Data science spans a defined U.S. occupation, adjacent computing roles, business technology adoption, and the rapidly changing economics of AI. The statistics below distinguish observed measures, survey estimates, forecasts, and model-cost estimates by period and population.
Key Data Science Statistics
The most broadly useful figures include:
- In 2024, 245,900 people were employed as data scientists in the United States under the BLS occupation definition.
- BLS projects 328,300 U.S. data-scientist jobs in 2034, a forecast rather than a guaranteed outcome.
- U.S. data-scientist employment is forecast to grow 34% from 2024 to 2034.
- About 23,400 data-scientist openings per year are projected in the United States during 2024–2034.
- The May 2024 U.S. median annual wage was $112,590 for data scientists.
- The highest-paid 10% of U.S. data scientists earned more than $194,410 annually in May 2024.
- Computer and mathematical occupations are projected to grow 10.1% in the United States from 2024 to 2034.
- California had 33,220 data scientists in May 2023, the largest state employment figure listed here.
- The District of Columbia had a 3.48 data-scientist location quotient in May 2023, measuring concentration against the national average.
- 3.8% of U.S. nonfarm employer businesses reported using AI to produce goods or services in the September–November 2023 Census snapshot.
- 13.8% of information-sector businesses reported current AI use in that Census snapshot.
- BTOS AI-use estimates rose from 3.7% in fall 2023 to 5.4% in February 2024 among U.S. businesses.
- The Census ABS found 68.0% of AI-using U.S. employer businesses adopted AI during 2021–2022.
- 40.0% of U.S. employer businesses adopted cloud-based technology during 2016–2020 in the ABS measure.
- 59.0% of U.S. employer businesses rated cloud technology very important to their processes or methods in the 2020–2022 survey reference.
- The estimated cost to train the original Transformer model was about $670 in 2017, according to an inflation-adjusted Epoch AI estimate.
- GPT-4’s estimated training cost was about $79 million in 2023, based on an inflation-adjusted Epoch AI estimate.
- The cost of querying a GPT-3.5-equivalent model fell from $20 to $0.07 per million tokens between November 2022 and October 2024.
- The smallest model above 60% on MMLU fell from 540 billion parameters in 2022 to 3.8 billion in 2024.
Contents
- Data science workforce size, pay, and job-growth statistics
- Where data scientists work: industries, states, and concentration
- Data science and related computing occupations
- Business data science and AI adoption
- Technology adoption, business impact, and motivations
- AI model scale, cost, and responsible-data trends
Data Science Workforce Size, Pay, and Job-Growth Statistics
The Bureau of Labor Statistics counted 245,900 U.S. data scientists in 2024 and projects 328,300 in 2034. The occupation-level forecast implies a 34% increase, while the projected annual average is about 23,400 openings, including labor-force exits and occupational transfers.
Source: Data Scientists: Occupational Outlook Handbook (BLS).
| Measure | Period | U.S. result |
|---|---|---|
| Employment | 2024 | 245,900 workers |
| Projected employment | 2034 | 328,300 jobs |
| Projected change | 2024–2034 | 34% |
| Projected annual openings | 2024–2034 | 23,400 per year |
Pay varies widely within the occupation. In May 2024, the median annual wage was $112,590 and the median hourly wage was $54.13; the lowest-paid 10% earned under $63,650, while the highest-paid 10% earned over $194,410.
The data-scientist median was $63,090 above the $49,500 median for all occupations in May 2024. That is a difference between two BLS medians, not an estimate of an individual earnings premium or total compensation.
Where Data Scientists Work: Industries, States, and Concentration
BLS industry data show different median wages for data scientists in May 2024. These are industry-specific medians, not a ranking of every employer or a measure of total compensation.
| Industry | May 2024 median annual wage |
|---|---|
| Computer systems design and related services | $128,020 |
| Management of companies and enterprises | $126,940 |
| Scientific research and development services | $120,090 |
| Management, scientific, and technical consulting services | $110,240 |
| Insurance carriers and related activities | $108,920 |
Source: Data Scientists: Occupational Outlook Handbook (BLS).
The May 2023 BLS OEWS estimates show concentration in several large labor markets. California had 33,220 data scientists, Texas 20,560, New York 16,390, Florida 8,400, and Pennsylvania 7,490.
The location quotient tells a different story from raw employment. In May 2023, the District of Columbia recorded a 3.48 location quotient, followed by Utah at 1.66, Colorado at 1.65, and Massachusetts at 1.60; a location quotient compares occupational concentration with the national average rather than showing a share of workers.
Source: Occupational Employment and Wages, May 2023: 15-2051 Data Scientists (BLS OEWS).
