Just as Anthropic approaches a $1 trillion IPO, the world’s most sought-after AI company has lost several key members in succession.
On November 5, Cat Wu, former head of product for Claude Code, announced she has joined OpenAI as Head of Core Product Management. She will be involved in developing OpenAI’s consumer and enterprise products, directly joining the rival camp of her former employer.
This is the third key team member to leave the Claude Code team, publicly confirmed within the past two weeks.
According to traditional Silicon Valley wealth logic, just before an IPO should be the time when employees are least likely to leave the company. After all, a successful listing could yield millions—or even tens of millions—of dollars from early stock options and restricted stock units.
However, today's AI industry seems to be rewriting these rules.
IPO countdown underway as key figures depart one after another
Cat Wu's departure has attracted significant attention, primarily because Claude Code has become one of Anthropic's most important commercial products.
After joining Anthropic in 2024, she collaborated with Boris Cherny, the creator of Claude Code, to bring this initially internal programming tool to market, where it gradually evolved into a standout product in the AI programming space.
According to Anthropic's official disclosure in February this year, the annualized revenue run rate of Claude Code has surpassed $2.5 billion, doubling since the beginning of the year, with enterprise customers contributing more than half of the revenue.
But just as the product was experiencing rapid growth, several key builders chose to leave one after another.
On October 23, Anthony Morris, creator of the Claude Code desktop version, left the company. He stated on social platform X that leaving a product he had invested immense effort into and close colleagues was one of the most difficult decisions of his career, and that he would share his next steps in the future.
On October 24, Claude Code engineer Thariq Shihipar announced his departure from Anthropic. He played a key role in the engineering development of Claude Code and witnessed its evolution from an early-stage product into a widely used AI tool. In his farewell statement, he praised the company and team highly, while expressing his intention to take some time off before taking on new challenges.
On October 30, Cat Wu officially concluded her work at Anthropic. On November 5, she announced her joining of OpenAI. According to reports, Wu will report to Tibo Sottiaux, Head of Core Products at OpenAI, and collaborate with the enterprise and consumer product teams on the development of next-generation AI agent products.
In fact, this is not the first time Anthropic has experienced the departure of key talent.
In February this year, Mrinank Sharma, Head of Security Research, departed and publicly expressed concerns about AI risks and whether the organization could truly live up to its stated values.
In July, Jenny Wen, who was responsible for the product design of Claude and Cowork, joined the AI programming company Cursor as Head of Design.
In September, researcher Jacob Coxon left Anthropic after participating in large model pre-training. In an interview with Axios, he expressed concern that AI companies are pushing technological development too rapidly under competitive pressure, even forfeiting equity that was about to vest upon his departure.
However, the departure of core employees does not mean Anthropic is unable to retain talent. According to the company’s prospectus filed with the SEC on November 2, the number of employees had grown from approximately 940 at the end of 2024 to 5,300 as of the end of September this year. Over the past 12 months, the “regretted attrition rate”—the proportion of employees the company sought to retain but lost—was only about 1%.
K-shaped divergence in the AI talent market
For many, Cat Wu’s move still seems unbelievable. Since Anthropic is about to go public, why not wait until after the IPO?
The answer lies first in the fact that the way top talent in Silicon Valley accumulates wealth differs from that of ordinary professionals.
AI companies typically offer compensation packages that include base salary, bonuses, and, most importantly, equity incentives. For early employees, the vested shares may far exceed the value of years of salary as the company’s valuation grows manyfold.
When Anthropic completed its $65 billion funding round in May this year, its post-money valuation reached $965 billion. Current market discussions around its IPO valuation are even higher, ranging from $1.5 trillion to $2 trillion.
According to the latest SEC filing, as of the end of September, Anthropic had approximately $32.3 billion in unrecognized equity-based compensation expenses related to restricted stock units (RSUs), making equity incentives a key component of Anthropic’s talent framework.
For core employees who have already accumulated significant equity, staying with the company and waiting for an IPO is not necessarily the only path to maximizing wealth.
If you can retain your vested equity while receiving a new, high-value compensation package from a competitor, you may potentially benefit from the growth of both companies in the future.
Of course, leaving may also result in the loss of unvested shares, depending on individual agreements. Cat Wu herself has explicitly stated that whether leaving allows her to sell her Anthropic shares earlier was not a factor in her decision to change jobs.
More importantly, Silicon Valley has an institutional environment that encourages the free movement of talent.
