Compa Takes NVIDIA's Pay War to Payscale

The Series B Signal

Compa closed a $35 million Series B led by Jump Capital in January 2026, bringing total funding to $48.9 million, per Simplify. CEO Charlie Franklin told Axios Pro the capital would support growth as boards demand precision on payroll decisions that now run into billions. The company's platform ingests 1.2 million live offer observations, 4.8 million employee records, and 1.2 million equity grants across 42 countries and 3,098 job families, a dataset built from HCM, ATS, and equity administration systems.

Traction at the top of the market validates the approach. Compa's customer roster includes NVIDIA, Stripe, DoorDash, OpenAI, Moderna, Workday, Ulta, and Target, per the company's own disclosures. Thirty percent of Fortune's Most Admired companies, PRNewswire's data shows, already trust the platform through its Syndio integration, which, per PRNewswire, embedded Compa benchmarks into Pay Finder in May 2025. Mercer struck a strategic alliance, Yahoo Finance reported, to feed its global compensation data into Compa's Analyst Agent. Micron's VP of Global Total Rewards, Athar Siddiqee, said the combined view (equitable range, internal range, live market data) lets his team "move faster, make more consistent and appropriate decisions, and avoid surprises down the line."

Compa lists 75 employees on Built In (201–500 on Simplify), with headquarters in Irvine, California and growing sites in Denver and San Francisco. The company is hiring an Enterprise Account Executive with a mandate to "drive strategic deals by targeting key peer companies in our anchor tenant strategy," per job postings on Simplify and Ashby. The role carries a $100,000–$125,000 compensation band on Simplify; board data shows a median of $200,000 across four salaried roles.

The anchor tenant language appears in Compa's own job descriptions: land one flagship customer in a vertical, then use that reference to pull in peers. It's a playbook the company is deploying now. The Enterprise AE will own the full cycle from creative prospecting to close, collaborating with marketing, insights, and product. Compa's August 2026 product updates, Simplify noted, include a new-grad compensation report; June 2026 releases covered K-shaped market analysis and leveling distribution data.

Why Surveys Lag

The compensation benchmarking model that dominated for decades rests on a simple premise: highly trained professionals triangulate between aggregated survey data points to set pay ranges for specific roles. Payscale describes the process as leveraging third-party salary data, with practitioners finding the best available job matches and interpolating between them. That workflow assumes the underlying data changes slowly enough that an annual or semi-annual refresh captures the market. Korn Ferry's March 2026 Global Total Rewards Pulse Survey makes clear that assumption no longer holds: "Pay models are under strain as AI-related jobs evolve faster than market data can keep pace."

The lag is measurable. Traditional surveys typically trail the labor market by six to twelve months, while AI-driven platforms ingest real-time job posting data and update estimates weekly, often within a 5 percent margin of error. The 2026 Winter Salary Survey from the National Association of Colleges and Employers illustrates the velocity: starting salaries for computer science majors are expected to rise nearly 7 percent year over year.

AI roles themselves are stretching the old frameworks. Korn Ferry reports the most common premium for AI-related talent sits at 10 to 15 percent above peer roles, while AI-enhanced positions such as Prompt Engineer and AI Product Manager have seen salary jumps of 30 to 45 percent over the past two years. Traditional software engineers now command 5 to 10 percent more when they demonstrate proficiency with AI-assisted coding tools like GitHub Copilot. Where AI reduces job scope, compensation reductions are limited to 5 to 10 percent, with organizations opting for role redesign over pay cuts.

"External benchmarking remained supreme, however, but surveys needed to evolve in real time to keep up." — Korn Ferry Global Total Rewards Pulse Survey, March 2026

Compensation teams are responding by layering tools. The same Korn Ferry survey finds most organizations at early to moderate maturity in applying AI to total rewards, with experimentation and operational adoption increasing over the past year. Simple applications around communication, grading, and pricing are the first to see impact. Payscale's January 2025 outlook noted that recent advances in HR tech integrations, AI techniques, and computational tools would enable processing and aggregation of pay data in near real time rather than yearly, with explainable models filling gaps transparently. WTW's January 2026 survey confirms growing use of scenario modeling to test budget trade-offs, identification of pay equity risks before decisions finalize, and smarter market pricing using broader, more dynamic data sets.

