Alibaba is selling a working games studio to pay for Qwen

Alibaba is selling its gaming subsidiary, Lingxi Games, to the Asian private equity firm Trustar Capital, so it can put the money into its Qwen models and its data centres ^[1]. The detail that matters for anyone modelling AI infrastructure spend is that Lingxi was not a problem asset: its flagship title, Three Kingdoms: Strategy Edition, built with Japan's Koei Tecmo, has been a reliable earner ^[5].

The price is unsettled. Reuters put the deal at more than $2bn; Bloomberg put it at more than $1.5bn ^[2][3]. That half-billion-dollar gap ^[4] is between 25 and 33 percent of the transaction value depending on whose number you take ^[15], and neither side appears in a hurry to reconcile it, according to the account published by The Next Web ^[4].

The strategic logic is explicit rather than inferred. Chief executive Eddie Wu has spent the past year restructuring the company around two pillars, AI and cloud computing, while shedding what is deemed non-core, and gaming, profitable or not, sits in the second bucket ^[6]. Lingxi chief executive Zhou Bingshu said Alibaba was "handing over the business to allow it to focus more closely on its strategic priorities" ^[7]. Alibaba is targeting $100bn in AI revenue over the next five years ^[8]. Set the sale against that target and the proceeds are worth roughly 1.5 to 2 percent of it ^[16], which tells you the sale is not really about the cash. It is about removing a business that requires management attention and capital allocation that the AI plan wants for itself.

Two things follow for operators. First, the divestment is the tell, not the announcement. A firm with Alibaba's balance sheet chose to convert a working revenue line into cash rather than fund AI capex out of growth, and the same pattern is spreading across China's technology sector, where the era of collecting businesses in gaming, retail, logistics and media is giving way to pruning as those firms race to fund AI and data-centre spending they now treat as existential ^[11]. If you are benchmarking a rival's infrastructure budget against its operating cash flow, you are probably reading the wrong line. Check what it sold.

Second, the market rewarded the subtraction. Alibaba's Hong Kong-listed shares rose about 2.67 percent on the report ^[9]. That is a small move, but it removes the usual internal objection to selling a healthy unit, which is that investors will read it as distress.

The technical claims underneath all this are contested. Alibaba recently launched its largest AI model yet, extending the Qwen line, and has claimed performance comparable to Anthropic's technology ^[10]. Anthropic has accused Alibaba of running the largest distillation campaign against Claude ^[12], which sits awkwardly beside any claim of home-grown parity ^[13].

What to watch: whether the final price lands nearer $1.5bn or $2bn, since a sale closing at the low end of a spread that wide says something about how motivated the seller is ^[2][3][4]. Watch also which asset goes next, and whether Alibaba starts reporting AI revenue against the $100bn figure in a way that can be checked ^[8].