Alibaba reported RMB 284.8 billion in revenue for the quarter ended December 31, 2025, but GAAP net income plunged 66% as heavy spending on quick commerce and AI infrastructure eroded margins. The group rolled out a new world model called Happy Oyster that can generate 3D environments and interactive video, part of a wider product push from its Token Hub unit while cloud and AI investments aim at a five-year revenue target. Tencent, meanwhile, is framing AI as a broad enterprise pivot and is pouring capital into compute and models even as its disclosures show rising B2B revenue and large infrastructure bills. Investors will watch whether Alibaba’s fiscal 2029 quick commerce timeline and the group’s $100 billion cloud and AI target deliver a return on this spending.

Alibaba has taken the wraps off Happy Oyster, a world model designed to create realistic physics and spatial properties for 3D content. The company said the model can produce game worlds and simulated videos that previously required separate pipelines and specialist 3D teams, making it possible to build interactive experiences from a single model.

What Happy Oyster does and where it sits

Happy Oyster is the latest output from Token Hub, a recently formed unit within Alibaba focused on model development and commercialisation. Alibaba described the release as a step toward producing games, films and other spatial content from one underlying model, rather than stitching together different tools for physics, rendering and animation.

The group has also published a separate video-generation model called Happy Horse, which it said ranked at the top of global model benchmarks on debut. Both models were disclosed as incremental product reveals rather than full general releases, and Alibaba has made Happy Oyster available only on a limited early-access basis. The company hasn't defined a broader commercial rollout timetable in its posts or coverage, leaving developers and customers to await further announcements.

Alibaba explicitly positions these launches to grab developer demand for content generation and simulation. The Happy Oyster announcement places the company directly against Tencent’s publicly available Hunyuan3D series of world models. That sets up a head-to-head contest between two of China’s biggest tech groups over 3D world modelling for games, video and robotics training.

The price of building models at scale

Alibaba’s product push comes against a backdrop of slowing core commerce growth and sharply weaker profitability. In the quarter ended December 31, 2025, the group reported total revenue of RMB 284.8 billion, a 2% year-on-year increase that missed expectations.

Customer Management Revenue, a key e-commerce metric, grew just 1% year on year, down from double-digit growth in the prior quarter.

Mastheads

Profitability suffered. Alibaba’s GAAP net income fell 66% year on year. Non-GAAP diluted earnings per ADS dropped 67% to RMB 7.09. Adjusted EBITDA declined 57%, and free cash flow fell 71%. Management attributed much of that pressure to heavy spending on quick commerce and investments in AI infrastructure.

The company reported that its quick commerce unit expanded revenue by 56% but remains loss-making. Under current plans, Alibaba signalled that quick commerce may not reach its profitability targets until fiscal 2029. At the same time, the group is racing to monetise cloud and model technology, and has publicly set a target to increase annual cloud and AI revenue to about $100 billion within five years.

Tencent has taken a different public tone. Li Qiang, vice-president of Tencent and president of its enterprise business unit, told reporters that the firm sees AI reaching a threshold for industrial-scale adoption. Tencent said it's building full-stack offerings across energy supply, computing power, model capability and intelligent agents.

The company disclosed heavy capital expenditure on AI infrastructure of roughly 79.2 billion yuan and reported AI-related R&D spending of about 18 billion yuan in the prior year, with plans to lift that figure further. Tencent said its B2B revenue grew to 229.43 billion yuan in 2025, up 22% quarter-on-quarter, which it cited as evidence that enterprise AI is already reshaping its revenue mix.

Tencent also cited national figures in its commentary, noting about 602 million generative AI users in China by December 2025 and a dramatic rise in daily token usage from 100 billion in early 2024 to 140 trillion in March 2026. The company uses those figures to justify scale investments in models and compute.

There is, however, an evidentiary gap between the two narratives. Alibaba’s recent quarterly filing shows a clear near-term hit to margins and cash flow from AI and quick commerce spending. Tencent’s accounts and public comments frame AI investments as a deliberate pivot and point to quarter-on-quarter B2B revenue gains. The bundle of reporting doesn't include a comparable set of quarterly losses for Tencent that would mirror Alibaba’s decline, so balancing the upfront cost against future returns remains a point of contention among analysts and investors.

For now, both groups are accelerating product launches while reorganising teams. Alibaba’s Token Hub will continue sequencing specialised models such as Happy Horse and Happy Oyster to capture developer demand. Tencent is signalling continued rises in R&D and compute investment both domestically and abroad as it expands enterprise offers.

Developers, game studios and media producers will be watching the limited early-access programmes. Companies that get early runs will be best placed to shape tooling and pipelines. But the economics of building and operating models at the scale both firms envisage are already visible on the profit-and-loss lines.

Related Articles

Happy Oyster is in limited early access. Alibaba's aim to lift annual cloud and AI revenue to about $100 billion within five years, and its plan to reach quick-commerce profitability by fiscal 2029, are concrete milestones investors can use to judge whether the group's heavy AI and quick-commerce spending pays off.

This article was created with AI assistance.