The shift is not a replacement of human judgment but a reshaping of it. WTW puts it directly: "AI is not replacing human judgment, it's reshaping it — compensation teams are gaining the ability to move faster, test assumptions, and explain decisions with greater confidence and transparency." Korn Ferry projects that most organizations expect greater levels of role transformation in the next two to three years, and AI adoption will accelerate as tools become more integrated into enterprise systems. The emerging requirement is clear: pricing and pay grading tools must become dynamic, and new pay models will be needed to support organizations through their AI transformation.

A Market Fractured Into Tiers

The talent shortage in frontier tech isn't cyclical — it's structural. ManpowerGroup surveyed 39,063 employers and found AI skills are now the hardest in the world to hire for, beating all of engineering and IT for the first time. Demand for AI-fluent workers grew sevenfold in two years, from 1 million to 7 million, per LinkedIn Economic Graph data via the World Economic Forum. Indeed Hiring Lab's January 2026 update makes the divergence concrete: AI postings sit 134 percent above their February 2020 baseline while total job postings are only 6 percent above the same baseline.

That pressure radiates well beyond pure AI labs. Defense contractors Anduril, Shield AI, and Palantir pay clearance-cleared engineers a premium that often exceeds frontier-lab base salaries. At Nvidia, RSU appreciation has made stock the dominant pay component regardless of what the offer letter says. OpenAI's profit-participation units vest like stock but settle on a different schedule, and recent hires have not yet seen a full tender cycle. At Anthropic, the equity is on paper until acquisition or IPO.

The numbers expose a market that has fractured into tiers. The Bureau of Labor Statistics pegs baseline software roles at $133,080 median. Mainstream AI/ML engineers earn $170,000–$245,000 total. Frontier-lab software engineers clear $600,000–$795,000 median total comp. OpenAI L5 employees pull $1.15 million total comp ($336,000 base + $774,000 stock per year) as of May 2026, with senior ICs clearing $1.28 million at the top of band. Anthropic's equivalent data shows median total comp at $600,000, with senior software engineers earning $316,000 base plus $247,000 in stock. The 2026 AI pay staircase spans six times from the BLS baseline to OpenAI's median — a spread no other software specialty comes close to matching.

Specialization compounds the spread. CUDA and GPU optimization specialists command $300,000–$500,000-plus total comp — the rarest skill in AI. AI safety and alignment expertise carries a 45 percent premium increase since 2023. LLM fine-tuning specialists earn 25 to 40 percent above generalist ML engineers. Agentic AI workflows drove a 25 to 35 percent spike in mid-level pay to $190,000–$270,000 in 2025. AI agent development, the newest high-demand specialization, runs $200,000–$320,000 and is growing at 136 percent year-over-year.

The talent deficit is structural: the US projects 1.3 million AI job openings over two years, but available supply covers fewer than 645,000. Entry-level hiring at the 15 largest tech firms fell 25 percent from 2023 to 2024, concentrating demand on mid-to-senior talent. Firms offering below a $200,000 base salary floor for senior AI talent face 114-day average time-to-fill, compared to 52-plus days for the broader tech market. Over half of AI roles now sit outside traditional tech companies, intensifying cross-industry salary competition.

Energy and biotech are feeling the same pull. Defense-AI, consumer-AI, infrastructure-AI, and foundation-model-AI roles pay differently even within the same company. International offices (Anthropic London, DeepMind London, OpenAI Dublin) pay 25 to 40 percent less than equivalent US roles on a USD basis. Most companies still pay a Bay Area premium of 15 to 25 percent over other US locations; the exceptions are frontier labs, which increasingly pay location-agnostic comp benchmarked to San Francisco for senior talent.

Static salary surveys published annually cannot track this. A compensation team relying on last year's Radford data would miss the 56 percent wage premium for AI-skilled roles (up from 25 percent the prior year), the junior-level inversion where AI professionals average $173,500 total comp exceeding director-level averages of $152,600 at some organizations, and the reality that 3.4 open AI roles exist for every qualified candidate. Real-time offer data from ATS integrations (what Compa's platform ingests) becomes the only way to price a requisition before the candidate accepts a competing offer.

