AI lab compensation in 2026 is at the top end of the technology compensation market, but the underlying structure is more complicated than at FAANG. Equity in private AI labs comes in multiple non-standard forms — OpenAI’s Profit Participation Units (PPUs), Anthropic’s pre-IPO stock options or RSUs, DeepMind’s Alphabet RSUs, smaller-lab pre-IPO equity. Liquidity varies. Tax treatment varies. Negotiation levers differ from public-company comp negotiations. Candidates entering the market without understanding the structure under-negotiate.
This piece is a deep dive on how AI lab comp actually works, what to negotiate, and what the typical 2026 numbers look like at each level.
The four major equity structures
OpenAI: Profit Participation Units (PPUs)
OpenAI uses a non-standard instrument called PPUs — units that give the holder the right to a share of future profit distributions, capped at a multiple of the original grant value. The structure resembles equity but has specific differences:
- PPUs vest over a 4-year schedule similar to RSUs.
- Liquidity is via secondary tender offers, typically annually. The company has run several large tenders since 2023.
- Value is capped at a multiple of the original grant. The cap has been adjusted upward in recent rounds but is a real constraint.
- Tax treatment is non-standard; treated as ordinary income at vesting in most cases.
The realized value of PPUs depends on the company’s profit trajectory and the cap. Recent senior+ packages have been substantial, but candidates should understand the cap before signing.
Anthropic: Pre-IPO equity
Anthropic uses a more standard pre-IPO equity structure: stock options or RSUs in the private company, vesting over 4 years with a 1-year cliff. Liquidity is via:
- Secondary tender offers (the company has run several since 2023).
- Eventual IPO or acquisition.
The structure is more familiar to candidates coming from public companies. Realized value depends on the company’s eventual exit, but the recent secondary rounds have valued the company aggressively, and tender-offer liquidity has been meaningful.
DeepMind: Alphabet RSUs
DeepMind employees receive Alphabet (GOOG) RSUs, the same instrument as Google product team employees. This is the simplest structure of the three: publicly tradeable equity that vests over 4 years.
The advantage: liquidity is immediate at vesting. The disadvantage: Alphabet stock appreciation is bound by Alphabet’s overall trajectory, not DeepMind-specific performance. Candidates targeting the upside of “the AI lab succeeds” will not capture it directly via the equity grant.
Smaller AI labs: pre-IPO with more uncertainty
Mistral, Cohere, xAI, Inflection, Character.AI — these companies use pre-IPO equity but with more variability. Some have run secondary rounds; others have not. Some have raised at consistently rising valuations; others have flat or down rounds. The realized value is more uncertain, and candidates should weight comp comparison toward cash and against equity at these labs.
Typical 2026 packages by level
Senior+ packages at the major labs (rough ranges, US-based):
| Level (approx) | Cash (base + bonus) | Equity / year | Total comp / year |
|---|---|---|---|
| Senior (L4-L5 equiv) | $220-300K | $280-500K | $500-800K |
| Staff (L6 equiv) | $300-400K | $400-1M | $700K-1.4M |
| Senior Staff / Principal (L7-L8) | $400-550K | $800K-2M | $1.2-2.5M+ |
| Distinguished / Fellow | $500K+ | $1.5M+ | $2M-5M+ |
Important caveats:
- The equity component value depends on liquidity and exit. PPU caps and pre-IPO equity haircuts are real.
- Sign-on bonuses are common at senior+, often $100K-500K to bridge buyout of unvested equity at a previous job.
- Refresh grants vary; some labs are aggressive on refreshes (matching original grant size annually), others are not.
Comparison with FAANG
FAANG senior+ ranges (US 2026):
| Level | FAANG total | AI lab total |
|---|---|---|
| Senior | $400-650K | $500-800K |
| Staff | $650K-1M | $700K-1.4M |
| Senior Staff / Principal | $1M-1.8M | $1.2M-2.5M+ |
The gap widens at senior levels and beyond. The reason is straightforward: top-end AI talent is a constrained supply, and the labs are paying premiums to attract and retain it. The gap was smaller in 2023 and has grown through 2024-2026.
What to negotiate
The negotiation levers at AI labs differ from FAANG:
- Sign-on bonus: the most flexible item. Push hard, especially if you are leaving unvested equity.
- Equity grant size: moderately flexible. Companies have target ranges per level; pushing to the top of the range is reasonable.
- Base salary: less flexible. Most labs have tight bands.
- Refresh promise: sometimes negotiable, often by getting written commitment to a refresh schedule rather than verbal.
- Level (which has cascading effects on all the above): the highest-leverage thing to negotiate. A level mismatch is the single biggest source of under-comp.
Tax considerations
For US-based candidates:
- RSUs (DeepMind) are taxed as ordinary income at vesting.
- Pre-IPO equity (Anthropic, Mistral, etc.) generally has no tax until liquidity. Tax treatment at liquidity depends on whether the equity is options or RSUs and on the holding period.
- PPUs (OpenAI) are generally taxed as ordinary income; the structure is non-standard so consult a tax advisor specifically for PPU mechanics.
- Sign-on bonuses are taxed as ordinary income, usually in the year received.
For European candidates (Mistral Paris, Anthropic London, etc.), local tax structures dominate. France’s marginal rates over €200K are above 45%; UK’s are above 45% over £125K; both higher than typical US state-and-federal effective rates.
What’s realistic vs aspirational
Realistic for 2026 senior+ candidates:
- $700K-1M total comp at senior levels at the major labs.
- Sign-on bonus of $100-300K.
- Refresh grants commensurate with role.
Aspirational (achieved by exceptional candidates with strong leverage):
- $1.5M+ at senior levels (rare, requires multiple competing offers).
- Sign-on bonus of $500K+ (requires meaningful unvested equity at prior employer).
- Above-band equity grants (requires being a high-priority hire, e.g., a known senior figure in the field).
Frequently Asked Questions
What if my prior employer is FAANG and my unvested equity is huge?
This is the strongest sign-on negotiation lever. Document the unvested value and ask the lab to bridge it. Multi-hundred-thousand-dollar sign-ons are common in this case.
How risky is pre-IPO equity?
It depends on the company. Anthropic and OpenAI have repeatedly tendered at rising valuations; the equity has appreciated meaningfully. Smaller labs have more variance. Treat pre-IPO equity as having some optionality value but discount it from the headline number when comparing offers.
Are PPUs worth their headline value?
The realized value depends on the cap and on the company’s profit trajectory. Recent senior+ packages have realized substantial value through tender offers. The cap is a real constraint but has been adjusted upward in recent rounds.
Should I take the AI lab offer over FAANG?
Higher comp and more interesting work, but more variance. If you need stability, FAANG is safer. If you want the highest-leverage outcome and are willing to accept more risk, an AI lab is likely better.
How often do AI labs run secondary tender offers?
OpenAI and Anthropic have each run roughly annual tender offers since 2023. Smaller labs are less frequent. Plan for liquidity windows of 12-24 months between tenders, not at-will.