On August 27, the U.S. Bureau of Labor Statistics released employment forecasts for the period from 2025 to 2035. Total employment is expected to increase from 170.3 million to 176.2 million, an increase of 5.9 million, representing a growth rate of 3.5% over ten years, which is significantly slower than the 10.9% growth rate from 2015 to 2025. This set of forecasts is based on population, labor participation, macroeconomic factors, and historical technological trends, and it is not an accurate prediction of employment for each individual year in the future, nor does it attempt to predict cycles of recession or prosperity in advance.
Growth will be highly concentrated in healthcare and social assistance. The private healthcare and social assistance sectors are expected to grow by 9.5%, creating more than 2.2 million new jobs, which accounts for about 37% of all new jobs created. Aging populations and the increase in chronic diseases such as heart disease, cancer, and diabetes are the main sources of demand. Two major occupational categories, healthcare support and medical professional technology, are expected to grow by 13.3% and 8.0% respectively, contributing nearly one-third of the new employment.
AI is both a source of new demands and a source of uncertainty in the report. BLS introduces for the first time a classification of occupational AI exposures, comparing occupations based on theory and observed levels of exposure. The report predicts that demands for AI systems, research and development, and consulting will drive growth in professional, scientific, and technical services by 8.6%, creating 926,700 new jobs; computing infrastructure, data processing, network hosting, and related services are expected to grow by 25.1%, resulting in 120,400 new jobs.
The fastest growth does not equate to the largest increase; the industry scale determines the actual number of positions.
Public utilities are expected to grow by 9.8%, which is the fastest among the major industries, but due to the smaller base, only 58,800 new jobs will be created. The increasing demand for electricity, including the power consumption of AI data centers, is an important driving factor. Solar, wind, and geothermal power generation are among the fastest-growing sub-industries, but the related power generation industries are expected to create only about 35,800 new jobs in total. A high percentage does not necessarily mean that they can absorb the largest number of workers.
Similarly, it is expected that the number of solar photovoltaic installers and wind turbine service technicians will increase by 36.5% and 29.5% respectively, yet the total number of new jobs created is less than 15,000. Nurses are expected to see a 41% increase, making it one of the professions with the fastest growth rates; data scientists are expected to grow by 34.6%, and computer and information research scientists by 21.8%. When evaluating investments in education and training, it is important to consider not only the growth rates but also the base number of employed individuals, annual job vacancies, and regional distribution.
The absolute scale of healthcare is even larger. It is expected that there will be an additional 625,400 jobs created in services for the elderly and people with disabilities, making this the sector with the largest increase in new jobs. The growing demand for home care means that employment growth may be concentrated in jobs that require face-to-face interaction and are difficult to fully automate. However, wages, occupational safety, and career advancement opportunities remain key factors. The presence of many jobs does not necessarily mean that those jobs are of high quality.
The report also predicts a 3.4% decline in federal government employment and a 0.2% decline in retail trade, with the latter resulting in a reduction of about 27,500 jobs. E-commerce continues to limit the need for employees in physical retail, while increasing demand for transportation and warehousing; this industry is expected to grow by 3.1%. This represents a shift in job structure, rather than just a change in overall numbers: consumers are shifting from stores to online platforms, and jobs are moving from cashier and sales positions to warehousing, delivery, software, and logistics coordination roles.
AI affects processing at historical speeds, therefore the results may underestimate extreme variations.
BLS clearly states that predictions assume that technological progress and labor productivity will generally continue along historical trends. If AI results in technological progress far exceeding historical rates, current methods may not be able to produce reasonable outcomes. Higher overall productivity under the assumption of "full employment by 2035" would increase GDP, but if productivity improvements vary significantly across industries, the occupational employment structure could change markedly. However, BLS currently lacks sufficient data to establish a reliable baseline for such differences.
This means that the AI exposure category cannot be directly interpreted as a "probability of substitution." A profession may have high exposure to AI, yet employment may increase due to growing demand; it is also possible that with the use of AI, individual output rises, but the overall growth in jobs slows down. Complementarity and substitution at the task level, new demands brought about by declining prices, and the speed at which regulations are adopted by organizations will all affect the final outcome. Enterprises and workers should not regard decade-long percentages as a foregone conclusion of fate.
The education system is more suited for using predictions in making directional choices. There are signs of long-term demand in healthcare, data science, energy, power infrastructure, and technical services, but the specific tools used will vary. Courses should focus on enhancing transferable skills: a foundation in mathematics and data, communication, domain knowledge, equipment maintenance, compliance judgment, and collaboration with AI, rather than just training for a particular software. For professions that may experience decline, short-term certifications and on-the-job retraining pathways are also necessary.
Regional planning cannot simply be copied from the national average. Data centers may be concentrated in states with suitable power and land conditions, while medical needs are related to population age and urban-rural distribution; areas with new job creations may not coincide with those where retail or government jobs are decreasing. States and cities need to recalculate based on local employment bases, wages, housing, and training capacity, rather than evenly distributing the national forecast of 3.5%. The number of jobs, their quality, and accessibility must all be evaluated together.
5.9 million new jobs represent a decade of slow growth and significant structural changes. The healthcare sector contributes the most; AI drives demand for jobs in areas such as computing power, electricity, research and development, and data processing, and may also alter the task combinations of many existing jobs. The most accurate approach is to treat these forecasts as planning scenarios based on a set of clear assumptions, and to continuously refine them using actual adoption rates, productivity data, and job-related information, rather than treating the figures for 2035 as a fixed future reality.
