The Insiders Are De-Risking: AI Employees Quietly Cash Out as Bubble Fears Go Mainstream

The people closest to the AI boom are behaving as though it might end. That is the uncomfortable signal emerging from Silicon Valley’s wealth-management industry this week, after Business Insider surveyed advisors who collectively oversee tens of billions of dollars belonging to employees at Anthropic, OpenAI, Cursor, SpaceX and the Magnificent Seven giants. Their near-unanimous read: bubble anxiety has stopped being a fringe worry and has become the default assumption inside the industry’s own ranks.

What the advisors are hearing

Compound Planning, an investment advisory with more than $5 billion in assets under management, says fears of an AI bubble are now voiced in “nearly every conversation” with clients. Many of those clients work at frontier AI labs — Anthropic, OpenAI, Cursor — or at the hyperscalers feeding them compute. What disorients them is not the direction of their wealth but its speed. “The dollar amounts have gotten so big, so fast,” principal wealth advisor Nicholas Garcia told Business Insider.

The macro backdrop explains the vertigo. The Nasdaq 100 has climbed 91% over the last five years. US households sat on a collective $55 trillion in stock wealth in the first quarter of 2026, up from $34 trillion in the fourth quarter of 2022 when ChatGPT debuted. And the private-market numbers are wilder still: Anthropic, valued at roughly $40 billion in 2024, is reportedly being pitched to investors at around $2 trillion ahead of a possible October IPO, per the Financial Times. SpaceX employees are watching shares unlock after an IPO that valued the rocket company at $1.75 trillion, while OpenAI carries a $1 trillion valuation as it eyes its own listing.

Mercer Advisors, which manages $111 billion, says the number one worry among its technology clients is “concentration risk” — the danger that a career, a stock portfolio and a housing market all depend on the same trade. Tidemark Financial Partners, a $400 million advisory working with employees from Nvidia, Meta and Microsoft, says it has been “inundated” by bubble-worried clients for the last 18 months. Jared Redfield, a financial consultant there, estimates about half of clients could stand to reduce tech exposure — while the other half are positioning to buy more if a crash actually arrives.

Five money moves now standard practice

Business Insider’s reporting distills the de-risking playbook into five recurring behaviors:

1. Structured selling. Rather than dumping shares, employees are setting liquidation schedules — selling at the next opportunity once their company’s stock clears a chosen level. The goal is locking in a quality of life: housing, education, retirement, generational wealth. Garcia estimates only about 5% of his clients are true panic sellers; the far bigger group is quietly mechanical about exits.

2. Tax engineering. Because gains accumulated so quickly, mitigating the tax hit dominates every plan. The toolkit includes tax-loss harvesting; direct indexing (buying the individual components of an index to harvest losses at the stock level); long-short strategies on diversified public portfolios; and variable prepaid forwards — contracts that let a shareholder cash in a large position now while deferring delivery, and the capital-gains bill, to a future, possibly lower-tax date.

3. Splurging. A minority — 5% to 10% by Garcia’s estimate — are “splurgers” cashing out large chunks of stock for Bay Area primary homes, vacation properties and luxury cars. Garcia notes many need “a lot of coaching” on lifestyle creep: “You buy a $5 million primary home, a $2 million vacation home, new cars, help family, and then they could take a breath six, 12 months and they’re like, ‘Where’d all my money go?’”

4. AI-adjacent rotation. Diversification increasingly means private vehicles — venture capital and other private assets — plus neocloud stocks, the “shovel sellers” of the AI economy. The subsector has broadly outperformed this year: Nebius Group +138% year-to-date, TeraWulf +24%, CoreWeave +10%, per the reporting. Redfield channels the prevailing logic: “Why go for the gold mine when you can be the one selling the shovels?”

5. Municipal bonds. The humble muni has re-entered the conversation as a tax-management tool, with interest usually exempt from federal income tax. The iShares National Muni Bond ETF is down just 1% year-to-date — boring, by design.

Why this matters beyond the Bay Area

Insider behavior is a classic sentiment indicator, and the direction here is unambiguous: the people with the best visibility into AI economics are spending real money on insurance against disappointment. The contrast with public rhetoric is stark. While executives project perpetual acceleration — and while OpenAI and Anthropic march toward fall IPOs at trillion-scale valuations — their own employees are harvesting losses, deferring gains and hedging concentration with instruments most retail investors have never heard of.

There is a second-order effect worth watching: the IPO window itself. Anthropic’s expected filing will list AI backlash among its risk factors, CNBC has reported, and a critical mass of employees liquidating at the listing could add selling pressure precisely when new public shareholders are most exposed. Lockup dynamics that once worked in tech IPOs’ favor may now amplify volatility in both directions.

The healthiest reading is also the most boring one. Mercer’s Govani describes the selling as “very measured” — insiders taking chips off the table after a “once in a lifetime opportunity,” which is what prudent portfolio theory has always recommended. If the boom continues, they merely got slightly less rich than they might have. If it does not, they will be the few who read the room in time.

But the signal cuts deeper than portfolio construction. An industry whose employees are quietly building exit plans is an industry whose internal confidence no longer matches its external valuation. In every previous cycle — dot-com, housing, crypto — the moment insiders started hedging en masse marked not the end, but the beginning of the end of the belief that prices only go up. Whether AI’s fundamentals (real revenue, real productivity gains, real cash flow) prove sturdier than those precedents is the trillion-dollar question the market will answer this fall.