The AI talent market in 2026 is unlike anything that came before it. OpenAI and Anthropic are competing for the same narrow pool of researchers and engineers who could move elsewhere for life-changing money. They know it. You should too. Across the 500+ offers I've negotiated, AI lab comp is the hardest to benchmark and the easiest to get wrong.
But AI lab offers don't negotiate like FAANG offers. The equity isn't liquid. The bands are tighter. The timeline is shorter. And the questions you need to ask before you counter are fundamentally different from what you'd ask Google or Meta.
Most guides on this topic were written by people who've negotiated at Big Tech and are applying the same playbook. That playbook is incomplete here.
Key Takeaway: Negotiating an AI lab offer is more about understanding the equity structure than finding the right words. Know what you're being offered before you decide what to ask for.
How These Offers Are Structured (And Why It Changes Everything)
OpenAI: Profit Participation Units (PPUs)
OpenAI doesn't offer traditional RSUs or stock options. They offer Profit Participation Units, which are contractual rights to a share of company profits rather than ownership of company equity.
The mechanics matter for negotiation:
- Vesting: 4-year schedule, 25% per year, with no cliff. You start vesting from day one.
- Liquidity: No public market. You access value through company-facilitated tender offers. OpenAI ran a significant one in October 2025 at a valuation implying roughly $500B, and as of June 2026 is preparing another at $687.69 per share with the company valued at approximately $852B. These are not guaranteed events.
- IPO context: OpenAI confidentially filed for IPO as of mid-2026. A public listing would change how PPUs convert and when liquidity becomes available. The terms in your grant documents govern that conversion. Read them carefully.
- Tax treatment: PPUs can be complex. Depending on elections made at grant, gains may be treated as ordinary income or capital gains. This is not a trivial distinction at the dollar amounts involved. Consult a financial advisor who specializes in pre-IPO equity before you sign.
Total compensation at OpenAI (2026 estimates):
| Level | Role | Total Comp Range |
|---|---|---|
| L4 | Mid-Level / MTS | $520K – $770K |
| L5 | Senior / MTS | $700K – $1.1M+ |
| L6 | Staff / MTS | $1.0M – $1.6M+ |
Base salary for technical roles typically runs $250K to $450K. The rest is PPUs. Research Scientists often command a 40 to 80% premium over Research Engineers at equivalent levels.
Anthropic: Pre-IPO RSUs
Anthropic uses a more conventional equity structure — RSUs — but in a pre-IPO context that introduces its own complexity.
- Vesting: 4-year schedule with a 1-year cliff, then monthly vesting.
- Liquidity: Tender events have historically occurred every 6 to 12 months. Not guaranteed, but more regular than many pre-IPO companies.
- Valuation context: Anthropic was valued at approximately $965B as of May 2026. RSU grant values are calculated at the internal 409A valuation (used for tax purposes), which will differ from the funding round valuation. The gap between those two numbers matters for how you evaluate the equity.
- Equity as a share of TC: For L4+ roles, equity often represents 70% or more of total compensation. This is extremely high, even by AI lab standards. It means your headline TC number is significantly dependent on a valuation that changes over time.
Total compensation at Anthropic (2026 estimates):
| Equivalent Level | Role | Base Salary | Total Comp Range |
|---|---|---|---|
| L3 (Entry/Mid) | Software Engineer | $220K – $260K | $300K – $380K |
| L4 (Senior) | Senior SWE | $260K – $320K | $380K – $490K |
| L5 (Staff) | Staff SWE | $320K – $380K | $490K – $620K |
| L6+ (Principal) | Lead / Principal | $380K – $450K | $620K – $760K+ |
Note: Anthropic doesn't use a formal L-series publicly. These are approximate equivalents. Some specialized MTS and research roles sit well above these ranges. H-1B filings for senior technical staff have revealed base salaries exceeding $1M for the most senior positions.
Why AI Lab Negotiation Is Different From FAANG
At Google or Meta, you're negotiating within well-documented comp bands that are partly visible on Levels.fyi. Competing offers from peer companies create direct, apples-to-apples comparisons that recruiters can act on immediately. There's a mature playbook.
At OpenAI and Anthropic, three things work differently:
1. The equity isn't liquid. A Meta RSU vests and converts to a share you can sell on a given date. A PPU or pre-IPO RSU doesn't. You're accepting risk that a FAANG candidate doesn't face. That risk should be reflected in the total compensation you accept, and it's a legitimate point to raise in negotiation.
