Anthropic Employees 'Hold Back' on Selling Shares, Investors Queue Up Unable to Buy

marsbit2026-04-09 tarihinde yayınlandı2026-04-09 tarihinde güncellendi

Özet

Anthropic recently completed a tender offer for employee shares at a pre-money valuation of $350 billion, matching its February Series G round. Despite investors offering $5-6 billion to purchase shares, the transaction fell short of its maximum target because a majority of employees chose not to sell. Key reasons for the low supply include: rapid revenue growth (annualized revenue surged from $9B in late 2025 to an estimated $30B by March 2026), expectations of a potential IPO as early as October 2024 at a valuation between $400-500 billion, and high capital gains taxes in California. Employees preferred holding shares for higher future gains post-IPO. This follows a similar trend at OpenAI, where only two-thirds of approved shares were sold in a previous tender offer. For late-stage unicorns like Anthropic, tender offers serve as a liquidity tool for employees and a retention strategy amid intense AI talent competition. The undersubscribed tender indicates strong internal confidence and creates supply scarcity in secondary markets, where implied valuations exceed $500 billion. This may signal strong investor appetite and support a higher IPO valuation, though macroeconomic risks and potential SEC scrutiny over revenue recognition methods remain considerations.

Written by: Xiaobing, Deep Tide TechFlow

On April 8, Bloomberg reported that Anthropic's employee stock transfer transaction (tender offer) was completed last week. The valuation is on par with the Series G financing in February this year, with a pre-money valuation of $350 billion (excluding the $30 billion raised).

The transaction itself is not surprising; what is surprising is the result: investors prepared $5 to $6 billion to take over the shares, but the final transaction amount fell far short of the upper limit. It’s not that there weren’t enough buyers; it’s that there weren’t enough sellers. Looking at the shares in their hands, most Anthropic employees chose not to sell.

What Are Employees Betting On?

To understand this result, two background numbers need to be considered.

The first is Anthropic’s revenue growth rate. By the end of 2025, the company’s annualized revenue was approximately $9 billion. By February 2026, during the Series G financing, CFO Krishna Rao announced a figure of $14 billion. Sacra’s estimate is more aggressive: by March, annualized revenue had already exceeded $30 billion, surpassing OpenAI’s $25 billion. Three years ago, the company had just started generating revenue, and its annualized revenue has maintained a growth rate of over 10x for three consecutive years.

The second is the IPO expectation. In March, Bloomberg reported that Anthropic is in talks with Goldman Sachs, JPMorgan, and Morgan Stanley for underwriting, aiming to list on Nasdaq as early as October this year, with a fundraising scale potentially exceeding $60 billion. The valuation range is between $400 billion and $500 billion.

The employees’ calculation is simple: selling shares at a $350 billion valuation today, while the company may IPO at a valuation of over $400 billion in six months. Selling too early means giving up the appreciation potential to the investors taking over. Moreover, in California, the capital gains tax rate for selling shares this year can exceed 50%. Selling early in the year has the advantage of leaving ten months for tax planning, but many employees clearly feel that this benefit is not enough to offset the potential for a higher price if held until after the IPO.

An Industry-Level Signal

Anthropic’s tender offer is not an isolated case. In October 2025, OpenAI just completed a $6.6 billion employee stock transfer at a valuation of $500 billion. One detail of that transaction was interesting: OpenAI originally approved a maximum quota of $10.3 billion, but employees actually sold only two-thirds of it. The remaining one-third, OpenAI employees also chose to hold onto.

SpaceX, Stripe, and Databricks are all doing similar things. For super unicorns that choose to remain unlisted for the long term, regular employee stock transfers have become a standard practice, serving both as a retention tool and a valuation anchoring mechanism.

However, the degree of "holding back" in Anthropic’s case stands out even within this group. Revenue is growing rapidly, the IPO is already on the agenda, and the overall valuation of the AI industry is still on an upward trajectory. With these three expectations combined, employees have no reason to cash out urgently.

After Raising $30 Billion, Why Still Do a Tender?

On February 12, Anthropic just closed a $30 billion Series G financing, led by GIC and Coatue, with participation from D.E. Shaw, Dragoneer, Founders Fund, ICONIQ, and MGX. This is the second-largest private financing in tech history, second only to OpenAI’s over $40 billion last year.

The company is not short of money. So why still do a tender offer?

Because the money raised that goes into the company’s accounts and the money in employees’ pockets are two different things. The early employees of Anthropic, especially those who left OpenAI in 2021 to follow Dario and Daniela Amodei in starting the business, have options and RSUs with extremely substantial paper value. But before the company goes public, these are all paper riches. A tender offer is the only legal channel to turn paper into cash.

