Three days ago, Alibaba released its latest financial report.
AI Cloud and computing services revenue grew 45% year-over-year.
In the same quarter, capital expenditures approached $10 billion.
One might expect that after spending so much, the company would pause to assess if the investments pay off.
But Alibaba is doing the exact opposite today.
It announced plans to raise:
HKD 80 billion.
Approximately:
$10.2 billion.
And the company made it very clear:
100% of the net proceeds will be invested in AI.
The truly noteworthy point here is not just "Alibaba is spending a lot again."
It's that the AI competition has introduced a new barrier:
You not only need models but also the ability to continually raise capital
Exactly how much is Alibaba aiming to raise?
On August 23, Alibaba officially announced plans for a new share placement in Hong Kong, with a total offering amount of:
HKD 80 billion.
Approximately:
$10.2 billion.
Reuters noted that if completed, this would be the largest primary follow-on offering in Hong Kong's history and the third largest globally in 2026, trailing only Alphabet and Intel.
However, a key detail must be emphasized:
This is currently a proposed placement.
Meaning:
The plan has been formally announced,
but the transaction is still subject to market and other conditions.
Alibaba explicitly stated:
There is no guarantee the placement will be completed.
So, it's inaccurate to say:
"Alibaba has secured $10.2 billion."
A more precise description is:
Alibaba intends to raise $10.2 billion.
The most remarkable part isn’t the amount, but that 100% goes to AI
Companies usually allocate raised funds to various purposes, such as:
Debt repayment.
Mergers and acquisitions.
Working capital.
Investing in new markets.
But Alibaba clearly stated this time:
100% of net proceeds will be invested into full-stack AI capabilities.
That means:
Full-stack AI.
This term is critical because it indicates Alibaba isn’t just preparing to:
Train the next Qwen model.
But rather to invest across the entire AI chain.
What does “full-stack AI” mean?
At the base level:
Chips
Alibaba has its own T-Head chip business.
Moving up the stack:
Computing power and cloud infrastructure
This involves numerous servers, data centers, networks, and compute capacity.
Further up:
Foundation models
For example, Qwen.
And higher still:
MaaS and applications
Enabling enterprises and consumers to actually use the models.
Alibaba’s annual strategy treats chips, cloud, foundation models, MaaS, and applications as a single AI stack.
So this isn’t about:
"Raising another $10 billion for a bigger chatbot."
But rather about:
Controlling the entire chain from chips to user entry points.
Why doesn’t Alibaba just buy Nvidia chips?
As AI scales up, the competition extends beyond:
Model benchmarks.
To include:
Computing cost.
Supply stability.
Inference cost.
Cloud customers.
Developer ecosystems.
Model usage volume.
If a company only owns models but must continuously buy expensive compute from others,
long-term profit margins and supply chains may be controlled by others.
Alibaba’s strategy is more like:
Building its own comprehensive AI factory.
Chips form one layer.
Cloud is another.
Qwen model another.
Enterprise and consumer applications represent yet another layer.
But didn’t it just spend a lot recently?
Exactly.
On August 20, Alibaba disclosed that its latest quarterly capital expenditure neared:
$10 billion.
A 75% increase year-over-year.
They stated this reflects ongoing expansion of AI investments to meet rapidly growing customer demand.
In other words:
Today’s planned $10.2 billion fundraising
roughly equals:
Another quarter’s worth of capital expenditure.
This reveals just how capital-intensive the AI battle currently is.
Is Alibaba seeing revenue from AI yet?
Yes.
This is a key reason they feel confident to continue investing.
In the latest quarter, Alibaba’s AI Cloud and Compute Services revenue reached:
$7.1 billion.
A 45% year-over-year growth.
The fastest cloud growth speed in 22 quarters.
Related AI product revenue was around:
$1.8 billion.
Maintaining triple-digit year-over-year growth for 12 consecutive quarters.
So Alibaba’s signal is not:
"We’re spending but no one is using it."
But rather:
Demand is truly increasing.
This is the crucial point tonight
If AI Cloud wasn’t growing,
raising another $10 billion today would make markets question:
What exactly are you burning money on?
But if enterprises are actively renting:
GPUs.
AI Cloud.
Model services.
Agents.
Alibaba’s logic becomes:
The biggest risk is not spending too much.
But that demand truly arrives and we don’t have enough compute to sell.
This contrasts with typical consumer spending logic.
AI infrastructure is tricky: you must spend first to serve customers later
Imagine running a restaurant.
If customers suddenly increase, you can:
Buy more ingredients.
Hire more staff.
AI data centers aren’t so simple.
You must first prepare:
Land.
Power.
Server rooms.
Chips.
Memory.
Networking.
Liquid cooling.
Servers.
Many of these can’t be ordered today and scale up instantly by tomorrow.
So large AI firms face:
Capital expenditures must precede revenue.
This is why AI increasingly resembles heavy industry
We usually think of software companies as:
Write code once,
then a million more users incur very low marginal cost.
AI isn’t exactly like that.
A million users mean:
More generation.
More inference.
More agents running.
Which requires:
More tokens.
More GPUs.
More power.
More data centers.
So today’s large AI companies may appear as software firms,
but their capital structure is more like:
Infrastructure companies.
Even “free models” aren’t cheap behind the scenes
Qwen’s open model approach helps accumulate developers quickly.
But when users integrate Qwen into:
Enterprise systems.
Apps.
