The economics of non-fungible engineers: why tech compensation defies conventional logic

Engineers are priced like real estate, not commodities

The 2024-2025 tech labor market has added new complexity: post-layoff hiring is recovering selectively, AI is restructuring which skills command premiums, and return-to-office mandates are reshaping geographic arbitrage. Understanding these dynamics requires abandoning the fiction of a unified "tech job market" and recognizing that at least three distinct markets exist simultaneously.

The data conclusively demonstrates that software engineers with identical titles experience 300-500% compensation variance depending on employer. Levels.fyi data shows a Senior Software Engineer (L5 equivalent) earns a median of $400,000 at Google but just $182,000 at Microsoft—a 2.2x difference for nominally equivalent roles. The spread widens further when comparing Big Tech to traditional enterprises: senior engineers at Meta average $380,000 while their counterparts at Walgreens earn $149,000-$190,000.

This variance isn't limited to the senior level. Entry-level positions show similar dispersion: a Google L3 engineer commands $184,000 in total compensation while non-FAANG entry-level roles typically pay $80,000-$120,000. At the principal level, the spread becomes extreme: Microsoft's principal engineers top out around $445,000 while Google principals can exceed $1.2 million, representing a 170% gap for supposedly equivalent seniority.

The percentile data reinforces this picture. Blind's salary database shows software engineers at the 25th percentile earn $125,781 while those at the 90th percentile earn $328,833, a 162% spread within the same profession. Within Meta alone, the 25th-to-90th percentile range spans from $200,000 to $513,000. These aren't outliers; they reflect the structural reality that engineering talent, like real estate, derives value from location and context rather than inherent standardization.

Value extraction explains the compensation stratification

Different company types extract fundamentally different value from engineering talent, which rationally justifies dramatically different pay scales. The compensation hierarchy follows a clear pattern tied to revenue-per-employee and directness of value attribution.

Quantitative trading firms sit at the apex because engineering improvements translate directly into profit. A C++ developer reducing latency by microseconds at Citadel can generate millions in P&L—the connection between technical work and revenue is immediate and measurable. This explains why quant firms pay $350,000-$500,000 for new graduates and $500,000-$800,000 for mid-career engineers. Jane Street, Hudson River Trading, and Two Sigma compete for a talent pool where losing an engineer to a competitor means that person will trade against you. The stakes justify compensation that would seem absurd in other contexts.

Big Tech occupies the next tier because software serves billions of users at near-zero marginal cost. Google generates $1.46 million in revenue per employee; Meta generates $1.93 million. When a single engineer's feature reaches two billion daily active users, the value multiplication is extraordinary. FAANG compensation reflects this: entry-level ranges from $150,000-$200,000, senior from $300,000-$500,000, and staff engineers from $500,000-$700,000.

Traditional enterprises pay dramatically less because technology functions as a cost center rather than a revenue driver. Walgreens software engineers earn $88,000-$140,000 not because they're less skilled, but because their code supports operations rather than generating direct revenue. The business model simply cannot sustain Big Tech compensation—retailers generate $170,000-$200,000 per employee compared to tech companies' $1-2 million.

The critical insight is that a $500,000 Citadel engineer might generate $5-50 million in value through trading system improvements, while the identical engineer at Walgreens cannot generate similar value regardless of skill. There's no capital to deploy against their improvements, no direct monetization path, and their output serves operations rather than revenue generation.

The myth of market rate collapses under examination

The concept of "market rate" fails as a meaningful benchmark for tech compensation because there is no unified market—only distinct tiers with minimal overlap. Gergely Orosz's research at The Pragmatic Engineer, based on analyzing 1,000+ compensation data points, found that tech compensation follows a trimodal distribution rather than a normal curve.

Tier 1 companies (local businesses treating tech as IT) pay €50,000-75,000 for senior engineers in markets like the Netherlands. Tier 2 companies (ambitious startups, competitive local firms) pay €75,000-125,000. Tier 3 companies (Big Tech, trading firms, well-funded scaleups) pay €125,000-250,000 or more. Public salary comparison sites like Payscale overwhelmingly capture Tier 1 data, making their benchmarks irrelevant for anyone competing in higher tiers.

Job posting data compounds the problem through systematic selection bias. Since 2024, job openings have outnumbered actual hires by more than 2.2 million per month according to BLS data. Many posted positions are "ghost jobs" that remain unfilled indefinitely because companies post below-market rates and wait for desperate or uninformed candidates. Research indicates 70-80% of jobs are filled without ever being posted publicly, meaning job boards capture the harder-to-fill positions rather than representative market rates.

Salary surveys suffer their own methodological failures. Data is typically collected annually and released months later, making it outdated by publication. Manual entry creates errors—if even a small percentage of participants misclassify roles or mistype numbers, benchmarks skew significantly. The mismatch between survey peer groups and actual competitive sets means companies benchmark against irrelevant comparisons. Radford's tech survey contains 899 unique job titles, leaving substantial room for misinterpretation.

The practical result is that a junior engineer at Meta (E4) earning $200,000-$250,000 out-earns a "senior engineer" at typical Tier 1 companies making $80,000-$120,000. Same street, same city, same profession—vastly different "markets."

The 2024-2025 hiring landscape shows selective recovery

The tech job market has entered a phase of selective recovery following the 2023 layoff wave that eliminated approximately 264,000 positions globally. Hiring has rebounded, with job openings up 29.5% from early 2023 lows and TrueUp reporting over 200,000 open positions at startups and public companies. However, this recovery is highly stratified by role and seniority.