2. The bands are tighter and less flexible. Both labs have been public, in recruiter conversations reported on Blind and Reddit, about maintaining strict comp bands. The salary component in particular is harder to move than at large public companies. This doesn't mean negotiation is futile. It means knowing where to push.
3. Leveling is the highest-leverage negotiation. Because compensation is tightly band-to-band, getting leveled one step higher is worth dramatically more than squeezing margin within a level. A successful level bump at OpenAI from L4 to L5 could mean $150K to $300K in additional annualized TC. Getting $20K more within L4 is a different conversation entirely.
The Negotiation Levers, Ranked
1. Leveling (Highest Impact)
If you believe your experience warrants a higher level than the one offered, this is where to focus. The conversation isn't "pay me more at this level." It's "I think my background maps to the scope and expectations at the next level."
Be specific. What you've shipped, the scale you've operated at, the size of teams or systems you've led. AI labs are technical organizations and they'll respond to technical evidence, not generic "I have more experience" claims.
The window to raise leveling concerns is before or immediately after receiving an offer, not two weeks later. Ask your recruiter early in the process: "How are you thinking about level for this role?" The answer tells you where you are and opens the door to discussing it.
2. Equity Grant Size (High Impact)
Even when base salary is fixed, the size of the equity grant (number of PPUs or RSUs) often has more flexibility, particularly at the senior and staff levels. (See staff engineer salary benchmarks across FAANG and AI labs for what the total comp bands actually look like at this level.) This is the component to push on once you understand the base is at or near band ceiling.
At both companies, equity refreshers are also standard practice after a certain tenure. Asking about the refresh cadence and typical amounts is reasonable and gives you a fuller picture of the total long-term package.
3. Sign-On Bonus (Medium Impact)
Sign-on bonuses are one-time costs and easier for companies to approve than permanent comp increases. Both OpenAI and Anthropic have been known to use sign-on bonuses to close gaps, particularly when a candidate is walking away from unvested equity at a prior company or has a meaningful near-term competing offer deadline.
Frame it accurately: "I have [X] unvested at my current company that I'd be leaving on the table. Is there flexibility to offset that with a sign-on?" That's a direct, quantifiable request — easier for a recruiter to bring to their comp team than a vague "I want more."
4. Vesting Acceleration (Situational)
Some senior candidates have negotiated partial vesting acceleration clauses — particularly for double-trigger acceleration in change of control scenarios. This is more common in offers to candidates at very senior levels (L6+) or for unique research talent. It's worth asking about if you're in that range, but it's not a standard lever for most offers.
5. Base Salary (Lower Impact)
Base salary is the hardest thing to move at AI labs. Both organizations operate with strict bands. You can make a case if you have clear evidence you're at the top of a band or the offer is below market for base (Levels.fyi and H-1B filings are your references here), but expect more friction here than on equity.
The Competing Offer Question
Competing offers from peer organizations are the single most reliable lever in this market, and the labs know it. (If you don't have one yet, here's how to negotiate salary without a competing offer.)
A FAANG offer (Google DeepMind, Meta FAIR, Apple ML) creates a useful comparison because it provides a liquid, verifiable alternative. The implicit argument: "You're asking me to accept illiquid equity at a pre-IPO valuation. The liquid alternative from [FAANG] is $X. To make that tradeoff, the total package here needs to be meaningfully higher to justify the risk premium."
That's a clean, logical case. It doesn't require aggression. It just requires that you're honest about the comparison and prepared to quantify it.
A competing offer from the other lab (OpenAI vs. Anthropic) is even more direct leverage. Both organizations have explicit processes to respond when a candidate is being recruited by the other. Disclosing it, professionally, often triggers faster internal review and real comp movement.
What not to do: invent or inflate a competing offer. At this level of compensation, companies will sometimes ask for verification or for the recruiter to speak with your contact at the other firm. The downside of getting caught is severe.
If you don't have a competing offer, your best alternative leverage is a strong FAANG offer at L5+ or a credible offer from another top AI lab (xAI, Google DeepMind, Scale AI). Having that in hand before you negotiate is worth the extra two to four weeks it takes to run those processes in parallel.
The Timeline Problem
AI labs move faster than traditional tech companies. OpenAI and Anthropic both operate with recruiter timelines that can compress the offer-to-decision window to 7 to 10 business days. That's not a lot of time to gather competing offers, run financial analysis on pre-IPO equity, and negotiate.
