Why wait for a company to go public when you can mass-produce millionaires?
The Wall Street Journal revealed a set of shocking data. Last October, more than 600 current and former OpenAI employees cashed out a total of $6.6 billion through stock sales, with about 75 people each cashing out $30 million.
This means that even before OpenAI went public, a group of executives and ordinary employees had already reaped the financial rewards of this AI boom.
This is one of the most noteworthy changes in the current AI industry. In the past, employees of startups typically had to wait until after an IPO to actually cash in their stock options. But now, leading AI companies are releasing wealth much earlier.
OpenAI is the most striking example, DeepSeek is catching up on the lessons of external valuation and equity incentives, and companies such as Anthropic, Cerebras, and Character.AI demonstrate that financing, tender offers, secondary market transactions, technology licensing, and team transfers... the ways to create wealth with AI are becoming more and more diverse.
For AI companies, this is a new weapon to attract top talent. For AI talent, technical skills are no longer just about high salaries and stock options; they are also more likely to translate into real benefits before the company goes public.
1
Let's first look at the "wealth creation myth" of OpenAI.
The fact that OpenAI executives earn a lot of money has been made public by the recent legal battle.
Recently, Musk's lawsuit against OpenAI, Ultraman, and others went to court.
During his testimony, President Brockman stated that his stake in OpenAI was worth approximately $30 billion. Meanwhile, former Chief Scientist Sutskevich also disclosed in the Musk v. OpenAI trial that his OpenAI stake was worth approximately $7 billion.
CEO Sam Altman stated that he does not hold any shares in the company, citing its non-profit nature. However, some investors anticipate that Altman may still receive an equity arrangement in the future if OpenAI's for-profit restructuring ultimately proceeds smoothly.
Many ordinary employees have already cashed out substantial sums of money.
According to a report by The Wall Street Journal, in October of last year, OpenAI organized a large-scale stock sale, in which more than 600 current and former employees cashed out their shares on the same day, totaling approximately $6.6 billion.
Of these employees, approximately 75 reached the company's maximum sales limit, each cashing out $30 million.
Some employees donated their remaining shares to donor-advised funds to support charitable causes and receive tax breaks for that year.
This sale is one of the largest employee stock option vesting events in the AI industry to date.
This transaction also marks the first time since the launch of ChatGPT that OpenAI has allowed newly hired employees to cash in their stock options.
This is a noticeable change; OpenAI is becoming increasingly generous when it comes to employee stock options.
Previously, the company stipulated that employees could not sell their shares until they had been with the company for at least two years, so many technical experts who joined the company could not cash in their shares before that time.
Compared to the initial stock issuance seven years ago, the equity value of early employees has increased more than 100 times, far exceeding the wealth growth of traditional technology companies, compared to the Nasdaq index's approximately three-fold increase during the same period.
OpenAI's equity incentive plan itself has also undergone adjustments.
Previously, the sale cap per employee was $10 million, but this will be adjusted to $30 million in the fall of 2025 to respond to the needs of investors and employees.
This system addresses the buying demand from external investors while also providing employees with a way to realize their paper wealth. Historical data shows that if early employees could only sell their shares after the IPO, their wealth appreciation might be affected by market fluctuations. OpenAI's early redemption mechanism effectively mitigates this risk.
Compensation and stock incentives are important means for OpenAI to attract and retain top talent.
OpenAI offers annual salaries of up to $500,000 for some technical positions, plus stock awards and one-time bonuses, some of which can be worth millions of dollars. This combination provides employees with significant financial rewards while also increasing the stability of key positions, supporting the company's rapid progress in technology development and product iteration.
Meta offered its top AI talent a compensation package of up to $300 million last year, highlighting the intense competition and higher compensation levels for high-end talent in the AI industry compared to traditional tech companies.
AI is creating new millionaires in San Francisco, even to the point of boosting the city's long-sluggish housing market.
Some properties, due to numerous competing bids, ultimately sold for far more than the asking price; for example, a property initially priced at $1.6 million sold for $2 million. According to data from Apartment List, San Francisco's citywide rents rose 14% year-over-year in February, the highest increase in the nation.
2
These developments—whether it's executives wielding enormous wealth or ordinary employees receiving high salaries, bonuses, and increasingly "generous" equity schemes—have one obvious benefit for OpenAI: they are bound to create a new attraction for talent.
This appeal goes beyond just "higher salaries." More importantly, it shows employees a path to monetization. Join a leading AI company, receive stock options or shares, and as the company's valuation continues to rise, realize your wealth through a takeover bid, secondary market trading, or a future IPO.
This is one of the reasons why DeepSeek's recent funding rumors are worth paying attention to.
According to Reuters, DeepSeek is pursuing its first external funding round, targeting a valuation of up to $50 billion, with a target funding size of approximately $3 billion to $4 billion. Less than a month ago, rumors suggested DeepSeek was valued at only $10 billion.
On the surface, this is a case of a rising Chinese AI company gaining recognition from investors. But when viewed in the context of the OpenAI case, this event carries another layer of meaning: DeepSeek needs more than just money; it needs a price that is recognized by the external market.
