tripling in half a year. The overall quotation of Nvidia chips and servers was raised by 5-15%.

Let me ask you, have you ever seen electronic products that become more valuable as time passes? That is exactly what the computing power market looks like right now.

On one side there is the expectation of surging demand, on the other side there are export controls, TSMC's production capacity bottlenecks, and constraints on power, land and infrastructure.

The supply cannot be expanded at all.

Therefore, as long as the expectation of surging demand and insufficient supply remains, computing power will be a business that guarantees stable profits in the short term.

A large part of the AI investment of tech giants is spent on purchasing computing power, and this business is extremely profitable at the moment.

Elon Musk's xAI company built Colossus, the world's largest supercomputer with 100,000 H100 chips in the first batch, in 122 days in Memphis. Now, the number of GPUs has expanded to around 220,000 to 230,000 units.

He leased Colossus 1 to Anthropic, what is the annual rent? 5 billion US dollars!

Meta also leases data center computing power to Anthropic, with a maximum value of around 10 billion US dollars for a two-year contract.

Some financial institutions estimate that Meta's computing power business will reach an annual revenue of 10 to 20 billion US dollars by the end of 2027.

Do you get the point?

Elon Musk and Meta are both losers in the large model competition, but they are not losers in business. They make huge profits just by leasing graphics cards and computing power centers.

How could Chinese tech giants fail to figure out this simple account? The graphics cards they bought a few years ago are now generating extremely high returns. Being a computing power landlord to collect rent is a huge business opportunity right now.

03

A Business Without a Moat

Technology companies are essentially capital goods businesses by nature.

What are capital goods? They refer to assets such as chips, computing power, and model training facilities. Their value depreciates rapidly as technology iterates.

Therefore, this type of business has no option to just maintain the status quo.

Who can win the market depends on the speed of your investment.

If you do not continuously invest in R&D, you will be left behind by the speed of your competitors.

In recent years, you can see how fast the entire AI market is changing! ChatGPT defined conversational AI, then Claude came out. Anthropic's revenue in the second quarter of 2026 was around 11.6 billion US dollars, surpassing OpenAI for the first time. Then DeepSeek disrupted the market with low-cost solutions.

No one can win permanently just by relying on one single model.

Therefore, tech giants are forced to keep burning money. Because once you stop, you are out of the game.

This is the future arms race. I may not know what will happen in the future, but I know for sure that if you do not invest, you will definitely lose.

04

Tech Giants Have No Management Advantage, Only Capital

Facing the new era, large enterprises not only have no management advantage, their existing systems even become a burden.

Even a strong company like Microsoft fell completely behind in the internet era. Nokia and BlackBerry held huge market shares, but they were still disrupted by startup competitors.

Why are startups faster?

OpenAI, Anthropic, DeepSeek, xAI all started as startups. They have short decision-making chains, dare to place big bets, and can adjust their strategies quickly.

What about the big tech giants? They have huge organizations, KPI-oriented mechanisms, and vested interest groups that hold back transformation. Google's search advertising business and Meta's social advertising business are their core foundations that they dare not easily touch.

Therefore, the management system of tech giants is not an advantage but a burden when facing the new paradigm.

But tech giants have an advantage that startups can never match, that is capital.

The four major cloud vendors have a total annual capital expenditure of 750 billion US dollars, and Goldman Sachs expects the number to reach 1.2 trillion US dollars in 2027. This is a number that startups can never even dream of.

The endgame option for tech giants right now may not be winning the large model war, but building computing power infrastructure with capital to collect rent.

If you cannot be a gold digger, then become the mine owner.

Just like in the gold rush, no matter who digs up the gold, the shovel seller always makes money.

So you ask why tech giants are so obsessed with AI? Why are they so anxious?

AI is that catfish that wakes all these giants up from the old world where they could collect rent easily and enjoy extremely high certainty, forcing them to take things seriously.

In the past, Google could make steady profits just by holding onto its search advertising business. Meta could count money lying down just by holding onto its social advertising business. Microsoft could rest easy just by holding onto Windows and Office.

Alibaba, Tencent and ByteDance all have a large number of hit products, and they made huge profits easily in their respective leading fields.

What about now? They are confused, not knowing what will happen in the future.

They have to pour real money into a new world where no one knows how to make profits. If they do nothing, they get eliminated. If they do it, they may also lose all their investment.

Some people say this is a bubble, but can the owners of these big companies afford not to take the risk? What if this is not a bubble? What if this is the next era?

Anyway, get on the train first, grab a ticket, and at the very least, become a computing power landlord.

This article is from the WeChat official account "Kongfu Finance" (ID: kongfuf), author: Liusha, published with authorization from 36Kr.