A 2026 photo posted on X by Alexandr Wang brought three old friends back into the same frame, all now leading major AI ventures after starting out together at a quantitative trading giant more than a decade ago.
In the image were Wang, who leads Meta's superintelligence lab, Scott Wu, founder of AI coding company Cognition, and Jeff Yan, founder of crypto trading platform Hyperliquid. Wang captioned it simply: "got the HRT gang back together!"
The reunion reflects a much larger shift in Silicon Valley, where people of Chinese background are now concentrated across nearly every layer of the artificial intelligence industry, from chips and data to models and consumer applications.
Where the momentum begins
At the hardware foundation, the world's four largest semiconductor companies are led by executives of Chinese heritage. NVIDIA (NVDA) is run by Jensen Huang, AMD by Lisa Su, Broadcom by Hock Tan, and Intel by Lip-Bu Tan, giving them control over global GPU and custom AI chip supply.
In the middle layer of data supply and frontier labs, Alexandr Wang of Scale AI and Edwin Chen of Surge AI split the market for high-quality training data. Meta's superintelligence lab is led by chief scientist Shengjia Zhao, while OpenAI's defense of its core technology is overseen by Mark Chen. xAI's founding technical team also includes Chinese core members such as Jimmy Ba and Yuhuai Wu.
At the application layer, Scott Wu's Cognition is pushing AI coding, Johnny Ho has helped drive Perplexity's valuation into the tens of billions of dollars, Jesse Zhang's Decagon has captured a multi-billion-dollar customer service market, Jolin Qiao's Fireworks AI has become a key hub for open-source model inference, and 24-year-old math prodigy Carina Hong is taking on machine theorem proving with Axiom Math.
An AI boom inside an old house
To understand where this density of talent comes from, it helps to look at an ordinary house that has become a gathering point for young founders. In the fall of 2025, a reporter stayed in the light blue bungalow in Los Altos where Mark Zuckerberg once lived and worked, now known as the "Facebook House."
The house is simple, with mattresses, bookshelves, and a few cats, but it regularly hosts intensive AI discussions. At one gathering, more than 30 people sat together while a whole lamb roasted in the yard, with conversation ranging from embodied intelligence and social products to large model architecture. One robotics product manager remarked that the atmosphere had become, "If you're not doing AI, it feels like you don't deserve to be alive."
This scene reveals how the so-called Chinese network actually operates. It has no central command and no secret alliance. It is first a talent network woven from a shared language, competition records, overseas study experiences, and technical communities. Newcomers to the Bay Area look for housing, engineers exchange testing details from frontier papers, and founders seek technical partners. Job leads and investment tips spread across overlapping WeChat groups and the social platform X.
As generative AI became Silicon Valley's largest destination for capital, networks once scattered across university labs, big tech firms, and quantitative trading shops quickly converged on commercialization.
Four founders, four different paths
With capital accelerating the shift, this dense talent network soon entered the AI business arena. Scott Wu, Alexandr Wang, Edwin Chen, and Carina Hong, graduates of MIT, Harvard, and Stanford with math and informatics competition medals and, in many cases, experience at quantitative giants like HRT, entered the same wave.
Scott Wu bet that AI could genuinely take on software development work. In March 2024, his company Cognition launched Devin, aiming to evolve AI from a code completion tool into an engineer capable of handling full development tasks. The release caused a sensation but also raised doubts about whether demo capabilities could hold up in real projects.
To bridge that gap, Wu focused on enterprise delivery and positioned the product around a cross-model scheduling architecture. "The world spends trillions of dollars a year on 'meat computers,' meaning human software engineers," he said. "We enjoy being Switzerland. If you do everything, you lose to Microsoft or Google. They do everything, have trillions more resources, and hundreds of thousands more people."
In 2026, Cognition announced more than $1 billion in financing at a post-money valuation of $26 billion, with annualized revenue run rate climbing to $492 million. Its valuation later rose to $48 billion, with Goldman Sachs and Mercedes-Benz among its customers. But the "neutral Switzerland" model soon faced a challenge: six days before the photo was posted, Meta, under Alexandr Wang, released Muse Wonder and restricted the strongest version, Muse Wonder Full, to its own Muse Code, cutting off external access and squeezing independent scheduling platforms as giants entered applications directly.