According to the official guidance issued by the California Department of Justice in 2024, California generally prohibits non-compete agreements in ordinary employment relationships. This means that even if an employee quickly joins OpenAI after leaving Anthropic, they typically would not be restricted by non-compete clauses solely for joining a competitor.
In fact, Cat Wu's professional journey is a microcosm of Silicon Valley's talent mobility culture.
In July 2025, she joined Cursor alongside Boris Cherny, the creator of Claude Code, but both returned to Anthropic less than two weeks later. Now, she has left again and joined OpenAI.
In Silicon Valley, this phenomenon is known as the Tech Carousel, where companies are both business competitors and sources of talent for one another.
This phenomenon is being pushed to its extreme in the AI industry.
In February this year, Reuters reported that Ruoming Pang, formerly head of Apple’s AI model team, was successfully poached by OpenAI after just seven months at Meta. According to Bloomberg, Meta’s multi-year compensation package to recruit him was worth over $200 million.
In comparison, the base annual salary range for the core model product manager position currently listed on the OpenAI website is $347,000 to $490,000.
Thus, AI is creating a "superstar" talent market: the value of a small number of individuals who can drive model breakthroughs, lead key products, and determine the direction of technological commercialization is being amplified by capital at an unprecedented rate. However, young engineers on the other end of the talent market face a very different reality.
On October 16, Anthropic released a research report 《How Employment Is Changing at Anthropic》, analyzing how AI technology is transforming internal workflows and hiring structures. The data shows that since early 2025, the proportion of junior employees in Anthropic’s technology department has decreased from 21% to 14%; among newly hired technical staff, the share of junior employees has dropped from 27% to 9%. Meanwhile, by August of this year, the cumulative runtime of interactive AI agents used by employees had reached six times the total hours worked by human employees.
This change reveals the emerging "K-shaped divergence" in the talent market during the AI era.
At the top of the K-curve are the elite few who can design systems, manage agents, drive model advancements, and commercialize products. AI enhances their productivity and increases the price companies are willing to pay for their skills.
At the lower end are entry-level professionals who rely on basic coding, repetitive tasks, and standardized knowledge work.
In the past, a company expanding its operations needed to hire more engineers; in the future, the same business growth may require only a few senior employees to oversee numerous AI agents.
Trillion-dollar IPO meets sharp decline as capital markets reprice
A similar divergence is also evident in the capital markets.
After the midterm elections, the U.S. Democratic Party regained control of both the Senate and the House of Representatives and achieved a "red-to-blue" flip in Texas. The newly elected Texas governor, Gina Hinojosa, made a central campaign promise to immediately freeze approvals for new data centers until the state legislature passes relevant safeguard legislation. Texas is no ordinary state: according to Wells Fargo, 27% of all planned data center capacity in the U.S. is located in Texas. Goldman Sachs further estimates that Democratic victories in Texas, Michigan, Georgia, and other states could expose approximately 45% of planned U.S. data center capacity to risks of suspension, stricter scrutiny, or cost reallocation.
The market responded to this political signal with selling pressure. On November 5, the Nasdaq Index fell as much as 5% intraday, while the Philadelphia Semiconductor Index dropped by over 11%. The storage and data center hardware sectors were hit hardest: SanDisk plunged more than 19%, Western Digital slid over 16%, Micron Technology fell more than 15%, and SK Hynix dropped over 14%. Until now, AI-related stocks have sustained high valuations based on the assumption that tech companies could continuously build new data centers and expand computing power. Today, this assumption encountered real-world political resistance for the first time.
This is also the test that Anthropic will soon face.
On October 26, Anthropic officially filed its IPO application; on November 2, it updated its prospectus, planning to list on Nasdaq under the ticker symbol ANT. The prospectus revealed third-quarter revenue of $16.96 billion, a 47% sequential increase, but the company turned operating loss, recording an operating loss of approximately $567 million. More notably, it disclosed future commitments for compute power procurement and leasing totaling $895.4 billion. Market expectations had previously discussed a potential IPO valuation of around $2 trillion, but Reuters reported that some investors have now lowered their expectations to approximately $1.5 trillion.
Thus, an interesting scenario arises:
For a select few top AI talents, the next step is entirely up to them. For these AI companies, even with explosive revenue growth, they must prove themselves in two areas: meeting the standards set by policymakers and demonstrating to capital markets that trillion-dollar investments in computing power will ultimately translate into profits.
As Anthropic’s IPO approaches, Wall Street may soon have its own answer.
Author: seed.eth
Twitter: https://twitter.com/BitpushNewsCN
BitPush Telegram community: https://t.me/BitPushCommunity
BitPush TG subscription: https://t.me/bitpush