斜杠中年斜杠中年AI × 沟通 × 商业 × 人生 × 健康
AI and Business Leverage

Tencent Said 'Life or Death.' Baidu's AI Revenue Crossed 53%. The Chinese AI Race Just Got Brutal.

2026-11-19Updated: 2026-11-1912 min readWesley Chong
#Tencent#Baidu#Chinese AI#DeepSeek#AI business strategy
Tencent Said 'Life or Death.' Baidu's AI Revenue Crossed 53%. The Chinese AI Race Just Got Brutal.|AI and Business Leverage 封面图

Summary

On November 16, 2026, two of China's largest technology companies sent signals from opposite ends of the AI race. Tencent published an internal speech calling the AI revolution a "life or death" matter. Baidu released quarterly earnings showing AI revenue has been more than half of its core business for three straight quarters.

On the same day, DeepSeek — the startup whose model is the most-used AI model on the planet — started throttling free users because it ran out of compute.

The Chinese AI race is no longer about who is participating. Everyone is. It is about who survives.

Tencent's "Life or Death" Moment

At Tencent's internal LinkTime conference, the company's top leadership (referred to internally as the "General Office" or 总办) published a set of notes that were unusually blunt for a company known for its cautious public communications.

Key quotes from the notes:

  • "The intelligent revolution now underway — if you don't embrace it, you will be disrupted. This is a life or death question."
  • "We must reset our mindset to zero. Every business must re-entrepreneur for the AI era."
  • "The wave has already hit. Some core businesses look stable, but AI is progressing so fast that distant waves are already swallowing hills one by one."
  • "In a sense, Tencent's 28 years of existence were preparation for today — so we can invest in a technology trend that changes the world."

The language — "life or death," "get on the boat to fight," "break down the silos" — signals a level of organizational urgency that Tencent has not publicly displayed before. This is a company with over 100 billion RMB in annual profits that is describing itself as existentially threatened.

Baidu's AI Revenue Milestone

On the same day Tencent was sounding the alarm internally, Baidu was reporting results externally. Its Q3 2026 financials told a story of a company that has already crossed the AI Rubicon:

  • Total quarterly revenue: 30.8 billion RMB (down 1% YoY)
  • Core business revenue (general business): 25 billion RMB (up 1% YoY, beating expectations)
  • AI business revenue: 13.2 billion RMB — 53% of core business revenue
  • AI cloud infrastructure revenue: 8.2 billion RMB (up 94% YoY, accelerating from 50% last quarter)
  • AI application revenue: 2.6 billion RMB
  • AI-native marketing service revenue: 2.4 billion RMB
  • Net profit (Non-GAAP): 2.4 billion RMB

AI has been more than half of Baidu's core revenue for three consecutive quarters. The AI cloud growth rate nearly doubled year-over-year and accelerated from the prior quarter. This is not a company experimenting with AI — it is a company whose revenue structure has fundamentally shifted.

At the Baidu World 2026 conference, the company also launched a "Token Marketplace" (词元所) for trading model services. Over 100 suppliers applied to join and over 1,000 customers connected on launch day.

DeepSeek's Capacity Crisis

While the giants were reporting and speechifying, the startup that has been eating everyone's lunch was running out of compute. DeepSeek began implementing usage restrictions on its App and web versions:

  • Long conversations inactive for more than 7 days are no longer supported
  • Single input text limited to approximately 40,000 characters (down from previous limits)
  • Daily conversation quotas implemented — users hitting the limit see "Server busy, usage quota reached"

DeepSeek V4.1 Flash has been the #1 most-used model on OpenRouter for four consecutive weeks, consuming 59.2 trillion tokens in a single week. The throttling is a clear signal that demand has outstripped the company's compute supply — a problem that competitors with deeper pockets (like Baidu's 94%-growing AI cloud) are happy to have.

What These Three Signals Mean Together

Read together, these three stories from November 16 tell you where the Chinese AI market is:

1. The incumbents are scared. Tencent's "life or death" language is not marketing — it is the internal acknowledgment that a 28-year-old internet giant could be disrupted if it does not move fast enough on AI. When a company with Tencent's resources and reach uses survival language, the threat is real.

2. The transition is already revenue-bearing. Baidu's 53% AI revenue share is proof that the AI transition is not a future bet — it is a present reality. Companies that moved early and built full-stack AI infrastructure (chips, frameworks, models, applications) are now seeing the financial returns.

3. The demand is overwhelming supply. DeepSeek's throttling shows that even the most popular model in the world cannot keep up with demand at current compute levels. The bottleneck is not adoption — it is infrastructure. This is why companies like Alibaba and Tencent are spending unprecedented amounts on AI capex (Alibaba and Tencent together spent over 120 billion RMB in Q2 2026 capex alone).

The Practical Implications

For businesses and individuals watching from outside China:

  1. Expect prices to stay low or go lower. The competitive intensity means Chinese models will continue to undercut on price. This is a structural feature, not a promotional phase.

  2. Expect instability. DeepSeek's throttling is a preview. As demand grows faster than supply, even popular services will have outages, rate limits, and capacity constraints. Build redundancy into your AI workflows — never depend on a single provider.

  3. Watch the organizational signals. When a company like Tencent uses "life or death" language internally, it means aggressive product launches, acquisitions, and strategic pivots are coming. The next 6-12 months will see Tencent push AI into every product surface — WeChat, Tencent Docs, Tencent Meeting, and enterprise tools — with a urgency that was absent before.

  4. The full-stack advantage is real. Baidu's numbers prove that owning the entire stack — from AI chips (Kunlun) to deep learning framework (PaddlePaddle) to models (ERNIE) to applications — creates compounding revenue. Companies trying to compete with just a model and an API are at a structural disadvantage.

The Chinese AI race in late 2026 is not a friendly competition. It is a war of attrition where the giants are spending billions, the startups are throttling users, and the language being used internally is "survival." For anyone building with AI, this intensity is actually good news — it means faster innovation, lower costs, and better products, as long as you build with redundancy and don't bet everything on one provider.

The last click belongs to a person — choosing which provider, which model, and which workflow to trust. But the infrastructure underneath is being rebuilt at a pace that makes every quarter look different from the last.

FAQs

What did Tencent's internal speech say?

On November 16, 2026, Tencent published internal LinkTime conference notes where leadership described AI as a 'life or death' issue, called for 'resetting mindsets to zero,' 'getting on the boat to fight,' and 'breaking down silos.' Every business unit was told to 're-entrepreneur for the AI era.'

How much AI revenue does Baidu generate?

Baidu's Q3 2026 AI business revenue was 13.2 billion RMB, accounting for 53% of core business revenue (general business). This is the third consecutive quarter AI has exceeded 50% of core revenue. AI cloud infrastructure revenue grew 94% year-over-year to 8.2 billion RMB.

Why is DeepSeek limiting usage?

DeepSeek restricted App and web version access due to insufficient compute capacity. Long conversations inactive for 7+ days are no longer supported, single input was capped at approximately 40,000 characters, and daily conversation quotas were implemented.

What does this mean for businesses using AI?

The Chinese AI race is intensifying, which means faster iteration, lower prices, and more capabilities — but also instability as providers hit capacity limits. Businesses should diversify across multiple model providers and build fallback routing into their workflows.

分享这篇文章 / Share Article
Wesley Chong

Author

Wesley Chong

Software developer, digital consultant, and Toastmasters speaker from Kluang, Malaysia.

Sharing practical ideas for AI, communication, business, personal growth, and healthy living.

Related Reading