Agents.
Large-scale services,
They still need:
Compute.
Alibaba’s strategy aims to:
Ultimately have more model users renting more compute on Alibaba Cloud.
Company data shows the Qwen series has been downloaded over 3 billion times, with more than 300,000 derivative models; Alibaba explicitly regards this open-source ecosystem as a key path to Cloud monetization.
So "free models" might actually be just an entry point
This is similar to traditional software strategies.
Let users:
Use the software.
Develop on it.
Form habits.
Then the actual monetization likely happens through:
APIs.
Cloud services.
Compute power.
Enterprise services.
Agents.
Deployment.
If Alibaba succeeds,
Qwen won’t need to generate profit on every download.
The critical point is:
How much workload in the ecosystem eventually runs on its infrastructure.
But the biggest risk on this path is the upfront cost
Alibaba’s latest quarter saw net profit drop by:
75% year-over-year.
Reuters noted the increase in AI capital expenditure is a key backdrop.
This doesn’t mean:
AI investments have failed.
But it reminds us that:
AI growth isn’t free.
A company may simultaneously experience:
Rapid AI revenue growth.
Even faster capital expenditure growth.
Short-term profit pressure.
All three can happen at once.
The real questions investors must ask now aren’t “Is AI growing?”
They are:
First
Can AI Cloud growth be sustained?
Second
How much capital is required to generate each additional dollar of AI revenue?
Third
Can new data centers and chips maintain sufficient utilization?
Fourth
Can model usage eventually convert to Cloud revenue?
Fifth
How long will it take to recoup invested capital?
These will become the true AI business KPIs in the coming years.
Alibaba itself is starting to calculate "payback period"
Reuters cited the company saying it expects the payback period for AI-related investments to be:
Originally about:
3 years.
Now expected to shorten to:
2.5 years.
Due to rapidly increasing demand.
If achieved, a $10 billion capital expenditure could translate into future revenue capacity.
If not, it would merely represent:
Very expensive data centers.
So the real test is never:
How much you build.
But rather:
How many users actually use it.
This logic is very similar to Amazon’s AI investment approach
Amazon also significantly ramped capital expenditure this year.
The market has come to accept this because
The key is not:
“Spending a lot is impressive.”
But because AWS growth shows:
Customers truly pay to use the infrastructure.
Alibaba’s story today follows the same logic:
Capital expenditure is only investment if it converts to long-term demand.
Otherwise, it’s just cost.
But Alibaba faces an additional factor in China’s AI competition
U.S. tech giants include:
Microsoft.
Amazon.
Google.
Meta.
China’s AI battlefield involves:
Alibaba.
ByteDance.
Tencent.
Along with various model and chip companies.
If Alibaba only focuses on Cloud,
model access may be grabbed by others.
If it focuses only on Qwen,
compute power might be controlled by competitors.
The true purpose of “full-stack” is:
To prevent the most crucial layers of the AI value chain from falling into competitors’ hands.
This also means AI company moats are changing
Previously, it might have been:
Who has the best model.
Going forward, it may be:
Who has models,
And chips,
Data centers,
Enterprise customers,
Developers,
Product entry points,
And capital.
The AI competition might no longer be:
A model ranking contest.
But more like:
An industry chain endurance race.
The most notable number might not be $10.2 billion
But:
100%.
Alibaba is willing to say:
All net proceeds from this equity fundraising
Will be invested in AI.
This means AI is no longer just:
A new division.
Or an experimental project.
It’s becoming:
The core of capital allocation.
But shareholders are paying a price
The company’s new equity issuance
Increases share count,
So existing shareholders face:
Dilution.
This isn’t:
Alibaba getting $10.2 billion for free.
The market is essentially:
Providing more capital today,
In exchange for a chance the company builds a bigger AI business tomorrow.
Whether this is worthwhile depends on:
Revenue.
Cash flow.
Return on investment.
This marks the biggest difference from the August 21 news
On August 21, we saw:
How much Alibaba has already spent on AI.
Today, we see further:
Where Alibaba plans to find its next round of capital.
When a company needs to invest:
Hundreds of billions of dollars
Or even more over several years,
One of the most important competitive strengths becomes:
The ability to access capital.
Great technology,
But no access to sufficient chip and data center funds,
May hinder growth.
The next phase of AI may eliminate not the worst models
But:
Those who can’t afford to maintain models.
Because better models mean more usage,
And thus much greater infrastructure demands.
So major AI players must combine:
Technical capability.
Customers.
Cash flow.
Capital market credibility.
Supply chains.
Energy.
This is no longer:
A contest decided by a few dozen engineers building a model.
Tonight’s key takeaway
The most important message from Alibaba’s fundraising is not:
Another tech company spending $10 billion.
But rather:
AI competition has officially entered a "capital endurance race."
Model capabilities decide whether you can enter the race.
Customer demand decides whether you have revenue.
But whether you can continuously build:
Chips.
Compute power.
Data centers.
Models.
Applications.
Determines whether you can compete for many years.
The strongest AI companies in the future may not be those with the best models,
But:
Those able to continuously convert massive capital into real customer demand and cash flow.
Today, let’s advance a bit with AI.
Learn an AI skill every day.
Save a little time daily.
Improve your abilities incrementally.
SasaDaily, growing with you.
Recommended Reading
Amazon Prepares to Invest $220 Billion in AI Over a Year: Why the Market Is Starting to Be Confident