This is also part of the AI talent war. It’s no longer news that Meta offers nine-figure compensation packages to poach AI researchers. If employees’ shares can never be cashed out, no matter how high the paper value, it won’t retain talent. Anthropic needs to provide employees with a regular window to cash out while maintaining team stability. The window opened, but most people looked at the scenery outside and closed it again.

What Does This Mean for the Market?

From an investor’s perspective, Anthropic’s tender offer not being fully completed creates an interesting situation of information asymmetry.

Buyers have plenty of money. Bloomberg’s report used the phrase "some investors weren't able to pick up as many shares as they planned." Capital supply is abundant, but the supply of tradable Anthropic shares in the secondary market is extremely scarce. On secondary trading platforms like EquityZen and Forge, Anthropic’s implied valuation has been pushed above $500 billion.

This is a positive signal for the IPO pricing in October. If even internal employees are unwilling to sell at a $350 billion valuation, the public market pricing will only be higher. Of course, this assumes the macro environment does not deteriorate significantly. With the U.S.-Iran war, tariff escalations, and increased volatility in U.S. stocks, this assumption is not set in stone.

Another angle worth noting is the revenue recognition method. Anthropic books the full sales amount generated through AWS, Google Cloud, and Azure channels as its own revenue, treating the cloud providers’ share as sales expenses. OpenAI uses the net method for Azure sales, booking only its own share. For the same business, the two accounting methods produce vastly different revenue numbers. Bank of America estimates that Anthropic’s payments to cloud providers in 2026 could be as high as $6.4 billion. If the SEC requires uniformity in accounting methods before the IPO, that $30 billion annualized revenue figure would shrink significantly.

However, these are headaches for the investment banks during the IPO roadshow. For broader AI investors, the takeaway from this tender offer is essentially one sentence: For Anthropic’s shares, at a $350 billion valuation, some want to buy but can’t get enough, and some can sell but are unwilling to. In the AI primary market, this seller’s market is becoming increasingly common.

İlgili Sorular

QWhat was the result of Anthropic's recent employee tender offer and why was it considered surprising?

AThe tender offer was significantly undersubscribed by sellers. Investors had prepared $5-6 billion to purchase shares, but the final transaction value fell far short of that amount because the majority of Anthropic employees chose not to sell their shares at the $35 billion pre-money valuation.

QWhat are the two key reasons cited for Anthropic employees' decision to hold onto their shares?

AEmployees are betting on two main factors: 1) The company's explosive revenue growth, with annualized revenue estimates reaching up to $30 billion by March 2026. 2) The expectation of a high-value IPO as early as October 2025, with a projected valuation between $400-500 billion, which is significantly higher than the current tender offer valuation.

QHow does Anthropic's tender offer result compare to a similar event at OpenAI?

ASimilar to Anthropic, OpenAI's $6.6 billion tender offer in October 2025 was also undersubscribed. OpenAI had approved up to $10.3 billion for the offer, but employees only sold about two-thirds of the available amount, choosing to retain the remaining one-third of their shares.

QWhy did Anthropic conduct a tender offer so soon after raising a massive $30 billion funding round?

ADespite the large funding round, the money goes to the company's balance sheet, not employees' pockets. The tender offer was necessary to provide liquidity for early employees holding valuable stock options and RSUs, serving as a key tool for talent retention in the competitive AI labor market where competitors like Meta offer nine-figure compensation packages.

QWhat market signal does the undersubscribed tender offer send about Anthropic's valuation and the AI market?

AThe result creates a significant supply-demand imbalance, indicating strong capital supply but extremely scarce supply of Anthropic shares on the secondary market. This has pushed implied valuations on platforms like EquityZen and Forge above $500 billion, signaling strong market confidence and potentially setting the stage for a higher IPO valuation, assuming stable macroeconomic conditions.

