The 4 a.m. alarm rings, the treadmill starts, and the emails go out before most of Beijing is awake. For decades this routine powered the search engines that shaped the modern internet. Now the same discipline is aimed at a very different goal: picking the next generation of Chinese AI builders.
A Shanghai boy who grew up without electricity
Lu Qi (陆奇) was born in Shanghai in 1961 and, during the Cultural Revolution, was sent to live with grandparents in a Jiangsu village without electricity or running water. He studied computer science at Fudan University, then earned a PhD at Carnegie Mellon in 1996 — arriving in Pittsburgh as only the second Chinese student admitted to its School of Computer Science.
The American climb
Lu's career reads like a tour of late-1990s and 2000s tech power. He spent a stretch at an IBM research lab, then rose through Yahoo! to become executive vice president of engineering for search and advertising technology. Steve Ballmer personally recruited him to Microsoft, where he helped drive the launch of Bing and later became an architect of the company's early AI and bots strategy. Fast Company once named him one of the ten most creative people in business.
The Baidu year
In January 2017 Lu returned to China as group president and chief operating officer of Baidu — the highest-profile western-trained operator the company had ever hired. He pushed the "All in AI" strategy and stood up the Apollo autonomous-driving group. It was a short, intense chapter: he stepped down in May 2018, citing health and family reasons, while staying on as vice chairman.
YC China, and the abrupt exit
In August 2018 Lu joined Y Combinator as head of its new China program. Barely a year later, in November 2019, YC announced it would shut down its China operations. Rather than walk away, Lu spun the effort into his own independent fund: MiraclePlus (奇绩创坛).
MiraclePlus (奇绩创坛) and the 7 percent bet
MiraclePlus runs two intensive cohorts a year out of Beijing. The model is deliberately YC-like: it writes an early check — roughly US$300,000 — for about 7% of each startup, then puts the founders through a three-month bootcamp before a demo day. By 2024 the program had backed well over 250 early-stage companies, with a large share in AI and large model (大模型) work.
Admission is brutal: thousands of applications per cohort, an acceptance rate reportedly under 1%. The firms Lu backs tend to be technical and young — the average founder age is around 29.
Betting early on robots
Lu's most interesting recent move is timing. He was slow to catch the first wave of Chinese large model (大模型) startups — the so-called "AI tigers" that formed in 2023 — and MiraclePlus missed most of them. But he had been quietly backing embodied AI and humanoid (人形机器人) startups since 2023, well before the category became crowded. As of 2024–2025 MiraclePlus was among the most active early-stage backers of embodied-AI and humanoid (人形机器人) companies in China, including bets on AgiBot (智元机器人) and a spread of simulation, dual-arm, and service-robot startups.
His thesis is that the next platform shift is "action" — autonomous driving, robots, spatial computing — not just language. Whether that call pays off is still open, but the early positioning is clear.
The honest part about timing
Lu is the first to live with the irony: a man who spent his career at the center of search and AI missed the headline generative-AI startup boom, then re-centered on robotics. Since 2024 he has also served as dean of the Shanghai Innovation Academy (上海创智学院), a role that mixes talent training with frontier-tech strategy. The through-line is consistent — he backs what he believes is the next paradigm, often before the market agrees.
Honest limitations
- The "over 250 companies" and "under 1% acceptance" figures come from Chinese tech and finance press and MiraclePlus' own pitching; we could not independently audit the portfolio count.
- Claims about MiraclePlus being a top robotics investor by deal count are reported by Chinese outlets (citing IT桔子 data) and are not independently verified here.
- Lu's exact stake, fund size, and returns are not public; this profile describes strategy and trajectory, not performance.
- We did not evaluate the success or failure of individual MiraclePlus portfolio companies.
What readers can do now
- If you're building in AI or robotics, study MiraclePlus' demo-day cohorts as a free, curated map of what young Chinese founders are actually attempting — especially in embodied AI and humanoid (人形机器人).
- Track Lu's public talks: his "world model" and "action is the next paradigm" theses are a useful counter-signal to pure large model (大模型) hype.
- For ecosystem watchers, compare MiraclePlus' seed-led model with U.S. accelerators to see how China's early-stage AI funding actually behaves on the ground.