Data Science and Related Computing Occupation Statistics
Data science sits within a wider U.S. labor market that includes software, security, research, and operations analysis. These adjacent measures should not be treated as data-scientist employment, but they indicate the broader occupational setting.
| Occupation or group | Forecast period | Projected change or size |
|---|---|---|
| Computer and mathematical occupations | 2024–2034 | 10.1% growth; 545,600 added jobs |
| Software developers | 2024–2034 | 15.8% growth |
| Information security analysts | 2024–2034 | 28.5% growth |
| Computer and information research scientists | 2024–2034 | 19.7% growth |
| Operations research analysts | 2024–2034 | 21.5% growth |
The information sector employed 2.9 million wage-and-salary workers in 2024, with a projection of 3.1 million in 2034. Computer systems design employment is projected to rise from 2,444.8 thousand jobs in 2024 to 2,831.6 thousand in 2034, while computing infrastructure providers, data processing, and web hosting are projected to grow 20.3% over 2024–2034.
Source: Industry and occupational employment projections overview and highlights, 2024–34 (BLS).
Business Data Science and AI Adoption Statistics
Business surveys measure AI use rather than data science alone, so the figures are adjacent indicators. The Census BTOS snapshot found that 3.8% of U.S. nonfarm employer businesses reported using AI to produce goods or services in September–November 2023, compared with 13.8% of information-sector businesses.
BTOS estimates rose from 3.7% in fall 2023 to 5.4% in February 2024. Respondents expected AI use to reach about 6.6% by early fall 2024, but that figure is an expectation, not a later observed outcome.
Source: Tracking Firm Use of AI in Real Time: A Snapshot from the Business Trends and Outlook Survey (U.S. Census Bureau).
The Census ABS used different populations, questions, and reference periods. Among businesses reporting AI use, 68.0% said they had adopted it during 2021–2022, so this is adoption timing among AI-using businesses rather than the percentage of all businesses.
During 2016–2020, ABS reported adoption of cloud-based technology by 40.0% of U.S. employer businesses, specialized software by 27.0%, and specialized equipment by 23.6%.
Source: Census Bureau’s 2023 Annual Business Survey Provides Insight into Technology Adoption by Businesses (U.S. Census Bureau).
Data Science Technology Adoption, Impact, and Motivations
The ABS also asked businesses how important technologies were to their processes or methods, using a 2020–2022 survey reference. Cloud-based technology was rated very important by 59.0% of businesses, specialized software by 58.9%, and specialized equipment by 57.4%.
Reported motivations focused on process quality and reliability. Among adopting businesses, 51.8% cited that motivation for cloud adoption, 49.8% for specialized software, and 45.8% for AI; these are self-reported motivations and do not establish causation.
AI’s reported effect on STEM-worker skills was mixed. In the ABS measure, 40.7% of AI-adopting businesses reported no overall impact, 22.1% reported boosted skills when AI altered those skills, and 3.3% reported a decrease in workers’ skill level.
These percentages are survey categories, not direct measures of data-science productivity, financial returns, or changes in headcount. The denominators and wording follow the Census ABS figure context.
Source: Census Bureau’s 2023 Annual Business Survey Provides Insight into Technology Adoption by Businesses (U.S. Census Bureau).
AI Model Scale, Cost, and Responsible-Data Trends
Estimates from the Stanford AI Index show how the economics surrounding data science have changed. Epoch AI estimated the original Transformer’s 2017 training cost at about $670, RoBERTa Large’s 2019 cost at about $160,000, GPT-3 175B’s 2020 cost at about $4 million, GPT-4’s 2023 cost at about $79 million, and Llama 3.1-405B’s 2024 cost at $170 million.
These are inflation-adjusted estimates based on cloud compute rental prices, not audited company disclosures. They therefore describe estimated training economics rather than reported corporate spending.
Source: Artificial Intelligence Index Report 2025 (Stanford HAI).
Inference became cheaper for a comparable benchmark level. The cost of querying a GPT-3.5-equivalent model fell from $20 to $0.07 per million tokens between November 2022 and October 2024, although the comparison does not represent every workload.
At the same time, benchmark-eligible models became smaller. The smallest model above 60% on MMLU fell from 540 billion parameters in 2022 to 3.8 billion in 2024; this is a benchmark-threshold comparison, not a general intelligence measure.
Source: AI Index 2025: State of AI in 10 Charts (Stanford HAI).
Responsible-data indicators also changed. The share of restricted tokens in actively maintained C4 domains rose from 5–7% to 20–33% between 2023 and 2024, under the study’s corpus and restriction definition.
The average Foundation Model Transparency Index score rose from 37% in October 2023 to 58% in May 2024. Responsible-AI papers accepted at leading AI conferences increased from 992 in 2023 to 1,278 in 2024, but paper counts do not measure research quality or real-world safety.
Source: Responsible AI: The 2025 AI Index Report (Stanford HAI).