Model

Alibaba stops giving away its best AI model, three days before earnings

Alibaba launched Qwen4-Max on November 16 and kept the 2.4-trillion-parameter flagship behind its own paid API, promising open weights only for smaller models. Thursday's report is the first test of whether Qwen's three billion downloads become cloud revenue, and it lands two days after a US rule put the chips behind the model under license.

Richard TangRichard Tangthree-minute read
Share
Jack Ma, Alibaba's founder, seated on a white armchair at a conference, hand at his chin
Jack Ma, Alibaba's founder, at the Valdai Discussion Club in October 2017. Qwen4-Max, the company's flagship model, now sells only through Alibaba's own cloud and apps.

Alibaba−1.56% — Alibaba, down 1.56 percent today has stopped giving away its best model, and on both sides of the Pacific the new rules make keeping it closed the easy call. Qwen4-Max, launched on November 16, sells only through Alibaba's own cloud and apps at $1.50 per million input tokens 12.

It is a 2.4-trillion-parameter model, and open weights are promised only for the smaller Qwen4 models, so developers keep a free way in at the small end while the flagship becomes the upgrade they pay for 12. For a group whose AI labs unit lost RMB 13.9 billion in the June quarter while capital spending nearly quadrupled in two years, that upgrade is the payback line 1.

Revenue kept rising while operating cash flow halved

  • Revenue
  • Operating cash flow
$0B$20B$40B$60BQ4 '23Q2 '24Q4 '24Q2 '25Q4 '25Q2 '26Capex of $18.2B reported thatquarter
Data
RevenueOperating cash flow
Q3 '23$30.79B$6.74B
Q4 '23$36.63B$9.11B
Q1 '24$30.73B$3.23B
Q2 '24$33.5B$4.63B
Q3 '24$33.7B$4.48B
Q4 '24$38.39B$9.72B
Q1 '25$32.58B$3.79B
Q2 '25$34.57B$2.89B
Q3 '25$34.81B$1.42B
Q4 '25$40.72B$5.15B
Q1 '26$35.3B$1.37B
Q2 '26$39.59B$3.38B
Quarterly revenue and operating cash flow, calendar-labeled quarters; Alibaba's fiscal year ends in March. Alibaba reports capital spending unevenly across quarters, so capex is not charted here. Source: Alibaba SEC filings.7

Beijing made the closed model the compliant shape

China's commerce and science ministries issued Announcement No. 48 on October 16, the country's first outbound controls on AI technology 3. The draft catalog covers model weights at two trillion parameters or above, while API and SaaS access is exempt 3. Qwen4-Max is a 2.4-trillion-parameter model: its weights sit inside the restricted band, and the API it sells through does not 3.

The public consultation closed on November 15, and the flagship launched the next day 3. Per Bloomberg reporting carried by DIGITIMES, the new Qwen dropped its weight settings so users cannot download and retune them, in line with an October policy direction curbing models' cyberattack capabilities 4. Alibaba has not said whether revenue or compliance drove the closure, and both point the same way.

Washington licensed the compute it runs on

On November 16 the US Commerce Department issued an interim final rule: exporters need a license to ship advanced chips to data centers in Singapore, Malaysia, Thailand, Indonesia and Brazil, and to any data center selling large amounts of compute to China-headquartered firms, with 90 days to comply 5. Computing for commercial recommendation systems, the engine of Alibaba's shopping business, is exempt 5. The shares fell 15.6% that week to a 52-week low 5. The rule aims at the rented capacity behind the AI cloud and spares the profit pool that funds the buildout.

Thursday prices the new bottleneck

Demand is no longer the open question. Revenue growth has accelerated for four straight quarters, AI cloud revenue grew 45% in the June quarter, and management guided AI-related annualized revenue toward $10 billion for the September quarter it reports tomorrow 157.

Alibaba's revenue growth has climbed for four straight quarters

Growth hit 14.5% in the June quarter, the fastest in the three years shown

0%5%10%15%Q4 '23Q2 '24Q4 '24Q2 '25Q4 '25Q2 '2614.5%AI cloud demand, before the chiprule
Data
Revenue growth, % YoY
Q3 '235.9%
Q4 '233%
Q1 '241.4%
Q2 '243.8%
Q3 '249.4%
Q4 '244.8%
Q1 '256%
Q2 '253.2%
Q3 '253.3%
Q4 '256.1%
Q1 '268.3%
Q2 '2614.5%
Year-on-year revenue growth by calendar-labeled quarter; Alibaba's fiscal year ends in March. Source: Alibaba SEC filings.7

A print below 11.6% means the chips bite first

The test is the margin underneath: cloud adjusted EBITA margin reached 11.6% in June, and a print below that says the chip squeeze is arriving before the revenue is 1. Qwen4-Max launched after September closed, so its own revenue first appears in the December quarter; Thursday shows the cloud line around it 1. Consensus sits near $40.5 billion of revenue 6. The rule is interim, and data-center operators still have to enforce it by vetting their customers 5.

The question is no longer whether anyone will pay for Qwen. It is whether Alibaba holds licensed chips to serve them, and a rule that spares the shopping business makes that a narrower problem than this week's selling priced in.

Get Model by email, free

The AI labs, their models, and the business behind them. In our daily email, every weekday.

Loading BABA

Reader comments

0 comments

    Model

    HeyGen's Avatar 6 scored 84% on realism, but its 42.2% pick rate was below chance

    Avatar 6 launched November 17, 2026, with an 84% realism figure and first place in a blind arena; the score doubles a below-chance pick rate, and the arena capped 97% of clips at 35 seconds. That is the evidence base under a company claiming $200 million of annual revenue.

    Model

    MiniMax hit a record low, about 90% below its January peak, as a US rule gates its rented Nvidia compute

    Washington's new remote-access rule puts a US license between China's listed model makers and the offshore Nvidia capacity they rent. MiniMax, the purest listed bet on Chinese foundation models, holds the price tag.

    Model

    OpenAI cuts flagship pricing 20% to match Anthropic as heavy users call GPT-6.2 Astra unusable

    OpenAI cut its flagship's price to match Claude Fable 5.2's rate, and the users who run models hardest rejected it, 24 hours before Anthropic priced a $1.4 trillion IPO on the assumption the quality gap holds. The independent labs say the gap is real, for now.

    Model

    Microsoft books Azure revenue on Chinese open models running on Baseten hardware it does not own

    Microsoft now bills customers' committed Azure dollars for open models from Chinese labs running on Baseten's hardware, a toll booth earning rent on compute it does not own.

    Model

    Perplexity claims $1 billion of revenue on five times less spend than Anthropic's $1.4 trillion IPO

    Perplexity says it hit $1 billion in recurring revenue on a fraction of Anthropic's spend, three days before Anthropic prices a $1.4 trillion IPO on the opposite assumption. Whose economics hold will only be settled by a filing Srinivas now says could come in early 2027.

    Model

    Anthropic prices a $1.4 trillion IPO as Databricks logs show its costliest model at 0.6% of enterprise tokens

    Databricks' gateway logs show the costliest Claude tier at 0.6% of enterprise tokens and open-weight models near a third, hours before Anthropic prices a $1.4 trillion debut on premium-model revenue.

    Subscribe to TheInference for free