Alexandr Wang's path moved toward institutional power. Scale AI, which he founded, had been universal data infrastructure for the industry. In June 2025, Meta announced an investment of about $14.3 billion in Scale AI, valuing the company at more than $29 billion, and Wang joined Meta to lead its superintelligence lab.
The deal brought Wang into the center of power at a tech giant but undermined the independence Scale AI relied on. OpenAI and Google quickly adjusted their cooperation to prevent leaks of core model assets. Reuters later reported that Google planned to cut an annual procurement contract worth about $200 million. For a data services provider nurtured by all the major platforms, losing neutrality itself was the biggest commercial loss.
Edwin Chen chose a different bet. A graduate of MIT with degrees in computer science, mathematics, and linguistics, he worked on algorithm engineering at Twitter and Google. When he founded Surge AI, he insisted on an independent model with no outside venture capital. "Venture capital makes companies bloated and distorts team incentives," he said.
According to data disclosed by The Information and Sacra, Surge AI generated $1 billion to $1.2 billion in revenue in 2024 with about 110 full-time employees, surpassing Scale AI, which had a team of about 1,000 and revenue of around $870 million. It remained profitable and naturally absorbed data orders flowing out from major platforms.
In the deeper field of mathematical verification, 24-year-old Carina Hong dropped out of a Stanford doctoral program to found theorem-proving AI company Axiom Math. In March 2026, it completed a $200 million Series A at a post-money valuation of more than $1.6 billion. She recruited renowned mathematician Ken Ono and focused on using formal verification to solve hallucinations in complex reasoning by large models.
She clearly defined the technical route in an interview: "We are not trying to build another chatbot that imitates solutions. We want to teach AI to prove theorems. This is a completely different challenge."
Four people, four different fates: Wu wants to turn technical excitement into reliable products, Wang moved from founder to the center of a giant's power structure, Chen seeks market advantage through independence, and Hong is looking for a commercial outlet at the frontier of mathematical reasoning.
From enterprise coding and neutral data to theorem verification, they quickly entered the frontier with strong technical judgment. Yet under the reshuffling of capital and tech giants, former acquaintances must also face hard market competition.
AI lab battlegrounds and the fight for talent
Behind the rapid advance of technical products, competition among frontier labs for core research talent has also intensified. In mid-2025, Zuckerberg personally intervened and poached four core researchers from OpenAI: Shengjia Zhao, Jiahui Yu, Hongyu Ren, and Shuchao Bi.
All four completed undergraduate education at top Chinese universities, including Tsinghua, the University of Science and Technology of China's youth class, Peking University, and Zhejiang University, before earning doctorates at Stanford, UIUC, and Berkeley. Among them, 29-year-old Shengjia Zhao was deeply involved in the architecture design of ChatGPT, GPT-4, and the o1 logical reasoning model, and was later appointed chief scientist of Meta's superintelligence lab. Zuckerberg said publicly that Zhao had already pioneered several breakthroughs, including new scaling paradigms, proving he is a leader in the field.
Mark Chen, senior vice president of research at OpenAI in charge of overall technical defense, previously worked at Jane Street and once served with Scott Wu as a coach for the U.S. informatics national team. Zuckerberg privately invited him to join but was rejected.
Why this group: from Olympiads and quant trading to AI labs
Stringing these companies and figures together raises a natural question: why are people of Chinese background so prominent in this AI wave? The answer cannot be reduced to being "good at math" or "more hardworking." Looking at them together shows their paths were not identical, but several intersections recur: training in mathematics and computer science, top universities, research institutions, quantitative trading or large tech companies, and Silicon Valley's startup network.
Rigorous math and science competition training in basic education built foundational thinking, top North American universities formed core technical and personal networks, and experience at quantitative firms such as HRT and major tech companies sharpened intuition for ultra-high-concurrency architectures and extremely complex systems. This path turned extreme mathematical talent directly into hard engineering capability, making it one of the most reliable incubators in this AI boom.
For example, Scott Wu's experience is tied to programming and math competitions; Alexandr Wang is known for mathematical ability; Carina Hong also moved from an interest in mathematics and intensive technical education into AI math entrepreneurship. This is not a route unique to Chinese people, but it offers a clue to understanding the current concentration of talent.