The fix is to start parallel processes early. If you're in final rounds at OpenAI, trigger final rounds at Anthropic and at least one FAANG company simultaneously. Don't wait for one offer to land before starting the others. By the time you have an offer in hand, your alternatives should be 7 to 14 days behind you — close enough to produce a real deadline, not so far behind that you're fabricating urgency.
When you receive an offer, ask for the full 7 to 10 days available to review. "I want to review the full package carefully and discuss it with a financial advisor given the equity components. Can I have until [specific date]?" That request is professional, reasonable, and almost always granted.
Questions to Ask Before You Sign
These are the questions most candidates don't think to ask, and they're the ones that determine whether you actually understand what you're accepting.
On PPUs or RSUs:
- What is the current 409A valuation used for this grant?
- What was the strike price / fair market value at the last tender offer?
- How frequently do tender offers or liquidity events occur?
- What are the lock-up provisions if the company goes public?
- How does the equity convert in an acquisition scenario?
On the grant itself:
- What is the total grant size in units, not just dollar value?
- What is the annual refresh cadence and typical refresh amount?
- Is there a performance component to refresh grants?
On leveling:
- What performance benchmarks define movement to the next level?
- What's the typical time-in-level for someone at my seniority?
On taxes:
- What elections are available at grant time and what are the implications? (Ask a financial advisor, not the recruiter.)
If You're Choosing Between OpenAI and Anthropic
This is a real decision for a meaningful number of candidates, and the financial comparison is not straightforward.
OpenAI PPUs and Anthropic RSUs are both pre-IPO instruments, but they're structurally different. PPUs are profit-sharing rights; RSUs are ownership stakes. In a successful IPO or acquisition, the mechanics of how each converts to value differ. The relevant questions: What happens to your units in each liquidity scenario? What are the tax treatments at vesting vs. sale?
Beyond the financial structure, the comparison usually comes down to mission alignment, team, research focus, and culture. Both organizations are intentional about culture fit in ways that most tech companies aren't. A misalignment there will surface quickly, and no amount of comp makes a wrong fit sustainable at an organization where the work demands deep engagement.
The financial comparison is a necessary input. It shouldn't be the only one.
Frequently Asked Questions
Q: Can you actually negotiate with OpenAI and Anthropic?
Yes, but the playbook is different. Leveling and equity grant size are more productive to push on than base salary. Competing offers from peer organizations are the most reliable lever. And the negotiation window is shorter than at traditional tech companies, so preparation before the offer lands matters.
Q: What are OpenAI PPUs worth?
It depends on when you're asking. OpenAI's October 2025 tender offer processed at a valuation around $500B. As of June 2026, the company is preparing another tender at $687.69 per share with a valuation near $852B. PPU value is pegged to those periodic tender prices or, eventually, to a public share price if an IPO proceeds. There is no guaranteed liquidity timeline.
Q: Should I take an OpenAI or Anthropic offer over a FAANG offer?
It depends on what you're optimizing for. FAANG offers provide liquid equity, well-defined comp bands, and predictable TC. AI lab offers provide higher potential upside, a different quality of technical work, and significantly more equity risk. The risk premium on illiquid pre-IPO equity should be real and reflected in the TC difference. If the total comp numbers are comparable, the FAANG offer is the lower-risk choice. If the AI lab TC is materially higher on a risk-adjusted basis, the calculus shifts.
Q: How do I negotiate level at OpenAI or Anthropic?
Raise it early. Ask your recruiter how they're thinking about level before you get deep into the process. When you have the offer, make the case with specifics: scope of prior work, scale of systems you've owned, teams you've led, research impact you've driven. Level negotiations at AI labs respond to technical evidence, not seniority claims.
Q: What's a good competing offer to have when negotiating at an AI lab?
An offer from the other top AI labs (OpenAI vs. Anthropic vs. Google DeepMind vs. xAI) carries the most weight. A strong L5+ FAANG offer is a credible alternative that also highlights the liquidity risk premium question. Any offer from a well-known organization in the same technical domain is useful. Generic or non-comparable offers add less.
Q: How should I think about the equity if an IPO is coming?
With caution and a financial advisor. A confidential IPO filing doesn't guarantee timing, valuation, or terms. Lock-up periods typically prevent employees from selling for 6 months post-IPO. The 409A valuation used for your grant may be different from the IPO price. Tax obligations can crystallize at vesting in ways that require cash on hand. None of this means you shouldn't take the offer. It means you should understand what you're accepting before you accept it.
