Physicist
| Median wage (USD) | 172250 |
|---|---|
| Projected employment growth (%) | 7 |
| AI exposure (%) | 64 |
| Maximum exposure (%) | 77 |
| Source | eloundou |
| What AI changes in this occupation | Analysing data from research conducted to detect and measure physical phenomena (96 out of 100), reporting experimental results by writing papers or presenting them at conferences (94), performing the complex calculations involved in analysing and evaluating data on a computer (92), describing and expressing observations and conclusions in mathematical terms (91) and writing research proposals (82) are five of the tasks the O*NET profile ranks as most important — and they are exactly the ones that get delegated. Work activities confirm the diagnosis: analysing data or information comes first (98 out of 100), working with computers right behind (97), processing information (92), updating and using relevant knowledge (92), documenting and recording information (77), and 90% of respondents use e-mail every day. Designing computer simulations to model physical data (74) and performing peer reviews of scientific papers (64) follow the same path: these are processing, writing and calculation tasks, not bench work. BLS puts the median wage for physicists at $172,250 in May 2025 — $128,820 for astronomers, $166,880 for the physicists and astronomers group, against $50,980 for all workers — and projects +7% between 2025 and 2035 for the group (25,600 jobs to 27,400, or +1,900), growth it calls "much faster than average", the average being 3%. For code 19-2012 alone the projection is the same: 23,200 physicists in 2025, 24,800 in 2035 (+1,700), and O*NET shows 1,300 cumulative openings over the decade where BLS announces about 1,500 openings a year on average for the group, mostly replacing people who transfer to other occupations or exit the labour force. The nuance that matters: the market is narrow in absolute terms — 23,200 posts, 1,700 net new jobs over ten years — and entry is locked by the degree. BLS puts the typical entry-level education at a doctoral or professional degree, with no related work experience and no on-the-job training required; only entry-level physicist jobs in the federal government require a bachelor's degree, and O*NET places the profile in Job Zone Five, "extensive preparation", defined as most of these occupations requiring graduate school and often a Ph.D. Growth is therefore real but selective: it is played out on specialisation — condensed matter and materials, optics and photonics, quantum information, medical physics, instrumentation, energy — not on volume. The bench and the instrument remain the core of the job: 90% of O*NET respondents work indoors in an environmentally controlled setting every day, but only 50% rate the job "slightly automated" and 40% "not at all automated"; 63% have a lot of freedom in determining their tasks, priorities and goals and 60% rate being exact or highly accurate as extremely important. Observing the structure and properties of matter and the transformation and propagation of energy with masers, lasers and telescopes (82 out of 100), collaborating with other scientists in the design, development and testing of equipment and instrumentation (83), developing theories and laws from observation and experiment (72), teaching physics (72) and conducting application evaluations on electro-optical devices (54) cannot be delegated — that is the share the Eloundou study leaves untouched, up to 77% at most. Add the working pattern: 70% of respondents work more than 40 hours a week and 47% hold face-to-face discussions every day. |
| Key skills | Quantum mechanics, relativity, statistical physics, electromagnetism, condensed matter, optics and photonics, nuclear and particle physics · mathematical methods: analysis, linear algebra, differential equations, numerical methods (finite differences, finite elements, Monte Carlo) · scientific computing and programming (Python, C++, MATLAB, Fortran, high-performance computing) · data processing and analysis, statistics, machine learning · experiment design and execution, instrumentation, lasers, spectroscopy, cryogenics, vacuum, detectors · modelling and simulation of physical systems · electronics, data acquisition and signal processing · scientific writing, grant applications, conference communication · laboratory safety, radiation protection and regulatory compliance · generative AI tools for literature review, data analysis, simulation code and drafting |
| Courses to get ahead | Numerical Methods for Engineers · Coursera (The Hong Kong University of Science and Technology) · 6 weeks at 4 hours a week · 90 · The course that attacks the profile's most delegable technical task — performing the complex calculations involved in analysing and evaluating data on a computer (92 out of 100) — and directly serves the first one, analysing research data (96). Interpolation, integration, differential equations, linear systems, model fitting: this is the toolbox that lets you set up a scientific computation, run it and judge whether the numerical result holds, at the very moment when code generation is being delegated. Writing in the Sciences · Coursera (Stanford Online) · 8 weeks at 3-5 hours a week · 91 · Reporting experimental results (94 out of 100) and writing research proposals (82) are, along with data analysis, the profile's most important and most directly delegable tasks: a first draft of a paper or a proposal can be generated, the ability to say what the data prove cannot. Stanford's course works on the scientific sentence, the structure of a paper and the art of cutting — what remains when AI produces version one, and what separates a defensible result from a plausible text. AI For Everyone · Coursera (DeepLearning.AI) · 4 weeks at 2-3 hours a week · 85 · For deciding what is delegated and what is not: the course explains what generative AI can produce — data analysis, literature review, simulation code, a first draft of a paper —, where it gets things wrong, and how to frame a use that does not lose the measurement trail. A physicist whose occupation is 64% exposed and 77% at most needs to know exactly what goes in their place. Particle Physics: an Introduction · Coursera (University of Geneva) · 11 weeks at 3-4 hours a week · 86 Introduction to Mathematical Thinking · Coursera (Stanford Online) · 4 weeks at 10 hours a week · 84 From the Big Bang to Dark Energy · Coursera (The University of Tokyo) · 3 weeks at 10 hours a week · 82 |
