Where do these salary figures come from?

The US Department of Education's College Scorecard Field of Study file, updated 2 November 2026. It reports median earnings one year after completion for graduates who are working and not enrolled in further study, for each specific programme at each institution.

Which US university pays off best for a computer science master's?

By median earnings one year after completion, Stanford leads at $159,729 from a cohort of 57. But Georgia Tech reports $121,610 on median graduate debt of only $21,664 from a much larger cohort of 375, which is the strongest earnings-to-debt result in the table.

What does a typical CS master's actually earn?

Across the 108 computer science master's programmes with a cohort of at least 50, the median programme reports $78,392. The programmes in our table are the top of a distribution running from about $44,120 to $159,729.

What do data science master's graduates earn?

The federal data cannot say. Of the 317 data science and data analytics master's programmes in the file, zero report median earnings — the programmes are too new for a reportable cohort to have graduated and been observed for a year. Any specific figure you see quoted is not from this source.

Is engineering or computer science better paid at master's level?

The median engineering programme reports $92,527 against $78,392 for the median computer science programme, so engineering has the higher floor. Computer science has the higher ceiling — its top programmes report more than any engineering programme in the file.

Why is my university missing from the table?

Most likely because its cohort is too small. The Department of Education suppresses figures drawn from small groups to protect individual privacy, and we additionally excluded any programme with a cohort under 50. Absence is not evidence that a programme is poor.

Does this data show which universities lead to H-1B sponsorship?

No. No federal dataset links a graduate's university to whether a later employer sponsored them. Sponsorship is a decision made by the employer, and it is governed by the H-1B cap rather than by where you studied.

Best US Universities for MS: Real Earnings & Debt Data (2026)

Q: Does the university I choose affect my STEM OPT eligibility?

Not the university — the CIP code of the specific programme. The 24-month STEM OPT extension depends on whether the CIP code printed on your Form I-20 appears on the DHS STEM Designated Degree Program List. Two similarly named programmes can carry different CIP codes, so confirm it with the department before accepting an offer.

Rankings measure reputation. The Department of Education measures what graduates of a specific programme actually earned a year later. Here is what that file says about CS and engineering master's — and the one field where it says nothing at all.

Ask which American university is best for an MS and you will be handed a ranking. Rankings are built from reputation surveys, faculty counts and selectivity. None of those is the thing you are actually buying when you move countries and borrow money.

The US Department of Education publishes something closer to it: what graduates of a specific programme at a specific university actually earned a year after finishing. This article is built entirely from that file. Where it cannot answer a question, we say so rather than reaching for a ranking.

What this data is, exactly

The figures below come from the College Scorecard Field of Study file, last updated by the Department of Education on 2 November 2026. For each programme it reports:

Median earnings one year after completion, for graduates who are working and not enrolled in further study, and whose highest credential is that master's.
Median cumulative graduate debt for students who completed that programme.
The cohort size the earnings figure is calculated from.

We have filtered to programmes with a cohort of at least 50. Below that the median moves too much on too few people to mean anything.

These are earnings in the US labour market, not a promise. They describe people who graduated and stayed and worked. They say nothing about your odds of admission, your visa outcome, or whether an employer will sponsor you. Those are separate problems, and this file does not model any of them.

Computer science: the top of the table

Institution	Programme	Median earnings 1 yr after	Median grad debt	Out-of-state tuition/yr	Cohort
Stanford University Stanford, CA · Private	Computer Science	$159,729	$33,500	$68,544	57
Cornell University Ithaca, NY · Private	Computer Science	$147,241	$36,706	$72,270	77
Columbia University in the City of New York New York, NY · Private	Computer Science	$133,851	$41,000	$74,180	54
Georgia Institute of Technology Atlanta, GA · Public	Computer and Information Sciences, General	$121,610	$21,664	$35,092	375
University of Washington-Seattle Campus Seattle, WA · Public	Computer and Information Sciences, General	$116,990	$37,297	$44,640	95
Southern Methodist University Dallas, TX · Private	Information Science/Studies	$115,112	$56,716	$69,722	70
Johns Hopkins University Baltimore, MD · Private	Computer and Information Sciences, General	$111,651	$30,500	$67,170	137
University of Southern California Los Angeles, CA · Private	Computer Science	$110,796	$41,882	$75,162	81
Carnegie Mellon University Pittsburgh, PA · Private	Computer and Information Sciences, General	$109,615	$65,734	$68,776	108
Pennsylvania State University-Penn State Great Valley Malvern, PA · Public	Information Science/Studies	$109,295	$41,000	n/a	53
Johns Hopkins University Baltimore, MD · Private	Information Science/Studies	$108,932	$34,914	$67,170	75
Georgetown University Washington, DC · Private	Computer/Information Technology Administration and Management	$105,843	$55,779	$71,338	114

The number that should change how you read this

Across all 108 computer-science master's programmes with a cohort of 50 or more, the median programme reports median earnings of $78,392. The table above is the top of a distribution that runs from about $44,120 to $159,729.