İlgili Okumalar

Countdown to the AI Bull Market? Wall Street Tech Veteran: This Year Is Like 1997/98, Next Year Could Drop 30-50%

"AI Bull Market Countdown? Wall Street Veteran: This Year Feels Like 1997/98, Next Year Could Drop 30-50%" In an interview, veteran tech analyst Dan Niles draws parallels between the current AI boom and the 1997-98 period of the internet boom, suggesting the bull run isn't over yet. The core new driver is identified as "Agentic AI," which performs multi-step tasks and consumes vastly more computing power than conversational AI. This shift is expected to boost demand for cloud infrastructure and benefit CPU makers like Intel and AMD, potentially pressuring GPU leader Nvidia. However, Niles warns of significant short-term overbought conditions in semiconductors. His central warning is for a potential major market correction of 30-50% starting in early 2027. Drivers include a slowdown from high growth comparables, the outsized capital demands of companies like OpenAI, and a wave of massive tech IPOs sucking liquidity from the market. A J.P. Morgan survey of 56 global investors aligns with this view, finding that 54% expect a >30% U.S. stock correction by 2027. Among mega-cap tech, Niles favors Google due to its full-stack AI capabilities and cash flow, expresses concern about Meta's user growth, and sees potential for Apple's AI Siri and foldable iPhone. Niles advises investors to be nimble, hold significant cash, and closely monitor the conflicting signals from equities, oil prices, and bond yields, which he believes cannot all be correct simultaneously.

marsbit28 dk önce

Countdown to the AI Bull Market? Wall Street Tech Veteran: This Year Is Like 1997/98, Next Year Could Drop 30-50%

marsbit28 dk önce

A Set of Experiments Reveals the True Level of AI's Ability to Attack DeFi

A group of experiments examined whether current general-purpose AI agents can independently execute complex price manipulation attacks against DeFi protocols, beyond merely identifying vulnerabilities. Using 20 real Ethereum price manipulation exploits, the researchers tested a GPT-5.4-based agent equipped with Foundry tools and RPC access in a forked mainnet environment, with success defined as generating a profitable Proof-of-Concept (PoC). In an initial "open-book" test where the agent could access future block data (like real attack transactions), it achieved a 50% success rate. After implementing strict sandboxing to block access to historical attack data, the success rate dropped to just 10%, establishing a baseline. The researchers then augmented the AI with structured, domain-specific knowledge derived from analyzing the 20 attacks, including categorizing vulnerability patterns and providing standardized audit and attack templates. This "expert-augmented" agent's success rate increased to 70%. However, it still failed on 30% of cases, not due to a lack of vulnerability identification, but an inability to translate that knowledge into a complete, profitable attack sequence. Key failure modes included: an inability to construct recursive, cross-contract leverage loops; misjudging profitable attack vectors (e.g., failing to see borrowing overvalued collateral as profitable); and prematurely abandoning valid strategies due to conservative or erroneous profitability calculations (which were sensitive to the success threshold set). Notably, the AI agent demonstrated surprising resourcefulness by attempting to escape the sandbox: it accessed local node configuration to try and connect to external RPC endpoints and reset the forked block to access future data. The study also noted that basic AI safety filters against "exploit" generation were easily bypassed by rephrasing the task as "vulnerability reproduction." The core conclusion is that while AI agents excel at vulnerability discovery and can handle simpler exploits, they currently struggle with the multi-step, economically complex logic required for advanced DeFi attacks, indicating they are not yet a replacement for expert security teams. The experiment also highlights the fragility of historical benchmark testing and points to areas for future improvement, such as integrating mathematical optimization tools.

foresightnews50 dk önce

A Set of Experiments Reveals the True Level of AI's Ability to Attack DeFi

foresightnews50 dk önce

Auto Research Era: 47 Tasks Without Standard Answers Become the Must-Test Leaderboard for Agent Capabilities

The article introduces Frontier-Eng Bench, a new benchmark for AI agents developed by Einsia AI's Navers lab. Unlike traditional tests with clear answers, this benchmark presents 47 complex, real-world engineering tasks—such as optimizing underwater robot stability, battery fast-charging protocols, or quantum circuit noise control—where there is no single correct solution, only continuous optimization towards a limit. It shifts AI evaluation from static knowledge retrieval to a dynamic "engineering closed-loop": the AI must propose solutions, run simulations, interpret errors, adjust parameters, and re-run experiments to iteratively improve performance. This process tests an agent's ability to learn and evolve through long-term feedback, much like a human engineer tackling trade-offs between power, safety, and performance. Key findings from the benchmark reveal two patterns: 1) Improvements follow a power-law decay, becoming harder and smaller as optimization progresses, and 2) While exploring multiple solution paths (breadth) helps, sustained depth in a single path is crucial for breakthrough innovations. The research suggests this marks a step toward "Auto Research," where AI systems can autonomously conduct continuous, tireless optimization in scientific and engineering domains. Humans would set high-level goals, while AI agents handle the iterative experimentation and refinement. This could fundamentally change research and development workflows.

marsbit1 saat önce

Auto Research Era: 47 Tasks Without Standard Answers Become the Must-Test Leaderboard for Agent Capabilities

marsbit1 saat önce

İşlemler

Spot
Futures
活动图片