AI Industry Salary Report: September 2026 Data
September 2026 AI salary bands for employers: Levels.fyi puts Research Scientist median TC at $305K and ML Engineer at $280K; Robert Half AI/ML base mid is $170,750 (+4.1% YoY). Budget framework across AI, ML, LLM, research, MLOps, and product seats.
U.S. Machine Learning Engineers report a $280,000 median total compensation on Levels.fyi as of 6 September 2026. Research Scientists clear $305,000. Robert Half's 2026 national AI/ML Engineer base midpoint sits at $170,750, with that specialty projected to rise +4.1% year over year against +1.6% for technology roles overall.
This Best Firms data report maps AI industry salaries for September 2026 across roles employers actually open: AI Engineer, ML Engineer, LLM-adjacent seats, research, applied science, data, MLOps-style ops, and AI product. You will get cash vs total-comp bands, regional signals where they exist, demand pressure, and a hiring budget framework. For the three-title gap story alone, see the companion software vs AI vs ML engineer salary gap 2026 report.
Key Takeaways
- Specialization still prices above the "AI" label. Levels.fyi (U.S., 6 Sep 2026): Research Scientist median total comp $305,000; Machine Learning Engineer $280,000; ML/AI SWE focus $248,000; Software Engineer $195,000; standalone AI Engineer title $153,750.
- National base cash is cleaner for most budgets. Robert Half 2026: AI/ML Engineer $134,000–$193,250 (mid $170,750); AI Architect mid $175,000; Data Scientist mid $153,750; Software Engineer mid $142,000. AI/ML and data science bases projected +4.1% YoY vs +2.3% for software development.
- Developer-survey cash confirms the premium. Stack Overflow Developer Survey 2025 (U.S.): AI/ML engineer median $189,500; applied scientist $157,500; data engineer $150,000; data scientist $145,000; full-stack $138,000.
- Demand is concentrated, not universal. LinkedIn ranks AI Engineer #1 on U.S. Jobs on the Rise again in 2026. Robert Half counted 49,200 U.S. AI/ML/data-science postings in 2025 (+163% YoY). Ravio 2026: AI/ML hiring +88% YoY with a 12% Professional/IC salary premium.
- Budget the seat by first 90 days of work. Pay research/ML rates only when modeling systems, eval ownership, or training loops are required. Price LLM product integration closer to a strong AI engineer or backend band, not a frontier-lab Staff package.
What "AI salary" means in September 2026
Here's the problem.
"AI salary" is not one number. Levels.fyi total compensation mixes base, annualized equity, and bonus at companies that grant equity. Robert Half publishes national starting base for someone new to the role. Stack Overflow is self-reported total annual compensation among survey respondents. BLS OEWS (May 2025) puts software developers (SOC 15-1252) at a national median wage of about $135,990, an economy-wide floor that includes non-tech employers.
Read them as stacked benchmarks. Do not average them into a fake "market rate."
Best Firms builds this report for employers setting bands, not for candidates shopping frontier-lab rumors. Company-level outliers still live in the highest paying tech companies in 2026 list and in startup-cash context from the 2026 startup salaries report.
Role salary bands (U.S.)
Use Levels.fyi for equity-heavy total-comp ceilings. Use Robert Half for cash floors. Use Stack Overflow as a developer-survey cross-check.
Levels.fyi median total compensation (U.S., updated 6 September 2026)
| Role | 25th | Median TC | 75th | 90th |
|---|---|---|---|---|
| Research Scientist | $179,000 | $305,000 | $410,000 | $540,000 |
| Machine Learning Engineer | $200,000 | $280,000 | $385,000 | $500,000 |
| ML / AI Software Engineer (focus) | $167,500 | $248,000 | $367,500 | $512,000 |
| Product Manager | $168,000 | $230,000 | $330,000 | $455,000 |
| Software Engineer | $137,500 | $195,000 | $280,000 | $388,000 |
| Data Scientist | $134,200 | $180,000 | $250,000 | $350,000 |
| AI Engineer (title) | $110,000 | $153,750 | $215,000 | $300,000 |
Against the Software Engineer median, Machine Learning Engineer sits +$85,000 (~44%). Research Scientist sits +$110,000 (~56%). The diluted AI Engineer title median sits below SWE, which is title inflation, not a bargain aisle.
Robert Half 2026 national starting base (low / mid / high)
| Role | Low | Mid | High |
|---|---|---|---|
| AI Architect | $142,750 | $175,000 | $196,750 |
| AI/ML Engineer | $134,000 | $170,750 | $193,250 |
| Data Engineer | $127,000 | $156,250 | $180,750 |
| Data Scientist | $121,750 | $153,750 | $182,500 |
| AI/ML Analyst | $119,250 | $145,750 | $174,000 |
| Software Engineer | $109,250 | $142,000 | $175,500 |
At the midpoint, AI/ML Engineer clears Software Engineer by about $28,750 (~20%). That lines up directionally with Ravio's 12% Professional/IC AI premium (European-tech-weighted sample: treat as a premium signal, not a U.S. cash band).