DeepSeek was not a typical venture capital-driven company in the past. Its funding mainly came from its founder, Liang Wenfeng, and his backer, Magic Square Quant.
This is why it was able to maintain a "research team" image for a long time: low-key, technology-oriented, and emphasizing model efficiency. But when a company truly enters the arena of global AI competition, it becomes difficult to sustain an organization solely on technological reputation. Models require computing power, products need commercialization, and teams need long-term incentives.
The primary function of financing is to value the company. Once a valuation is established, the prices of employee stock options and equity become negotiable. Otherwise, equity incentives are more like a long-term commitment: they have theoretical value, but employees don't know their actual worth or when they can be realized.
The reason OpenAI employees were able to cash out on a large scale before the IPO was that the company had already gone through multiple rounds of financing and tender offers, forming a pricing system that investors were willing to accept.
If DeepSeek wants to retain its core members in the long-term competition for AI talent in China, it also needs to fill this gap.
This is especially important for DeepSeek. Reuters reports that the funds raised will be used to strengthen computing infrastructure and improve employee benefits.
The report also mentioned that DeepSeek is facing competition for talent and capital from ByteDance, Alibaba, as well as Chinese AI companies such as MiniMax and Dark Side of the Moon, and there have already been cases of talent loss, such as Luo Fuli joining Xiaomi.
In an industry where companies like OpenAI, Anthropic, and Meta have already raised salary standards, providing sufficiently convincing long-term rewards for core talent is a new challenge DeepSeek faces.
3
The fact that OpenAI employees cashed out demonstrates that leading AI companies can already generate massive wealth before going public; DeepSeek's pursuit of external funding shows that newcomers are also making up for lost value through equity incentives and increased investment in computing power.
This round of AI-driven wealth creation is not limited to "waiting for a company's IPO".
In the past, the most standard way for startups to generate wealth was to go public; now, money is flowing through more complex channels.
Employees can cash out in advance on the secondary market, startups can exit through mergers and acquisitions, and chip companies and infrastructure companies can also enter the public market by taking advantage of the AI boom.
The most direct exit strategy remains an IPO. Besides OpenAI, Anthropic is another example of a model company.
Anthropic is currently believed to be able to go public as early as 2026. Its uniqueness lies in the fact that, unlike DeepSeek, it is not in its initial external funding stage, nor does it have the complex controversies surrounding its transition from non-profit to for-profit like OpenAI. It has Claude, enterprise clients, and support from cloud providers such as Google and Amazon.
Another type of IPO example is the chip startup Cerebras.
Reuters reports that Cerebras plans to raise its IPO price range from $115-$125 per share to $150-$160 per share, and increase the number of shares to be issued from 28 million to 30 million, due to strong investor demand.
Based on the highest bid, the IPO raised approximately $4.8 billion; the deal was oversubscribed by more than 20 times and is scheduled to list on Nasdaq under the ticker symbol CBRS. The report also stated that this could become the world's largest IPO in 2026.
The AI boom has not only made model teams more expensive, but has also turned chips, computing power, and data centers into new outlets for wealth.
Mergers and acquisitions are another option.
In June 2023, Databricks announced the acquisition of generative AI platform MosaicML for approximately $1.3 billion, with the transaction including a retention incentive package. MosaicML at the time focused on helping companies train and deploy their own generative AI models. Databricks' acquisition of it essentially meant directly buying the model training platform, the team, and the company's AI capabilities.
MosaicML had only about 62 employees at the time. Therefore, the media described the deal as being expensive at "approximately $21 million per employee".
Mergers and acquisitions are no longer simply about "a company being completely bought out".
Character.AI is a more typical example of this new sample.
In 2024, Google entered into a technology licensing deal with Character.AI worth approximately $2.7 billion, acquiring a license to its model technology and hiring co-founders Noam Shazeer and Daniel De Freitas, along with some core research members, to join Google DeepMind.
The Financial Times later reported that after the deal, Character.AI abandoned training cutting-edge large models and shifted its focus to enhancing consumer-grade chatbot products.
In addition, the company used the funds to buy out the investors' shares and transfer ownership of the company to the employees, who also received a one-time cash severance. Approximately 30 employees joined Google, and about 100 remained at Character.AI.
In other words, in this case, Google did not fully acquire Character.AI, but it obtained the technology and the most scarce talent through high licensing fees; the original company continued to exist, and investors and employees also obtained liquidity in advance.
The company may not necessarily be bought, but its technology, team, and future revenue rights have already been repriced by major companies.
This is also what makes this wave of AI boom different from many past technology cycles: wealth is no longer released only at the moment of IPO, nor does it belong solely to founders and investors.
The people behind the models, data, computing power, products, and infrastructure are gaining earlier and more complex opportunities to realize their gains through secondary markets, technology licensing, team transfers, mergers and acquisitions, and IPOs.
For AI companies, this is a new weapon to attract talent; for AI talent, it means they don't necessarily have to wait for an IPO and may be able to turn their technical skills into real benefits sooner.
This article is from the WeChat Official Account "Letter List" (ID: wujicaijing), author: Xiaojinya