Demand concentrates in specialized areas: AI/ML job postings increased 170% from January 2024 to January 2025, while frontend engineering positions declined 24% and mobile engineering dropped over 20%. The hardest roles to fill remain backend engineers (41% of companies report difficulty) and engineering managers (38%). Entry-level positions face the tightest squeeze, with many companies reducing junior hiring while competing fiercely for senior and specialized talent.

Salary trends show moderation rather than continued growth. Stack Overflow's 2024 Developer Survey found most developers reporting $10,000 less than the previous year. Full-stack developer median salaries fell 11% globally, backend developers saw a $9,000 decrease, and site reliability engineers dropped $15,000. ADP research found developer base pay rose only 24% from January 2018 to January 2024, compared to 30% for all U.S. workers—indicating relative salary compression for the profession overall.

The exception is AI specialization. Levels.fyi reports AI engineers earn a 17.7% premium over non-AI peers and saw 34% year-over-year compensation growth. Senior AI roles approach $200,000-$300,000, and AI-specific job postings now dominate the hiring landscape.

Remote work's geographic arbitrage is narrowing

The pandemic-era experiment in location-agnostic compensation is consolidating into more structured frameworks. Robert Half data for Q3 2025 shows 64% of tech workers now fully on-site, 24% hybrid, and only 12% fully remote, a dramatic shift from 2021-2022 levels.

Companies have settled into three compensation models. Location-agnostic employers like GitLab, Automattic, and Zapier pay San Francisco rates regardless of location, but these represent a minority. Most companies, including Google, Microsoft, and Meta, now apply geographic adjustments with Tier 2 cities (Denver, Austin, Boston) paying 5-15% below Bay Area rates and Tier 3 markets paying 15-25% below.

Return-to-office mandates are accelerating. Amazon and Dell implemented 5-day office requirements in early 2025; Ford, Toyota, and 3M moved to 4-day mandates. Over two-thirds of companies globally have formal RTO policies, though fewer than 5% require full five-day presence. The employee response has been significant: surveys indicate 41% of workers would job hunt in response to strict RTO mandates, and companies with rigid policies report twice the turnover of flexible competitors.

Smaller companies are exploiting this dynamic. Seventy percent of companies under 500 employees still offer full remote options, using flexibility as a competitive advantage against larger employers who can offer higher base compensation but less autonomy.

Rational paradoxes govern career decisions

The compensation landscape produces behaviors that appear paradoxical but are entirely rational under analysis. The most striking is junior engineers rejecting "promotions" to other companies. A Google L3 earning $186,000-$192,000 would take a pay cut to accept a "Senior Engineer" offer paying $130,000 at a Tier 2 company. The economically rational choice is to decline the title upgrade.

Conversely, experienced engineers accept apparent demotions for higher compensation. Blind discussions document patterns like "Downleveled to join AWS, added $100K/year to comp, couldn't care less if my title no longer says 'Sr' in front of it." One engineer reported dropping from L7 to L6 while moving from a $900,000 to $600,000 package—still a substantial income, but reflecting Big Tech's strict leveling standards.

Title inflation has accelerated these dynamics. Research from Datapeople found 25% of junior jobs now carry senior titles and twice as many junior positions include "Lead" in the title. A Pearl Meyer survey found 54% of organizations use title inflation explicitly to attract talent. Startups, unable to compete on cash, hand out VP and Director titles to engineers with a few years of experience. Meanwhile, Google under-levels almost everyone: Directors from other companies typically enter as L5, and external Staff+ hires are rare regardless of prior experience.

The result is systematic title arbitrage where "Senior Software Engineer" at Netflix means $350,000+ while the same title at a local company means $90,000. Companies benchmark against their actual competitive sets, not a universal standard. Roblox benchmarks against Dropbox, DoorDash, Snap, and Twilio, not other gaming companies, because those are the employers who might poach their engineers.

Conclusions: toward realistic compensation frameworks

The evidence supports several counterintuitive but empirically grounded conclusions for both employers and engineers navigating this market.

First, compensation bands should be tied to peer groups, not industry averages. A company benchmarking against "the market" is benchmarking against a statistical fiction. The relevant comparison set is the specific employers competing for the same talent pool—which varies dramatically by geography, specialization, and company ambition level.

Second, titles are signals, not standards. The same title carries 2-4x compensation differences across employers. Career decisions based on title progression rather than total compensation or learning opportunity often result in worse outcomes. An L4 at Google with a clear path to L5 may have better prospects than a "Staff Engineer" at a company with no further growth potential.

Third, AI specialization commands the strongest premiums in 2024-2025, but the broader market shows salary compression for generalist roles. Engineers should calibrate expectations: the era of universal 15% annual raises has ended for most positions, while specialists in AI/ML, security, and cloud infrastructure continue seeing strong demand.

Fourth, geographic arbitrage is real but narrowing. Location-agnostic pay exists primarily at companies with distributed-first cultures. Engineers accepting remote roles at companies with strong office cultures should expect eventual RTO pressure and potential compensation adjustments.

The fundamental insight is that software engineering compensation reflects value extraction potential, not labor market equilibrium. Engineers are priced like real estate, same apparent characteristics, vastly different values based on context. Recognizing this reality is the prerequisite for making rational career and compensation decisions.