Two young men from ordinary Guangdong families are rewriting the global AI landscape.

41-year-old Liang Wenfeng, from a teacher's family in Zhanjiang, built DeepSeek in Hangzhou after making his fortune through quantitative trading. 34-year-old Yang Zhilin, from a working-class family in Shantou, burst onto the global stage with Kimi/Moonshot AI.

The two Guangdong natives took very different paths, but in August both ignited the investment world.

DeepSeek launched a second funding round seeking 50 billion yuan, with a pre-money valuation up to 500 billion yuan. Moonshot AI secured a Pre-IPO round targeting a $50 billion valuation, about 340 billion yuan. Together the two are valued at over 800 billion yuan, roughly one and a half PetroChina.

AI wealth creation has shattered expectations. DeepSeek's valuation jumped from 350 billion to 500 billion yuan in two months, up 43%. Moonshot rallied from $4.3 billion to $50 billion in half a year, an increase of more than 11 times.

Liang Wenfeng's net worth soared from $1 billion to $39.5 billion in one year, an annual growth rate of 3,850%, ranking first globally. The next highest, from Anthropic's founding team, rose only 1,191%. Even top American AI talents cannot keep up.

Liang is extremely low-key: no social media, only a desk phone for work, and almost no personal traces online. Yang is outgoing and flamboyant, once forming a rock band at Tsinghua; Moonshot's meeting rooms and plans use music themes. Beneath these differences, both are pragmatic, clear-headed, and willing to bet big.

Neither stayed in Guangdong; one rooted in Hangzhou and the other in Beijing, competing globally.

Liang took the "make money first, then chase the dream" path. A top scorer, he entered Zhejiang University, started quantitative trading with 80,000 yuan, founded High-Flyer Quant in 2015, and surpassed 100 billion yuan in assets by 2021. He then built the supercomputer "Firefly No. 2", later DeepSeek's key asset. DeepSeek was founded in 2023 and for three years relied entirely on High-Flyer's profits. After it broke out, investors rushed in: Tencent invested 10 billion yuan, CATL 5 billion, JD.com and NetEase 3 billion each, while Liang kept key voting power.

Yang followed another road: prodigious talent backed by capital. Born in 1992 in Shantou, he learned programming from scratch in high school, won a provincial informatics first prize, and entered Tsinghua. Ranked first in computer science, he finished a Carnegie Mellon PhD in four years. His startup was smooth: a $200 million angel round within two months, Alibaba's $800 million, and later Tencent, Meituan, Xiaohongshu, Sequoia and IDG, totaling about $3.5 billion, topping China's large-model funding rankings.

One self-funded and steady, the other VC-backed and fast—both reached Chinese AI's first tier.

Now China is competing head-to-head with global giants. Kimi K3 beat Claude, GPT and other top closed-source models to top a major international programming benchmark—the first Chinese open-source model to do so. Musk wrote "Impressive" and said xAI aims to surpass Kimi. Wall Street felt the shock; U.S. tech stocks fell, with the market calling it the "Kimi moment."

DeepSeek's V4-Flash later hit 7.22 trillion tokens in weekly API calls, ranking first globally and forcing OpenAI to cut prices—proof that Chinese large models have reached a global large-scale adoption turning point.

Both have fallen before. Liang's crisis came from his old business: this July, High-Flyer had a drawdown of over 20% in a month, nearly breaking its five-year annualized 114% myth. The quant "money printer" faltered, pushing DeepSeek to raise outside capital. Yang's darkest moment was more dramatic: DeepSeek's explosion slashed Kimi's monthly active users, the old ad-burning model failed, and product, capital and public pressures hit at once. He halted ad spending, cut staff to 300, and focused on core R&D. Extreme contraction brought extreme breakout: K2 landed in six months, and K3 topped the world a year later.

This is not just a valuation carnival; it is an industry turning point. DeepSeek recently launched peak/off-peak API pricing and raised core model prices. Zhipu, Moonshot, MiniMax and others followed, and Kimi's paid membership sold out. The shift is driven by scarce computing power, so the vicious price war is ending.

The AI first half—"burning money for traffic and scale"—is over. The second half is a hard battle over technology, commercialization, and profitability.

The landscape is clear: DeepSeek has raised over 100 billion yuan and is the largest; Moonshot AI grew 11 times in half a year and is the fastest. They have become China's AI duopoly.

But the real test has just begun. Moonshot reports annualized revenue of about $300 million, implying a price-to-sales ratio of 105 times; it must grow at least 50% annually for a decade to justify that price. DeepSeek's monetization has improved after price hikes, but it still faces user complaints and declining cache hit rates.

From technology to commercialization, from valuation to real profit, the two AI giants are at a critical stress-test point. What is certain is that the AI second half will be a duel, and China's tech innovation golden age has only just begun.

Seller Home Comment

The rise of DeepSeek and Moonshot AI proves that AI requires deep technology and business insight. Cross-border sellers should watch how Chinese large models are applied in supply chain optimization, multilingual customer service, and intelligent product selection—these may become new levers for cutting costs and improving efficiency.

Source: E-commerce Pai
Original link: https://www.pai.com.cn/p/01m0rn97361k725bkwk443mmt2