OpenAI raised the floor on senior-research compensation by approximately 28 percent across a quiet, internally-named "Q1 reset" that landed in employee pay statements between mid-February and the second week of March. The reset, confirmed by two current OpenAI staff and one senior recruiter who has tracked offer letters from the lab through the period, lifted the bottom of the senior-IC band — roughly the L5-equivalent and Member of Technical Staff line — to a new floor reportedly above $850K total comp, with the median for the same band sitting in the $1.1M to $1.4M range and the top of the band well past $2M for staff with multi-paper publication records or model-launch ownership. Meta and Google responded within fourteen days. The cascade is the most consequential compensation event in AI hiring of the year so far.
The structural shape of OpenAI's reset is what matters. The lab did not raise base salaries across the company. It raised the floor. The decision, per the recruiter we spoke with, was driven by a Q4 2025 internal analysis showing that OpenAI was losing offers at the senior-IC level not at the top end of comp negotiations but at the floor — candidates were declining offers that came in at the company's then-current senior-IC starting band because Anthropic, Thinking Machines Lab, and the larger compensation packages emerging from the Microsoft AI unit had reset what a credible senior-research offer looked like. OpenAI's senior leadership, with Sam Altman closing the decision personally per internal accounts, recalibrated the floor and pushed the change through quietly enough that the public press did not surface it for nearly six weeks.
Meta's response landed first. Inside a fortnight of OpenAI's reset hitting the recruiter network, Meta's Superintelligence Labs unit — the rebranded successor to FAIR after the 2025 reorganisation — raised its own senior-research floor in what one external recruiter described as "a calibration to the new market," with package terms structured around equity refreshes rather than base lifts. Meta's structural advantage in this market remains its public-equity liquidity: a senior research engineer at Meta SL receives Meta common stock that vests on a four-year cliff and can be sold immediately, in contrast to OpenAI's tender-offer-and-PPU structure which depends on the company running the next secondary round at or above the current $500B valuation. The cash-equivalent gap between OpenAI and Meta at the senior-IC level narrowed materially after the reset, and Meta's recruiting machine has been pressing the liquidity advantage in candidate conversations through April.
Google's response landed second, and was structurally different. Google DeepMind did not match OpenAI's floor lift on a like-for-like basis. Demis Hassabis's people-side organisation instead introduced a one-time "scale-of-impact" cash component for senior research staff with multi-paper or model-launch involvement — reportedly in the $400K to $900K one-time range — paid against a 24-month retention cliff. The mechanism is closer to a retention bonus than to a band reset, and the structure preserves Google's overall compensation architecture while addressing the specific senior-IC retention risk OpenAI's reset created. Per a Google DeepMind insider, the design choice was deliberate: Hassabis and his leadership team have been resistant to structural band lifts that create permanent cost-base inflation, and the one-time mechanism gives them a defensible alternative.
The named principals here are worth tracking. Sam Altman closed the OpenAI decision. Dario Amodei has been operating Anthropic's compensation architecture, which we wrote about separately, on a comparable but structurally different model that emphasises offer-stage founder closing rather than band-wide resets. Mustafa Suleyman, running Microsoft AI's two-track hiring framework, has been able to pay frontier-lab rates inside Microsoft's enterprise structure without forcing a Microsoft-wide reset. Mark Zuckerberg has been personally involved in Meta SL retention conversations through Q1, per one external recruiter who described a candidate-side outreach that included a Zuckerberg-led closing call. Demis Hassabis has stayed out of the public conversation but has been the architect of Google's one-time-bonus response.
Three observations matter for the comp landscape through the rest of 2026.
The first is that the senior-research compensation band is now structurally above $1M total at every credible frontier lab. The era when a "frontier-lab senior research engineer" could be hired at $600K to $800K total is over, and the recruiters still working off pre-2026 comp models are losing offers at a measurable rate. Per the senior recruiter we spoke with for this piece, the offer-decline rate at $700K total comp for a senior research engineer with a publication record is now north of 60 percent, against roughly 25 percent eighteen months earlier.
The second is that the cascade from frontier labs is now reaching enterprise. Microsoft's Copilot product groups, Salesforce's Agentforce engineering, Amazon's AGI unit, and the senior AI groups inside the largest US banks are all reportedly recalibrating bands in response. The enterprise comp architecture for AI senior-IC roles in 2026 will look meaningfully different by Q4 than it did at the start of the year.
The third is that the cascade is asymmetric across geographies. The reset has hit San Francisco hardest and is being absorbed in London with a roughly six-month lag. Dubai and Riyadh, where the comp architecture is built around cash-and-tax advantages rather than equity, have absorbed the reset already through the compensation arithmetic we wrote about in the Bangalore-Dubai-SF corridor briefing. Tokyo has not yet adjusted; Singapore's IMDA-aligned employers are at the start of their own reset cycle.
Forecast: a second OpenAI reset will land in Q4 2026 if the compounding effect of Anthropic's offer pressure and Microsoft AI's enterprise comp absorption continues at the current rate. The senior-IC band at the top frontier labs will likely sit between $1.4M and $2.5M median by year-end. The labs that absorb the cost without breaking equity dilution discipline will be the ones that retain the talent they hire. The labs that try to fight the reset on principle will lose the next eighteen months of senior-IC mobility to the ones that don't.
For the funnel-side dynamics behind the comp pressure, see How Anthropic restructured its talent stack. For the geographic arbitrage that compounds it, see The new AI talent corridor: Bangalore → Dubai → San Francisco.