While in the West the focus is usually on OpenAI, Anthropic, or Google, in China another equally intense race is unfolding: among its three major tech giants—ByteDance, Alibaba, and Tencent—to become the “super gateway” to the next generation of digital services. According to an extensive report published on Sina, this is no longer just about language models: computing, products, distribution, talent… and ecosystem power are all part of the mix.
A symbolic appointment: “stopping training for the exam”
The story begins with a very specific scene: Yao Shunyu arriving at Tencent, an investigator known for co-authoring the ReAct (Reason + Act) approach, tasked with diagnosing why the Hunyuan model was not performing as expected. According to the report, Yao criticized the team for chasing ranking results at the risk of “contamination” in evaluations—improving test scores but being unstable in real-world applications.
Three strategies, three weaknesses
The narrative paints a clear triangle:
- Tencent: historically more cautious, with an edge in social traffic (WeChat) and gaming, but arriving late to the “pace” of consumer AI assistants.
- Alibaba: seeks to build a “full stack” narrative (models + cloud + chips). The report mentions an internal axis like “Tongyi/Cloud/T-Head” and a sustained investment in infrastructure.
- ByteDance: views AI as a “center of the world” opportunity and pushes product growth, leveraging its distribution machinery and viral formats.
It’s important to note: many financial and operational data points (budgets, internal metrics, campaign figures) are presented “according to the report’s sources” and may not always be independently verified. Still, they align with observable realities: Doubao (ByteDance) has been recognized as the most popular AI chatbot in China by active monthly users, according to international coverage.
The ignition point: DeepSeek and the Lunar New Year as the battleground
According to Sina, the “before and after” shift occurs when DeepSeek changes industry perception and prompts more aggressive reactions. From that point on, the rivalry shifts from labs to showcases: the Lunar New Year / Spring Festival becomes a peak moment for capacity, marketing, and service reliability pressures.
The report describes product moves and promotional efforts (incentives, integrations, campaigns) to capture users precisely when traffic surges; and the underlying logic is simple: if your assistant fails when the whole country is testing it, you lose momentum.
The real race isn’t just “who responds best”: it’s who has the computing power, data, and distribution
One of the most interesting threads in the article is that the “war” is fought across multiple layers:
- Computing (GPU/infrastructure): suggests that giants operate at scales that exclude startups except in rare cases.
- Talent: the battle for top profiles is almost like a “war of recruitments,” where timing is as important as salary: “Is there a seat?” to do influential work.
- Ecosystem: in Tencent, the multiplier effect of WeChat; in Alibaba, cloud leverage + a suite of services; in ByteDance, distribution and viral product learning.
The core message: a great model without channels, product, and stable infrastructure is merely a demo; with channels and infrastructure, it becomes a platform.
The uncomfortable question: More open or more closed?
The report closes with a reflection: AI can simplify everything into “a chatbox” capable of executing tasks… but this also concentrates control in fewer decision points (what you see, what you do, in what order). It’s compared to a new “Tower of Babel”: the promise of unification, with the risk of confinement.
via: Sina