| Occupations to pivot toward | Astronomer · 128820 · 8 · 42 Materials Scientist · 117790 · 7 · 52 Nuclear Engineer · 133970 · 0 · 48 |
| Skills to acquire for the pivot | Eloundou measures the physicist's exposure at 64% (human β) and up to 77% (human ζ): analysing research data (96 out of 100), reporting experimental results (94), performing complex calculations on a computer (92), expressing observations and conclusions in mathematical terms (91) and writing research proposals (82) are the O*NET profile's most important tasks and its most delegable ones, while collaborating on the design and testing of equipment and instrumentation (83), observing the structure and properties of matter with lasers and telescopes (82), developing theories and laws from observation and experiment (72), teaching physics (72) and conducting application evaluations on electro-optical devices (54) remain human. Within the repository, 64 puts the physicist 11th of the ninety-nine occupations: ten occupations have a higher exposure — from the translator (84) to the mortgage broker (65) —, eighty-eight a lower one, and none is tied with it; its nearest neighbours above are the mortgage broker (65), the financial adviser (67), the tax examiner, the programmer and the web designer (68) and the mathematician (69), below the journalist (63), the tester and the compliance officer (61), then the data engineer, the database administrator and the proofreader (60). Its ζ of 77 places it 44th: thirty-nine occupations have a higher ζ and four are tied with it (the administrative assistant, the secretary, the compliance officer and the researcher), the closest above being the general office clerk (79) and the lawyer (80), below the HR manager (76), the project manager and the bill collector (75). The SOC code used is 19-2012.00 "Physicists": it is confirmed by the occupation's O*NET profile — exact code and title, updated 2026, Bright Outlook flag — and by the Eloundou dataset, which carries a line titled exactly "Physicists" under that code: two sources that agree, nothing assumed. O*NET files under this profile job titles as different as theoretical physicist, biophysicist, health physicist, medical physicist or research physicist: the profile covers a family, not a single post. The GPT-4 self-rated series (β 75, ζ 95) is kept in expositionGpt4 and is not displayed, in line with the decision of 21 August 2026. BLS puts the median wage for physicists at $172,250 in May 2025, and for astronomers at $128,820, on the same page; the physicists and astronomers group stands at $166,880, against $50,980 for all workers. The projection is +7% between 2025 and 2035 for the group (25,600 jobs to 27,400, +1,900), "much faster than average", the average being 3%; for code 19-2012 alone, BLS gives 23,200 physicists in 2025 and 24,800 in 2035 (+1,700), and O*NET shows the same growth with 1,300 cumulative openings over the decade, where BLS announces about 1,500 openings a year on average for the group — two different measures, one cumulative over ten years for a single code, the other annual for a group. A degree gates entry: BLS puts the typical entry-level education at a doctoral or professional degree, with only entry-level federal physicist jobs making an exception with a bachelor's degree, and O*NET places the profile in Job Zone Five, "extensive preparation". The work is mainly office and laboratory work: 90% of respondents work indoors in an environmentally controlled setting every day and 90% use e-mail daily, but 70% work more than 40 hours a week, 47% hold daily face-to-face discussions, 63% have a lot of freedom in determining their tasks and priorities and 60% rate accuracy as extremely important; 50% of respondents rate the job "slightly automated" and 40% "not at all automated". This card's three pivots are neighbouring occupations documented by BLS: the astronomer shares the physicist's page ($128,820, +8% between 2025 and 2035, human β 42), the materials scientist appears in the similar-occupations table under "chemists and materials scientists" ($117,790, +7%, human β 52) and the nuclear engineer appears there too ($133,970, no change — "little or no change" from 2025 to 2035 —, human β 48). The median salary shown is $172,250 a year, or $82.81 an hour according to the O*NET profile that reports the 2025 BLS wage data. It is the second-highest salary in the repository of ninety-nine occupations: only the physician ($239,200) pays more, and the physicist is ahead of the dentist ($166,300), the lawyer ($151,160), the HR manager ($149,280) and the data engineer ($139,500), while the repository's median salary is $72,910 — the physicist sits at more than twice that. The degree explains the gap: O*NET places the profile in Job Zone Five and BLS puts the typical entry-level education at a doctoral or professional degree, which narrows the pool and supports pay even though headcount grows slowly — 1,700 net new jobs over the decade for 23,200 posts. Progression runs through specialisation (condensed matter and materials, optics and photonics, quantum information, medical physics, scientific instrumentation, energy) and through the ability to carry a research programme, an instrument or a funding line that AI-assisted drafting does not replace. Every value on this card comes from a named source; only the task shares, derived from O*NET importance scores, and the course scores are editorial estimates, documented in _ponderation. |
| GPT-4 exposure — beta / zeta (%) | 75 · 95 |