Four Engines, One Race

The enterprise compensation market has long been a three-horse race between Payscale, Visier, and Radford — each built on a different architecture of data collection. Payscale, founded in 2002, processes 55 million profiles via API. Visier, founded in 2010 by John Schwarz and Ryan Wong, bet on cloud-based workforce analytics. Radford standardized the consultancy-led survey model that Fortune 500 compensation teams have relied on for decades. Compa, founded in 2020, enters with a fundamentally different data engine: real-time offer data pulled directly from applicant tracking systems.

Payscale's scale is undeniable. Its Market Match algorithm processes 55 million profiles to generate role-specific reports, and SelectHub's 2026 analyst scorecard rates it 79 out of 100, best-in-class for mobile capabilities and rewards recognition. User satisfaction sits at 85 percent across 201 reviews. Users flag accuracy gaps for niche roles, a clunky interface, integrations that demand heavy IT involvement, cumbersome custom report building, restricted widget customization, and pricing that can be steep for smaller organizations.

Visier carries a $1 billion valuation on $147.3 million in revenue with 574 employees, backed by $216.5 million in funding. Its 5,000 customers use it for workforce planning and analytics beyond pure compensation. SelectHub gives CompAnalyst (Visier's compensation module) a 70 analyst score but 88 percent user satisfaction from 12 reviews, best-in-class for dashboard and reporting. Pain points include an outdated interface, cumbersome custom reports, IT-heavy integrations, and no live pay data. Visier's strength is breadth; its compensation module is one piece of a larger people-analytics platform.

Radford operates on a different timeline. Its benchmark data comes from exhaustive manual submissions that take months to compile, require dedicated analyst teams, and introduce error surfaces at every handoff. As Compa's own blog noted in 2023: "Survey data lags. Tech stock valuations are down as much as 80%. Surveys don't show this because the data it's compiled happened months, or even years, before." That lag is structural. When equity-heavy compensation packages swing violently with private-market valuations, a survey snapshot from six months ago isn't just stale — it's misleading.

The market is splitting between platforms that tell you what people were paid and platforms that tell you what companies are offering right now.

The Series B signals investor conviction in the offer-based model. The company's partnership with Mercer (announced on Compa's blog) is a signal that the incumbent survey giant sees real-time offer data as a necessary complement. Mercer brings the survey depth and global coverage Compa lacks; Compa brings the live market pulse Mercer's surveys can't capture.

The competitive dynamic is no longer purely zero-sum. Payscale and Visier are layering AI onto their existing databases. Radford is embedding its data into broader Mercer workflows. Compa is betting that the anchor-tenant strategy (landing one flagship enterprise per vertical, then expanding through peer referrals) can overcome the switching costs that protect incumbents. The role's Simplify band is $100,000–$125,000; board data shows a $200,000 median across four salaried roles. That hire is a marker that the Series B runway is being deployed into go-to-market.

Landing the First Flagship

Compa's job posting for the Enterprise Account Executive makes the strategy explicit: "You'll pursue strategic deals by targeting key peers in that strategy, helping compensation leaders unlock real-time market intelligence." The Series B capital is funding a sales motion built on this logic. The platform ingests the dataset outlined above. That dataset becomes more valuable with each enterprise customer that plugs in their ATS — especially when those customers are frontier-tech companies generating the very compensation outliers the platform tracks.

AI deal sizes run two to four times larger than equivalent SaaS deals because buyers are replacing headcount or process, not just buying software. Sales cycles stretch six to eighteen months for enterprise. Purchases involve IT, legal, security, data teams, and often a CISO review. Buyers expect sellers to speak competently about model capabilities, data privacy, inference costs, and integration patterns. The Enterprise Account Executive role carries a compensation band consistent with the 15 to 35 percent AI premium over comparable SaaS roles that market data shows.