Robert Half also flags emerging 2026 titles without full national tables yet: LLM engineer, Agentic AI engineer/developer, AIOps engineer, and AI strategy consultant. Until those pages publish low/mid/high, price LLM product work against AI/ML Engineer and strong backend bands, and price AIOps against DevOps plus an AI/ML premium.
Stack Overflow 2025 U.S. median total annual compensation
| Developer type (U.S.) | Median |
|---|---|
| AI/ML engineer | $189,500 |
| Cloud infrastructure engineer | $189,000 |
| Architect, software or solutions | $180,000 |
| Developer, back-end | $175,000 |
| Applied scientist | $157,500 |
| Data engineer | $150,000 |
| Data scientist | $145,000 |
| Developer, full-stack | $138,000 |
Against full-stack, AI/ML is +$51,500 (~37%). Against back-end, the gap shrinks to +$14,500 (~8%). Specialty beats buzzword.
Levels and seniority: how bands usually widen
Most employers do not hire "an AI person." They hire a level.
On Levels.fyi-style total comp, the spread from 25th to 90th is the real planning range. Machine Learning Engineer runs $200,000 to $500,000. Research Scientist runs $179,000 to $540,000. Software Engineer runs $137,500 to $388,000.
That means a mid-level production ML hire and a Staff research hire are not the same budget line, even if both show up in an ATS under "AI."
Best Firms' take: publish one cash band per seat family, then a separate equity philosophy. Matching a $280,000 Levels.fyi median as cash when you cannot grant equity is how Series A teams blow runway.
For company outliers (Anthropic, OpenAI, quant shops), use the highest paying tech companies report. Those packages are fundraising theater for most employers, not a midpoint.
Regional signals (where public data exists)
Levels.fyi location medians for Software Engineer (U.S., 6 Sep 2026):
- San Francisco Bay Area: $295,000
- Greater Seattle Area: $256,000
- New York City Area: $200,000
- Greater San Diego Area: $190,000
- Greater Austin Area: $187,200
Data Scientist location medians on the same date stamp: San Francisco Bay Area $272,239; New York City Area $170,000.
Robert Half's Salary Guide defaults to national averages and lets you select a city in its calculator. National midpoints remain the cleanest published cash floor when your remote policy pays one U.S. band.
If you hire remote, decide whether you anchor to SF/Seattle TC, a national Robert Half mid, or a hybrid. Publishing that choice in the req reduces offer-letter churn.
Demand signals behind the premium
Compensation follows scarcity and posting volume.
LinkedIn's 2026 labor-market release ranks AI Engineer #1 on U.S. Jobs on the Rise for a second year and cites more than 1.3 million new AI-enabled jobs globally over two years, with U.S. jobs requiring AI literacy skills up about 70% year over year in that research window.
Robert Half counted nearly 1.1 million U.S. technology and IT postings in 2025. AI/ML/data-science roles hit 49,200, up 163% from 2024. In the same Salary Guide cycle, 87% of technology and IT leaders say they typically offer higher salaries for specialized skills, with AI, machine learning, and data science at the top of that list.
Ravio's 2026 Compensation Trends analysis shows AI/ML hiring +88% year over year, unique AI/ML job titles +50%, AI/ML Engineer at 45% of AI/ML titles (Senior AI/ML Engineer another 15%), and a 12% salary premium at Professional/IC levels vs 3% at Management.
Python remains the default language across many of these seats, which is why teams still source from the pool in Best Firms' hire Python developers guide, then filter for evals, data contracts, or LLM systems experience. For classical modeling seats, start with hire machine learning engineers. For applied LLM product seats, start with hire AI engineers.
Need candidates, not just bands? Open the Best Firms hire hub or browse live AI roles.
Hiring budget framework for employers
Do not open three overlapping AI reqs. Pick the mechanism the first 90 days require.
| Seat family | First 90-day job | Cash budget signal (U.S.) | TC ceiling signal | When you overpay |
|---|---|---|---|---|
| Software / platform | Product features, services, reliability | Robert Half SWE mid $142K | Levels.fyi SWE median $195K | Paying ML equity for CRUD plus a vendor LLM call |
| AI / LLM applied | RAG, eval harness, agents, cost caps | Often $150K–$190K base outside frontier labs | AI Engineer title median $153,750 is diluted | Staff-research TC for prompt-and-glue work |
| ML Engineer | Dataset → baseline → eval → deployed inference | Robert Half AI/ML mid $170,750 | Levels.fyi MLE median $280K | Notebook-only modeler for production ranking |
| Research / Applied science | Novel methods, papers-to-product | Stack Overflow applied scientist $157,500 | Levels.fyi Research Scientist $305K | Opening research when you only need product LLM features |
| Data / MLOps-adjacent | Pipelines, feature stores, serving, monitoring | DE mid $156,250; DS mid $153,750 | Levels.fyi Data Scientist $180K | Paying research TC for pipeline ownership alone |
| AI product | Roadmap, eval product sense, vendor tradeoffs | Price near senior PM cash | Levels.fyi PM median TC $230K | Paying Staff MLE TC for a PM who never owns evals |
Load the offer before you celebrate the base. Employer payroll tax, benefits, equipment, and tooling commonly add 25–40% on top of cash for a W-2 hire. A $170,750 AI/ML midpoint can become about $213K–$239K fully loaded before recruiting fees or equity. That is the number finance should see.
