AI Research Scientist Salary: What BLS Wage Data Shows by Percentile and State

AI research scientists earn a median of $140,300 a year, according to the May 2025 BLS wage survey. BLS counts them with computer and information research scientists, so that occupation is the source of every figure on this page. The middle half of the group earns $103,570 to $188,700, and your state moves the number more than anything else.

Which Occupation Do These Numbers Describe?

The federal government doesn’t publish a wage line titled “AI research scientist.” The occupation that covers the work is computer and information research scientists, code 15-1221. It includes researchers across computing, not only machine learning, so these figures are a close match for the job and not a count of AI specialists alone.

We use the federal survey because it counts employment and wages by occupation and state, with the same method every year. Job-board estimates and self-reported surveys tend to drift and can’t be compared from one year to the next.

This is a small occupation. BLS counts 37,200 workers, so a state with a few hundred of them can post an unusual average. We flag those below.

If you want the job described first, start with our page on AI research scientist jobs. It covers duties, degrees and how to get hired.

What Is the National Median?

The median is the middle of the line: half of the workers earn more and half earn less. For this occupation it is $140,300 a year, which works out to $67.45 an hour.

The mean, or average, is $153,930, or $74 an hour. The mean sits above the median because a group of high earners pulls it up. When you see a salary quoted as an average, expect it to read higher than what a typical worker takes home.

Hold the median up against the middle of your own plans. If you are weighing a master’s, this is the number to compare against what you earn today.

How Wide Is the Pay Range?

Percentiles show how pay is spread, and they also work as a rough stand-in for experience. BLS doesn’t report pay by years on the job, so we don’t invent experience bands. Lower percentiles mix in newer workers and lower-paying states. Higher ones mix in senior researchers and expensive metros.

The gap between the 10th and 90th percentiles is wide. The top of the range is nearly three times the bottom. That tells you the occupation holds very different jobs, from small employers in low-cost states to well-funded labs in expensive ones.

Do not read the 10th percentile as a starting salary. It is the nearest thing to a floor, but it blends entry-level workers with experienced people who work in lower-paying places.

Which States Pay the Most?

State is the biggest lever. This table shows the highest medians, with the number of workers BLS counts so you can judge how solid each figure is.

Oregon has the highest median and mean, but it has only 730 workers, so a few high salaries can shift its numbers. Washington, New York and Massachusetts rest on far more workers and tell a steadier story.

California holds 8,440 jobs, the most of any state, and its median is $159,100. If you want the largest pool of openings, that is where the survey finds the most people.

Which States Pay the Least?

The low end is just as instructive. Michigan’s median is $82,590 and Indiana’s is $85,460. Georgia’s is $92,250 with 810 workers, Oklahoma’s is $93,100 and Utah’s is $94,700 with 1,250.

Texas sits in the middle at $121,210 with 1,960 workers, and Florida’s median is $119,810 with 1,070.

Pay is not the whole picture, because cost of living differs a lot between Oregon and Indiana. A lower salary in a cheaper state can leave you with similar spending power.

Connecticut has a wage in the survey but no worker count, and Idaho has a worker count but no wage figure. Treat both as gaps in the survey, not as signals.

What Moves Pay Within the Occupation?

BLS gives us three things that move the number, and we’ll stick to what the data supports.

State. The gap between Oregon’s median of $206,220 and Michigan’s of $82,590 is larger than the gap between the national median and the 10th percentile. Where the job sits matters more than almost anything else we can measure.

Where you fall in the range. The 90th percentile is $230,630. The people up there are not typical, and a path to that range depends on the employer and the work, which the survey can’t see.

Occupation group. Your job title decides which BLS line you are counted in. Many people who build AI products are counted with software developers, whose median is lower. Our AI engineer salary page compares the groups.

What the data doesn’t tell us is the effect of a degree on pay, because BLS doesn’t split wages by education. We won’t promise a raise from one.

BLS also reports pay by industry, but we don’t use those tables here, so we can’t say which sectors pay most.