"Analyst Agent gives comp teams instant visibility into what's moving, where, and why, turning months of analysis into minutes." That claim from Compa's site captures the value proposition an anchor tenant buys into. Compa's customer quotes name specific outcomes: "Being able to use Compa's data allows us to decide where we want to be more competitive and to be more curated." Another: "Compa helps us layer real-time market data with survey data so we can have fact-based discussions with recruiters and adjust our pay guidance accordingly." These are testimonials about leverage in a talent war where a single mispriced offer can delay a critical hire.

McKinsey research estimates 30 percent of sales tasks are automatable, and reps leveraging AI tooling see a 35 percent productivity boost. Compa's own sales team operates with a lean model: That source shows the same median band for four salaried roles, consistent with a model where AI agents handle research, outreach personalization, and analytics that used to require headcount.

The anchor tenant strategy only works if the anchors are real. The next anchor tenant will likely come from the same peer set: space, defense, robotics, AI, energy, biotech. Each win deepens the dataset for the rest. Real-time compensation intelligence is a perishable good. The first mover in each vertical captures the freshest signal. Compa's Series B buys the sales capacity to reach them before the incumbents do.

Transparency Forces the Issue

The pay transparency wave has moved from compliance headache to strategic imperative for frontier-tech companies. By early 2026, roughly half the U.S. workforce (over 60 million workers) will fall under salary disclosure requirements across more than a dozen states plus Washington, D.C., up sharply from 2023 coverage levels. For space, defense, robotics, AI, energy, and biotech firms that hire across state lines and rely on remote or distributed teams, the patchwork is immediate: California, New York, Colorado, Washington, Illinois, Minnesota, Massachusetts, Vermont, New Jersey, and Oregon all have active laws with different employee thresholds (from 5 to 50), varying definitions of "pay range," and distinct posting rules. Several explicitly address remote positions. Delaware joins in September 2027; Columbus, Ohio enforcement begins January 2027; Maine's bill awaits concurrence. The EU Pay Transparency Directive adds another layer for any firm with European operations — all member states must transpose by June 2026.

The compliance penalties are concrete. Colorado fines up to $10,000 per non-compliant posting. New York City can levy $250,000 for repeated violations. But as MorganHR frames it, "the panic isn't about posting numbers — it's about posting numbers you can defend." Most organizations discover they don't actually have a compensation structure when transparency forces them to articulate one publicly. Three structural failures emerge immediately: inconsistent job architecture, unexplainable range overlap, and undocumented pay decisions. Recruiting velocity suffers when you can't quickly determine appropriate ranges for open roles. Internal equity erodes when similar positions receive different treatment based on which manager hired them, when they joined, or how aggressively they negotiated. Manager credibility collapses when they can't explain differences to their teams.

For frontier-tech hiring, the specialization problem compounds the compliance problem. Software engineering compensation in 2026 spans a four-to-five-times range depending on company tier, level, location, and timing. AI-enhanced roles (prompt engineers, AI product managers) have seen 30 to 45 percent salary jumps over two years. Traditional survey-based benchmarking (Radford, Payscale) looks backward; Compa's ATS-integrated offer data looks at what's happening now. The anchor tenant strategy Compa is pursuing (landing key peer companies in each vertical) reflects this reality: frontier-tech firms need benchmarking from their actual competitive set, not broad industry aggregates.

Cornell research adds a gender-equity dimension that frontier-tech cannot ignore. Across four studies, women showed a stronger preference for jobs with narrower salary ranges compared to men, and this preference correlated with less assertive negotiation behaviors. Wide ranges in job ads may discourage female applicants and perpetuate the very pay gaps transparency laws aim to close. The mitigation is structural: when ads included context about typical starting salary and how final offers are determined, the gender gap in application decisions and negotiation behaviors disappeared. That context requires a documented compensation philosophy and consistent range methodology — typically 30 to 50 percent spread from minimum to maximum, with narrower ranges (20 to 25 percent) for standardized roles and wider (up to 50 percent) for specialized positions with significant experience variation.

The job posting for Compa's Enterprise Account Executive is live. The anchor tenant it lands will determine whether Compa's real-time offer data becomes the benchmark the rest of the market chases — or whether the incumbents absorb the innovation and keep the old playbook running.

Working in frontier tech? Zero G Talent tracks the openings: see every open Compa role, browse frontier tech jobs, the companies hiring, and the people building the field.