The company was sold for 700 million yuan, and the employees achieved financial freedom in one year.

Another AI wealth-creation myth.

Having been established for only one year, with its product still in internal testing and the number of users undisclosed - the AI startup Torch was acquired by OpenAI at a price of over $100 million (approximately 700 million RMB).

Surprisingly, the Torch team consists of only four people, including one with an Asian face. They will all join OpenAI. Roughly calculated, the per capita return is about $25 million, and the early investors have also exited successfully.

Their story is a microcosm of the current AI talent war: in the wave of global tech giants frantically "scrambling for talent", the truly valuable ones are those few who can define the future.

Just now, OpenAI bought four people in one go

Let's first look at the details of this deal.

OpenAI announced that it has completed the acquisition of the medical startup Torch and will integrate it into the ChatGPT Health department.

The core of the deal is the people. Along with the deal, the entire core team of Torch - only four people: CEO Ilya Abyzov, co-founder Eugene Huang, Ryan Oman, and James Hamlin - will all join OpenAI. Among them, Eugene Huang is a Chinese-American born in the 1990s and taught at the University of California, Berkeley in his early years.

According to The Information, the total consideration for this deal exceeds $100 million. About $60 million will be cashed out immediately, and the rest will be distributed to the team members as retention incentives. This is a typical talent acquisition. Roughly calculated, each founding member will receive an average of about $25 million.

Notably, Torch was only established just one year ago. The company's product is an application program that aims to centralize users' medical and health information scattered in hospitals, laboratories, wearable devices, and consumer-level testing institutions in one place to form a personal medical file that can be called by AI.

Simply put, Torch will classify and organize the medical records, examination reports, and consultation content uploaded by users, convert obscure medical terms into plain language, extract key information, identify trends and potential associations in multi-source data, and finally output more understandable and actionable health insights. However, Torch itself relies on the underlying support of large models including OpenAI, and its product is still in a small-scale internal testing phase, and the specific user scale has not been disclosed.

There were early signs of this acquisition. Just one week before the news was announced, OpenAI launched ChatGPT Health, attempting to connect the large model with users' health data to provide more context-aware medical advice. Now, Torch's technology just fills in the key link - using AI to build "individual health memory". OpenAI also said that the combination of Torch and ChatGPT Health will open up a new way for users to understand and manage their own health.

Torch CEO Ilya Abyzov said, "I can't imagine a better next chapter. Now, we can put this technology directly into the hands of hundreds of millions of users who consult ChatGPT about health issues every week."

The team further emphasized in a statement: "We are joining OpenAI to realize this vision on a scale far beyond our own capabilities - there is nothing more important than bringing our technology and ideas to global users at this moment."

Serial entrepreneurship, the previous project went bankrupt

Torch's story begins with a failed startup.

Ilya Abyzov graduated from Dartmouth College and worked at institutions such as Morgan Stanley and HBK in his early years. In 2012, he joined Uber as a regional general manager. In 2016, he founded Forward with former Google executive Adrian Aoun and others, aiming to reshape healthcare services.

The company's model was quite ambitious: users could pay a monthly subscription fee of $149 to use a digital clinic system that integrated AI, intelligent sensors, full-body scanners, and diagnostic algorithms. At its peak, Forward operated physical clinics in 19 locations across the United States, with more than a hundred healthcare doctors on staff, and even boasted of being "the Apple of the healthcare industry".

With elite founders and the highly imaginative track of AI healthcare, Forward had a smooth journey in raising funds. At the beginning of its establishment, the company received $30 million raised by many investors, including Google Chairman Eric Schmidt, Benioff, Oscar founder Joshua Kushner, and Uber co-founder Garrett Camp. By 2019, the company's valuation had reached $453 million.

The Series D financing in 2021 pushed Forward to the peak - well-known investment institutions such as Founders Fund, Khosla Ventures, and SoftBank Vision Fund, as well as individual investors such as Salesforce founder, chairman, and CEO Marc Benioff and singer The Weeknd, formed a shareholder team that brought $225 million in financing to Forward, and the company officially joined the unicorn club, with a valuation exceeding $1 billion.

The last round of financing was in 2023. Forward announced that it had successfully raised $100 million in Series E financing provided by shareholders such as SoftBank, Khosla Ventures, and Founders Fund. At this time, the company had raised a total of more than $657 million in financing, and everyone was full of confidence in the future.