How Wide Is the Range Inside One State?

Percentiles also work inside a single state, and the spread there can be wider than the national one. In California, the 10th percentile is $88,920 and the 25th is $112,170. The 75th is $213,630 and the 90th is $275,320.

New York is higher at every step. Its 10th percentile is $95,260, its 25th is $133,350, its 75th is $280,710 and its 90th is $314,130.

Texas runs lower. Its 10th percentile is $65,480 and its 90th is $204,250, with a median of $121,210.

The lesson is that a state’s median tells you where the middle sits, not what you’ll be offered. Two researchers in the same state can differ by six figures. That is why we suggest reading a few local postings next to these tables, and why a percentile is a better guide than an average.

What Does Hourly Pay Look Like?

If you are moving from hourly work, these figures are easier to compare. The median is $67.45 an hour and the mean is $74. California’s median is $76.49 an hour. Massachusetts is at $81.97, and New York is at $87.50.

Michigan’s median is $39.71 an hour and Indiana’s is $41.09. Full-time work is about 2,080 hours a year, which is how BLS converts the annual figures.

What Does the Education Cost Next to This Pay?

BLS lists a master’s degree as the typical entry education for computer and information research scientists, so the master’s is the cost to price. Our data on online AI master’s programs gives a median total tuition of $32,463.

The spread is large. The lowest total is $9,990 at the University of Texas at Austin, for 30 credits at $333 a credit. The highest is Columbia’s online MS at $81,000 for 30 credits.

Here are programs that matter for a research career, because each includes a thesis or dissertation.

The University of Bridgeport and NJIT make the thesis an option, so confirm that when you apply. For more, see our online AI master’s programs with a thesis option.

How Does the Cost Compare to a Year of Pay?

Put the tuition next to the pay range and the math is easier to see. The lowest master’s total of $9,990 is far less than one year of the median wage of $140,300. The median master’s cost of $32,463 is under a quarter of one year at the median.

Columbia’s $81,000 is just under the 10th percentile wage of $82,200, so it costs about as much as a year of pay at the low end of this occupation. That doesn’t make it a bad choice, but it does raise the bar for the return you need.

Tuition isn’t the whole cost. If you leave a job to study, you lose that income, and some programs add technology fees or required campus visits. A part-time online program lets you keep earning while you study, which is often the biggest saving of all.

We have no data that says a pricier school pays more. BLS doesn’t report pay by school, and we make no such claim. Our cheapest online AI degrees list sorts programs by per-credit price, and the online master’s ranking compares them on more than price.

Is the Demand There to Support the Pay?

BLS projects 22 percent growth in this occupation from 2025 to 2035, which it calls much faster than the average. It expects about 8,400 added jobs on a base of 38,600 in 2025, with about 2,900 openings a year.

Openings include replacements as well as new jobs, so don’t read 2,900 as the number of new roles. And since the occupation is small, each opening draws real competition.

Projections are estimates. BLS revises them each cycle, and they can’t tell you what one employer will pay.

What Are the Limits of This Data?

The survey counts occupations, not titles. An employer may call your job AI research scientist and file it under another BLS code, and the pay would then follow that code.

Wages include everyone in the occupation, so a new graduate and a veteran with two decades of experience share one median. Percentiles are the closest we can get to an experience view.

Employment counts and wages are missing for some states, and small states can post unstable averages. And the survey measures wages, not total pay. We have no data on bonuses or equity, so we don’t estimate them.

None of that makes the numbers useless. Use them to set a range, then check a few real postings in your state.

What Should You Do With These Numbers?

Treat them as a range. Look up your state in the tables above, then compare it with the program you’re considering.

Then talk to people who hold the job. Ask what they were hired as, what degree they have and what their pay looks like next to the figures here. A federal survey can’t say what one title pays at one company, but those conversations can.

When you’re ready to look at the work itself, read our guide to AI research scientist jobs. To weigh the degree against its cost, see whether an AI degree is worth it.