Alibaba Turns a $60,000 Start into a $53 Billion AI Bet

Alibaba started with $60,000 that 17 people pooled in the apartment of Jack Ma, a former English teacher. Today the company plans to invest about 380 billion yuan, roughly $53 billion, in cloud and AI infrastructure over three years. Between those two figures lie 27 years, a defeated eBay, a record $25 billion IPO and an open conflict with Chinese regulators.

The marketplace that beat eBay

In 1999, Alibaba was conceived as a B2B platform where Chinese factories could find foreign buyers. The startup money was modest, but the idea hit a real pain point for exporters, who did not know how to reach customers abroad.

The company won the consumer market through Taobao, launched in 2003. The platform charged sellers no commission, while eBay, which had bought the local EachNet, made money on fees. The logic was simple: first gather sellers and buyers, monetize later. The payment service Alipay, with its escrow scheme, played a decisive role: money went to the seller only after the buyer confirmed receipt of the goods. In a country where people did not trust online payments, this removed the main barrier. eBay gradually ceded China.

In 2014, Alibaba listed on a US exchange and raised $25 billion. At the time, it was the largest IPO in history.

Investors and Beijing

The success was not smooth. In 2011, Ma moved Alipay out of the company's control, which triggered a conflict with shareholders, including Yahoo and SoftBank. The dispute was settled with a deal, but Ma's reputation as a hard-nosed negotiator stuck.

Pressure from the authorities proved more serious. In October 2020, Ma publicly criticized financial regulators, and soon afterward Beijing halted the IPO of Ant Group, Alibaba's fintech affiliate. In 2021, the company was handed an antitrust fine of 18.2 billion yuan, about $2.8 billion. Ma stepped out of the spotlight, and in 2023 Alibaba split into six business groups to make its structure more flexible and to prepare parts of the company for independent investment.

AI as a continuation of the old logic

Today Alibaba's main bet is its Alibaba Cloud and its family of in-house language models, Qwen, many of which are released with open weights. The idea is a familiar one: just as in 1999 the company earned money from other people's transactions, it now wants to earn money from how others build AI products. The models attract developers, while the revenue comes from cloud compute, tools and enterprise contracts.

This strategy runs into the compute bottleneck. US export restrictions block China's access to Nvidia's best accelerators, so Alibaba, like its competitors, has to develop its own chips and optimize models for the hardware available. The scale of the announced investment shows that the company expects to solve the problem with capital.

Key figures:

• 1999: 17 co-founders, $60,000 in startup capital
• 2014: $25 billion IPO, a record at the time
• 2021: antitrust fine of 18.2 billion yuan (about $2.8 billion)
• 2023: restructuring into six business groups
• AI plan: about 380 billion yuan (~$53 billion) over three years, announced in February 2025

What to watch next

For developers, the main practical takeaway is that Qwen remains one of the most accessible options among open models: they can be run locally or on your own infrastructure, without being tied to a single cloud. Before deploying, check the license of the specific version, since terms differ from model to model.

For companies that work with China or elsewhere in Asia, Alibaba Cloud is one of the few real regional providers with a ready-made set of AI services. For everyone else, the more interesting question is different: can Alibaba turn capital expenditure into cloud revenue faster than sanctions restrictions hit its access to chips? The answer will come from quarterly reports on Cloud Intelligence growth and the size of capital expenditure.