Earlier Chinese and Indian immigrants had already accumulated decades of technical and entrepreneurial experience in Silicon Valley. A 1999 study by UC Berkeley scholar AnnaLee Saxenian found that by 1998, Chinese and Indian engineers already held leadership roles in about a quarter of Silicon Valley's high-tech companies. Today's AI startup wave did not start from zero; it builds on long-formed talent flows and professional networks.
China's education system also continues to supply talent to global AI research. Tsinghua's Yao class, the University of Science and Technology of China's youth class, and key science and engineering universities have for decades sent graduates who formed a deep talent pool in North America.
A September 2026 global AI talent tracking report by Carnegie China, "Who's Ahead in the Global AI Talent Race?", showed that among the world's top AI researchers, 57% completed undergraduate education at Chinese universities, up 11 percentage points from 46% in 2022. Over the same period, the share trained at U.S. undergraduate institutions fell to 13%. The study wrote that China has established a clear lead as the largest source country of elite AI talent.
The Carnegie study, based on a sample of 10,280 NeurIPS paper authors, measures educational background, not ethnicity or nationality, and cannot directly represent the entire AI industry. But it suggests an important link between talent trained in China and research and employment opportunities in the United States and elsewhere.
The industry also feels this depth of talent. In an interview with Axios, NVIDIA (NVDA) CEO Jensen Huang said directly that "half of the world's AI researchers are Chinese."
The attractiveness of the U.S. tech industry gives some of these people the chance to turn research capability into startup opportunities. Investors, early employees, and potential customers can connect through shared schools, former employers, and industry gatherings. Spaces like the "Facebook House" are a daily-life reflection of this network.
That does not mean Chinese people hold comprehensive commercial dominance. But in this generative AI wave, other groups are indeed far less densely represented. Most Western elites remain in the capital and vision narratives of the previous wave, while Indian professionals have more often followed the path of big-company CEOs, as with Nadella and Pichai, rather than startup leaders or chief scientists. Eastern European and Russian scientists still make breakthroughs in theoretical exploration, but the group producing founders and core R&D talent at scale is Chinese.
Why now? AI pushes founders into the spotlight earlier
Returning to the cases of Scott Wu and Carina Hong, this technology cycle has profoundly changed how technical talent enters business. A decade ago, building complex software required large teams to construct toolchains and infrastructure, followed by long development cycles before reaching the market. Now, foundational large models deliver general reasoning through APIs, and cloud providers offer ready-made computing infrastructure, allowing startup teams to focus entirely on scheduling optimization, prompt engineering, and specialized delivery.
A small team no longer needs to build its own foundation model to develop a coding agent or theorem verification system. This sharply lowers the threshold for building technical prototypes, enabling small engineering teams with strong math and code optimization skills to launch products capable of supporting enterprise procurement within months.
But that speed is a double-edged sword: the application layer is easy to start, and giants can follow quickly. Once a product is absorbed by an upstream model platform, independent startups must endure enormous pressure to survive. Technical breakthroughs only secure a ticket to entry; avoiding being swallowed by platforms is the long-term challenge.
Two generations, two paths
Looking only at young founders can make it seem that the rise of Chinese people in Silicon Valley began with generative AI. In fact, older-generation leaders such as Jensen Huang and Lisa Su had already built hardware barriers in the chip industry. Huang leads NVIDIA (NVDA), Su leads AMD, Hock Tan heads Broadcom, and veteran investor Lip-Bu Tan became Intel's CEO.
But the paths of the two generations are very different. The older generation built industrial power through decades of manufacturing, supply chains, and ecosystems. The new generation, building on ready-made computing power, moves quickly into applications with extremely high capital efficiency. Most of the huge sums raised by the younger generation ultimately flow to data centers from NVIDIA (NVDA), AMD, and Broadcom, because software speed always depends on hardware foundations.
Immigration has long been an important driving force in Silicon Valley history. In different eras, talent from different regions entered core positions through education, research, engineering, and entrepreneurship. Today's AI wave has once again brought a group of researchers and founders of Chinese background into the spotlight. This time, the faces in the front row have changed.
In this AI gold rush, those surging to the front are Chinese. They are not one "gang," yet they share similar training, networks, and ambitions. The real question is not how they compete, but how this Chinese talent network will define the next phase of Silicon Valley.
Comments