Alibaba unveiled a lower-cost AI model on Wednesday, giving investors a reason to examine whether the company can make its AI push more economically efficient.
Qwen3.8-Flash is a multimodal model aimed at coding and office tasks while reducing training costs.
Its launch came as Alibaba completed an HK$80 billion ($10.2 billion) share placement to finance AI expansion.
Shares plunged 8.5% on Monday after the offering was announced, before recovering 1.5% to HK$114.20 on Tuesday.
Qwen gets cheaper as Alibaba’s AI bill gets bigger
Alibaba’s June-quarter results explain why efficiency matters.
Net profit fell 75% year over year, while capital expenditure jumped 75% to 67.68 billion yuan as the company bought chips and expanded computing infrastructure.
Growth is visible, as AI Cloud and Compute Services revenue increased 45% to 48.44 billion yuan, while AI-related products delivered triple-digit growth for a 12th consecutive quarter.
Qwen3.8-Flash potentially attacks the other side of that equation: cost. More efficient training and inference could allow Alibaba to serve rising AI demand without expenses increasing at the same pace.
Alibaba says its Qwen family has surpassed 3 billion global downloads and generated more than 300,000 derivative models.
That ecosystem becomes meaningful if developers using Qwen consume more Alibaba Cloud infrastructure.
Analysts see efficiency as the route to better economics
Daiwa analyst John Choi believes the economics can improve.
Dow Jones said Choi expects losses at Alibaba’s AI Labs and Applications business to decline sequentially, helped by “better model training efficiency, improved inference efficiency and lower marketing intensity.”
Daiwa maintained a Buy recommendation and raised its US ADR target to $185 from $180.
That argument fits Qwen3.8-Flash. Lower model costs could support better margins while allowing Alibaba to keep competing aggressively on performance and price.
Citi analyst Alicia Yap has made the strategic case. In commentary cited by Bloomberg, Yap said long-term AI success will require enormous resources and a loyal customer base.
She argued that companies spanning chips, cloud infrastructure, models and applications are better positioned to lead.
One cheaper model will not repair the stock alone
The sceptical case has not disappeared, as Alibaba has completed a $10.2 billion equity raise after committing 380 billion yuan over three years to AI and cloud infrastructure.
Roughly 60% of the new proceeds will fund global computing infrastructure, with 40% going towards hyperscale AI data centres and cloud upgrades.
Shareholders are being asked to tolerate dilution, heavy capital spending and weaker near-term profits for larger cloud earnings later.
BofA said the placement could initially weigh on sentiment because of dilution, higher depreciation, uncertain elevated capex and possible additional financing. It nevertheless maintained a Buy rating and $172 target.
CGS International also retained Buy on Tuesday but cut its Hong Kong target to HK$185 from HK$209.
The post Alibaba unveils cheaper Qwen AI model: is this what its stock needed? appeared first on Invezz