Joining an AI lab from a FAANG company often means leaving a significant amount of unvested equity behind. The labs know this and have institutional muscle for negotiating equity buyouts and sign-on bonuses to bridge the gap. But candidates who do not understand the negotiation playbook leave money on the table — sometimes hundreds of thousands of dollars. This piece is a practical guide to negotiating sign-on and buyout at the major AI labs in 2026.
Why this negotiation matters
For a senior+ engineer leaving FAANG with 2-3 years of unvested RSU stack, the unvested value can easily be $500K-2M+. If the AI lab does not bridge that, the candidate takes a meaningful pay cut in years 1-2 even if total compensation goes up long-term. The buyout is the mechanism that makes the move financially viable.
AI labs treat this as a routine cost of senior hiring. The expected default is that they will buy out some portion of unvested equity. The negotiation is about how much, in what form, and on what schedule.
Documenting your unvested equity
The first step is concrete documentation. Bring to the negotiation:
- A spreadsheet listing each unvested grant: original grant date, total grant size, vesting schedule, current unvested portion, current market value.
- For RSUs: the value at the company’s current public stock price.
- For pre-IPO equity (if leaving a private company): valuation from the most recent funding round, with caveats about liquidity.
- Refresh grants you would expect to receive in the next 12-18 months at your current employer (if stable refresh policy).
Total it. The number you walk into the negotiation with is the high-end target.
What labs will typically agree to
- Sign-on bonus that covers roughly the next 12 months of expected vesting at your current employer.
- Sometimes a multi-year sign-on (year 1 + smaller year 2) to bridge a longer vesting schedule.
- Refresh grants in subsequent years that ramp to a steady-state level.
What labs typically will not do:
- Match the full unvested stack at face value.
- Pay sign-on as additional equity at full pre-IPO valuations (sometimes done partially in equity, but heavily discounted).
- Match speculative future refreshes that you cannot prove are committed.
Typical sign-on ranges in 2026
Approximate sign-on bonus ranges for senior+ engineers leaving FAANG:
| Unvested equity at prior | Typical sign-on at AI lab |
|---|---|
| Under $200K | $50-150K |
| $200K-500K | $150-400K |
| $500K-1M | $300-700K |
| $1M-2M | $500K-1M |
| $2M+ | $700K-1.5M+ |
Above $2M, sign-on negotiations get more bespoke and depend heavily on the candidate’s overall leverage and the lab’s hiring priority for the role. Multi-year sign-ons (e.g., $750K year 1 + $400K year 2) are common in this range.
Sign-on structure variations
Cash sign-on
Most common form. Paid as a lump sum on the start date or first paycheck. Subject to clawback if the candidate leaves within 12-24 months (terms vary).
Multi-year sign-on
Year 1 cash bonus + year 2 cash bonus. Often used to bridge a longer remaining vesting schedule at the prior employer. Each year typically has its own clawback window.
Equity sign-on
Some labs offer additional equity grants as part of the sign-on package, typically vesting on the same 4-year schedule as the standard grant. Less common because the value depends on liquidity, which is uncertain at private labs.
Hybrid
Cash + additional equity. Most flexible. Often the best structure when the candidate has both significant unvested cash equivalents and meaningful long-term upside expectations at the new lab.
The negotiation conversation
The opening move:
“I’m leaving roughly $XYZ in unvested equity over the next 24 months. Can we discuss how the package addresses that?”
This is direct, specific, and treats the buyout as a normal part of the negotiation rather than an unusual ask. Most recruiters will respond with a sign-on number; if the first number is below your expectation, push back with documented unvested value and ask for a second look.
Common counter-tactics from recruiters:
- “Our standard sign-on is $X.” — sometimes negotiable beyond standard for senior+ candidates with significant unvested stack.
- “The equity grant is intended to make up for it.” — push back: the equity grant has uncertain liquidity, while your prior RSUs would have been publicly tradeable.
- “We don’t typically buy out future refreshes.” — fair pushback if you cannot document the refresh; if you can document a stable refresh history, it’s negotiable.
Refresh grants in years 2+
Sign-on bonuses bridge year 1. Years 2+ depend on refresh grants. AI labs have varying refresh policies:
- OpenAI: generous refreshes, often matching or close-to-matching the original grant size.
- Anthropic: meaningful refreshes; specifics vary by team and role.
- DeepMind: Alphabet refresh policy applies; Alphabet’s standard refresh has been moderate historically.
- Smaller labs: variable; ask explicitly during negotiation.
Strong negotiators get refresh expectations in writing during the offer process. Verbal commitments are harder to enforce.
What to avoid
- Treating the equity grant as cash. Pre-IPO equity is illiquid and uncertain. When comparing offers, discount the equity component meaningfully.
- Accepting the first sign-on offer. The first number is rarely the best. Push back with documented unvested value.
- Threatening to walk without leverage. If you do not have a competing offer, threats are weak. Documentation of your unvested stack is the leverage.
- Negotiating without documentation. “I’m leaving a lot of equity” is weak. “$847K in unvested RSUs over the next 22 months” is strong.
- Underselling future refreshes. If your current employer has refreshed you stably for 2-3 years, that pattern is documentable and worth bridging.
What competing offers do
The strongest negotiation lever is a competing offer at a peer lab. If you have an Anthropic offer in hand and are talking to OpenAI, both companies will move on sign-on and equity to compete. Without a competing offer, you can still negotiate, but the leverage is documentation and walk-away credibility, not a head-to-head bidding war.
Tax considerations
For US candidates: sign-on bonuses are taxed as ordinary income, often in the year received. Large sign-ons can push into higher marginal brackets. Consider with your tax advisor whether deferred or split sign-on structures are tax-advantageous.
For European candidates (Mistral, Anthropic London): UK and France have higher marginal rates than US. Sign-on bonuses still useful but the after-tax value is meaningfully lower than the headline number.
Frequently Asked Questions
Should I disclose my current total compensation?
You generally have to disclose past compensation; in many jurisdictions employers cannot ask. Disclosing your unvested equity (the buyout target) is necessary and useful. Disclosing your current total comp is sometimes useful, sometimes not — it depends on whether your current comp is above or below the lab’s typical band.
What if my unvested stack is small but I want a big sign-on?
The buyout case is weak. Frame the sign-on differently — as a relocation cost, as compensation for the year-1 ramp, or as bridging vesting cliffs at the new role. These framings produce smaller sign-ons but are still negotiable.
Is sign-on different at smaller labs vs OpenAI / Anthropic?
Yes. Smaller labs (Mistral, Cohere, xAI) have less institutional flexibility on sign-on. Sometimes the sign-on is replaced with larger equity grants instead, which is a different kind of deal.
Should I take a smaller sign-on for a larger equity grant?
Depends on your view of the lab’s liquidity timeline. Cash now is certain; equity is uncertain. For senior+ engineers with strong financial cushion, equity-heavy makes sense. For candidates without cushion, cash-heavy is safer.
Can I negotiate refresh grants in writing?
Sometimes yes, sometimes no. Some labs will commit a refresh schedule in the offer letter; others insist refreshes are at-will. Ask explicitly.