However, in 2024, Forward suddenly announced that it would stop operating.

Looking back, the collapse was not without signs. As early as the Series E financing, Forward had admitted that it was facing an "extremely difficult market environment" and had undergone multiple rounds of layoffs and closed some clinics. The new product CarePod became the last straw that broke the camel: the cost of each unit exceeded $1 million, but it frequently malfunctioned, such as failed blood draws and users being trapped inside the pod. The more fundamental reason was that since its establishment, the company's revenue had been less than $100 million, and its self-sustaining ability was seriously insufficient.

When technology cannot replace trust, even the most dazzling hardware cannot conceal the lack of the essence of service. Forward came to an end, but the flame was not extinguished. In early 2025, Ilya Abyzov founded Torch with three core members of Forward - Eugene Huang, Ryan Oman, and James Hamlin.

This time, they chose to travel light: no hardware, no clinics, and no longer trying to reshape the medical system. Instead, they focused on a more practical thing - using AI to make scattered medical data truly useful to users.

In the initial stage, Torch only completed a small seed round of financing, supported by old friends such as former Forward CEO Adrian Aoun and First Round Capital.

By no longer emphasizing grand narratives, they got closer to reality. After only one year of establishment, the team was acquired by OpenAI for over $100 million. The founders entered the core battlefield of large models, and the early investors exited successfully. For a startup that restarted from failure, this is almost a perfect ending.

The AI talent scramble

The talent war is in full swing.

A shocking scene occurred last month when Meta announced that it would acquire Manus for billions of dollars. Calculated by the amount, this deal ranks third in Meta's historical acquisitions, second only to WhatsApp and Instagram. After the deal is completed, the parent company of Manus, Butterfly Effect, will continue to operate independently, and founder Xiao Hong will serve as Meta's vice president.

This news quickly ignited the domestic venture capital circle: remember that in March 2025, Manus officially launched its product, and it became extremely popular overnight during the internal testing. Nine months later, this Chinese startup team won the favor of a tech giant. Before the acquisition, Manus had completed four rounds of financing, and the investors included ZhenFund, Sequoia China, Tencent, and the veteran Silicon Valley venture capital firm Benchmark.

However, the deal has not been finalized yet. The Ministry of Commerce recently said that it will conduct an assessment and investigation on the export control and technology import and export compliance involved in the deal.

Manus is not an isolated case. Earlier, Meta acquired Scale AI for about $14 billion and brought 28-year-old founder Alexandr Wang under its wing to lead the superintelligence project.

Jensen Huang is also frantically scrambling for talent. Just last month, NVIDIA acquired the assets of chip designer Groq for $20 billion in cash. According to the agreement, Groq founder and CEO Jonathan Ross, Sunny Madra, and several core executives will join NVIDIA.

Not only did NVIDIA acquire the technology and talent, but the incentive package it offered was also very generous. The potential equity value per person for about 600 Groq employees reached $5 million.

The logic behind this is easy to understand: Top AI talent is extremely scarce, and acquisitions have become the most efficient way to "scramble for talent". "I've been an executive in a tech company for 20 years, and I've never seen this kind of price before," Meta CTO said bluntly. The market is setting an "incredible" price for high-level AI talent.

Looking across the ocean, the AI talent scramble is also intensifying. In December 2025, Tencent appointed 27-year-old former OpenAI researcher Yao Shunyu as its chief AI scientist, who will be fully responsible for the large model infrastructure. Almost at the same time, Luo Fuli, a post-1995 large model leader who joined Xiaomi, made her debut. The younger generation is quickly stepping onto the core stage.

During the 2026 campus recruitment season, large companies such as Baidu, ByteDance, Alibaba, Tencent, and JD.com have all increased their recruitment for AI positions. On social platforms, it has become normal for top university doctoral students to receive five or six job offers, and there are also constant rumors that the total package easily exceeds one million yuan - the feeling in the talent market is more direct.

Ultimately, the competition in AI is not just a race between models and computing power, but also a competition in talent density. A more decisive change is that more and more young Chinese scientists and entrepreneurs are moving to the center - they will determine how the next round of the AI landscape will be rewritten.

This article is from the WeChat official account “Investment World” (ID: pedaily2012), author: Wang Lu. It is reprinted by 36Kr with authorization.