Inconsistent Kai-Fu Lee

"The Media Sends Its Regards"

"AI products launch daily, foundation models update monthly. Benchmarks and pitch decks multiply, invite codes and agents take flight together. Yet no one cares except AI content creators.

Does anyone still remember, before Hangzhou's Six Little Dragons, the previous generation of gods were the AI Six Little Tigers? 😭

In light of this, I declare this week Savage Roast Week. Going through the AI Six Little Tigers one by one, keeping it real.

One piece per day. Today is 01.AI, tomorrow is StepFun."

In May 2024, startup godfather Kai-Fu Lee set an extraordinarily ambitious goal for 01.AI: "If we're going to do this, we're building a trillion-dollar company," vowing to make it "the Microsoft of the AGI era."

It was a thrilling opening act.

Yet merely six months later, 01.AI splintered apart. Core technical talent was absorbed into Alibaba's Tongyi. The expensive dream of chasing AGI got packaged into an "Industrial Foundation Model Joint Laboratory" with Alibaba. The pre-training team received offers from Tongyi; the infra team received offers from Alibaba Cloud.

This is the godfather's final ripple 😭

The co-founders who've left Baichuan and 01.AI could be recited like a restaurant's menu: Zonghong Dai, Xuemei Gu, Xiangang Li, Wenhao Huang, Xin Pan, Dapeng Cao—one short of forming a boy band. Who knows whether co-founders have become worthless, or if they deliberately inflated their titles before exiting.

The godfather has mentioned on several occasions that large model companies will sooner or later face "the investment banker's interrogation" and "the soul-searching on business model."

He anticipated his own interrogation, so when media questioned him, he preemptively used the rhetoric of "changing lanes" to prove this wasn't a forced pivot driven by resource constraints, but a deliberate, active choice.

A master communicator can always find the most dignified framing for every strategic retreat.

In the first half of 2024, Kai-Fu Lee expressed firm rejection of B2B business. He saw the model as packaging systems integration companies into high-growth AI companies during the AI 1.0 era—essentially "low-quality revenue" where every deal loses money—and considered B2B "the curse of AI 1.0."

Teacher Kaifu's strategy was clear: overseas B2C applications and APIs. Because "the United States already had its ChatGPT moment," the market was educated and mature.

But by October 2024, 01.AI loudly launched its "AI 2.0 Digital Human" solution targeting the B2B market, signing deals with Yum China, Kidswant, and others. By January 2025, Kai-Fu Lee confirmed in an interview that of the company's 2024 revenue exceeding 100 million RMB, "around 70%" came from B2B.

"The curse of AI 1.0" became the company's primary revenue source.

Facing this reversal, Kai-Fu Lee's explanation was that what he opposed before was unprofitable systems integration. What they're doing now is a second growth curve that finds profit for clients, charges monthly recurring fees, and becomes a cash-printing machine for enterprises.

"If it's B2B where every deal loses money, 01.AI would rather not do it—that statement hasn't changed." The godfather truly is the ceiling of professional managers; any position can be spun.

As for the path of pre-training large models, 01.AI initially projected full confidence.

In October 2024, responding to rumors of "abandoning pre-training," Kai-Fu Lee personally debunked them. He assured the outside world that for China's top AI startups, doing pre-training "funding and chips won't be problems."

Yet three months later, Kai-Fu Lee admitted that internally, as early as May 2024, the company had begun reflecting on the Scaling Law path. He gave a very specific reason: "Finance colleagues review cash flow with me every few weeks, seeing compute costs month after month—it's really not sustainable." They abandoned the originally planned larger Yi-X-Large model, pivoting to small, beautiful, and cheap models.

The godfather also creatively proposed the concept of "frugal Scaling Law," arguing that "blindly insisting on things you can't afford is not the correct and healthy choice for a startup."

A strategic contraction driven by real resource constraints was once again successfully reframed as an active, wise lane change.

When discussing founding intentions, the godfather once quoted Steve Jobs, saying he "came into this world to connect these dots." He hoped 01.AI would carry the AGI dream, ensuring China wouldn't fall behind in the AI revolution.

But now, 01.AI's greatest value is contributing material for media coverage.

Great enterprises like OpenAI pioneer trends. First-class AI like DeepSeek benefits all people. As for companies with dim prospects, their greatest value is employees leaking daily to second-tier media.

The media teachers will miss the godfather's KPI contributions 😭

(Images in this article generated by ChatGPT o3, with writing assistance from Gemini 2.5 Pro.)