Startup cash still sits closer to Wellfound/Robert Half than to equity-heavy Levels.fyi medians. See Best Firms' 2026 startup salaries report for that context.
How to set bands without breaking payroll
- Write the first project into the requisition. Ranking model in production means ML. Customer-facing LLM feature with evals means AI. Platform service with an optional LLM call means Software.
- Split cash and equity. Match Robert Half midpoints for cash when you cannot print Levels.fyi equity.
- One ATS scorecard per seat. If "AI Engineer" and "ML Engineer" are synonyms in your ATS, your pipeline will be too.
- Cap exceptions. Ravio's reward advisors warn that spike premiums baked into base are hard to reverse. Prefer time-limited market adjustments, sign-on, or equity for non-differentiating seats.
- Name the downside. Paying ML TC for API-wrapper work burns runway. Underpaying a real ML systems hire burns six months of roadmap.
Named downside: who these bands do not help
This report does not argue that every company should hire a Research Scientist at the $305,000 Levels.fyi median.
If you are a seed-stage B2B SaaS team shipping one LLM-backed feature, a strong software engineer plus a clear eval budget often beats a classical research hire. If you need custom training loops, leakage-sensitive ranking, or on-device models, the opposite is true, and cheap "AI" titles will fail your on-call.
Frontier-lab packages are also the wrong benchmark for most employers. Competing with Anthropic or OpenAI senior medians is a fundraising strategy, not a hiring strategy.
Best Firms' take
As of 6 September 2026, the AI industry compensation stack is clear when you separate titles. Research and ML systems sit at the top of Levels.fyi medians. Robert Half still shows AI/ML base about 20% above software engineering, with +4.1% projected YoY gains in that specialty. Stack Overflow's U.S. AI/ML median clears full-stack by more than $50,000.
The trap is treating every "AI" req as a $280,000 problem. Open one precise seat. Price cash against Robert Half. Use Levels.fyi for equity-heavy ceilings. Then source.
FAQ
What is the median AI engineer salary in the United States in 2026?
Levels.fyi's U.S. AI Engineer title median total compensation is $153,750 as of 6 September 2026 (25th $110,000, 75th $215,000). That label is diluted. Machine Learning Engineer median TC is $280,000, and Robert Half's AI/ML Engineer national base midpoint is $170,750.
How much should employers budget for an ML engineer in 2026?
Budget Robert Half's AI/ML Engineer national starting base of $134,000–$193,250 (mid $170,750) for cash, then add benefits load (often 25–40%) and equity separately. Levels.fyi's Machine Learning Engineer median total comp is $280,000, which is an equity-heavy ceiling, not a cash midpoint for most startups.
Why are AI and ML salaries rising faster than general tech pay?
Robert Half projects +4.1% YoY for AI/ML engineer and data scientist starting salaries vs +1.6% across technology and IT overall and +2.3% for software and applications development. LinkedIn ranks AI Engineer the top U.S. growth title again in 2026, and Robert Half measured 163% YoY growth in AI/ML/data-science postings to 49,200 in 2025.
What salary premium do AI skills carry versus non-AI peers?
Ravio's 2026 Compensation Trends report measures about a 12% salary premium for AI talent at Professional/IC levels and about 3% at Management, inside Engineering/IT/Data families. That is a European-tech-weighted premium signal. On U.S. national base, Robert Half's AI/ML mid ($170,750) sits roughly 20% above Software Engineer mid ($142,000).
How do AI salaries differ by U.S. region?
Levels.fyi Software Engineer medians (6 Sep 2026) show San Francisco Bay Area at $295,000, Seattle at $256,000, New York City at $200,000, San Diego at $190,000, and Austin at $187,200. Data Scientist SF median is $272,239 vs NYC $170,000. Robert Half publishes national midpoints and city adjustments in its calculator.
When should a company hire an LLM engineer versus an ML engineer?
Hire LLM/applied AI when the first 90 days center on RAG, evals, agents, and production LLM product work. Hire ML when you need dataset-to-model-to-deployed-inference ownership. Robert Half lists LLM engineer as an emerging 2026 title without a full national low/mid/high table yet, so price it against AI/ML Engineer cash ($170,750 mid) and a strong backend bar unless the seat owns training systems.
What is the difference between Levels.fyi and Robert Half AI salary numbers?
Levels.fyi reports verified total compensation (base + annualized stock + bonus) and skews toward equity-granting employers. Robert Half reports national starting base salary levels for someone new to the role. Use Robert Half for cash floors and Levels.fyi for total-comp ceilings.
If you are opening a seat this month, define the first project, pick the title that matches that project, and set cash against Robert Half midpoints before you argue about Levels.fyi equity. Browse the hire hub or live AI jobs when you need market signal.