In other words, the gap between a top programme and a typical one is roughly double. That gap is far larger than the gap between any two adjacent rows in a published ranking, and it is the gap that actually matters.

Georgia Tech is the outlier worth studying

Look at the fourth row. Georgia Tech reports $121,610 median earnings on $21,664 median graduate debt, from a cohort of 375 — by far the largest in the table.

Compare that with Carnegie Mellon at $109,615 on $65,734 of debt. Both are excellent computer science departments. One leaves its median graduate with roughly three times the debt of the other, and earns less a year out.

Why the large cohort matters. A median drawn from 375 graduates is far more stable than one drawn from 54. When you compare programmes, read the cohort column before the earnings column — a $160,000 median on 57 people is a noisier claim than $121,000 on 375.

Engineering

Institution	Programme	Median earnings 1 yr after	Median grad debt	Out-of-state tuition/yr	Cohort
Massachusetts Institute of Technology Cambridge, MA · Private	Electrical, Electronics, and Communications Engineering	$144,716	not reported	$64,730	100
Northwestern University Evanston, IL · Private	Engineering, Other	$140,321	$115,500	$70,589	79
Massachusetts Institute of Technology Cambridge, MA · Private	Mechanical Engineering	$135,266	not reported	$64,730	54
Massachusetts Institute of Technology Cambridge, MA · Private	Systems Engineering	$133,685	$62,750	$64,730	66
Johns Hopkins University Baltimore, MD · Private	Systems Engineering	$114,155	$26,223	$67,170	119
University of Southern California Los Angeles, CA · Private	Electrical, Electronics, and Communications Engineering	$108,989	not reported	$75,162	61
George Washington University Washington, DC · Private	Systems Engineering	$106,809	$42,593	$70,170	55
University of Washington-Seattle Campus Seattle, WA · Public	Electrical, Electronics, and Communications Engineering	$106,287	$41,166	$44,640	50
University of Colorado Boulder Boulder, CO · Public	Engineering, General	$105,752	$39,500	$47,142	55
University of Michigan-Ann Arbor Ann Arbor, MI · Public	Systems Engineering	$102,847	not reported	$63,962	56
Embry-Riddle Aeronautical University-Daytona Beach Daytona Beach, FL · Private	Systems Engineering	$101,436	$32,139	$46,181	78
Johns Hopkins University Baltimore, MD · Private	Electrical, Electronics, and Communications Engineering	$101,186	$22,874	$67,170	101

Across all 59 engineering master's programmes meeting the same cohort threshold, the median programme reports $92,527 — noticeably higher than the computer-science median of $78,392. The ceiling in CS is higher; the floor in engineering is higher.

Note Northwestern in the second row: $140,321 earnings against $115,500 of median graduate debt. That is the highest debt figure anywhere in either table, and it is close to a full year of earnings.

Data science: the data cannot tell you yet

This is the finding most likely to be useful to you, and no ranking will tell you.

The file contains 317 master's programmes in data science and data analytics. The number of them that report median earnings is zero. Only 32 report debt.

The reason is timing. These programmes are new. The earnings measure needs a cohort that graduated, entered the workforce and was observed for a full year, and most of these programmes do not have one yet at a reportable size.

So be careful what you are told about data science salaries. Any specific figure quoted for a named US data science master's is not coming from this federal file, because this file does not have one. It is coming from self-reported surveys, the university's own marketing, or a general salary site that is not programme-specific.

How much of the picture is missing

Honesty about coverage matters more than the table itself:

Field	Master's programmes in the file	With usable earnings (cohort ≥ 50)
Computer & information sciences	1,454	108
Engineering	2,431	59
Data science & analytics	317	0

Most programmes are missing because their cohorts are too small to report without risking individual privacy — the file suppresses those. A programme's absence from the table is not evidence that it is bad. It usually means it is small.

One quirk to know when you read the raw file

Some university systems report a single figure across several campuses. The University of Washington's Seattle, Bothell and Tacoma campuses carry an identical earnings figure, as do four Penn State campuses. We have collapsed those duplicates to one row each. If you pull the file yourself, you will see the repeats.

The part that is about visas, not money

One field in this data has a direct immigration consequence: the CIP code, the federal classification number attached to every programme.

Your eligibility for the 24-month STEM OPT extension depends on whether the CIP code printed on your Form I-20 appears on the DHS STEM Designated Degree Program List. Not on the degree's title, not on how technical the coursework feels — on that number.

Check the CIP code before you accept an offer, not after. Two programmes with almost the same name can carry different CIP codes, and only one of them may be on the DHS list. Ask the admitting department which CIP code your I-20 will show, then check it against the official list linked below. We cover how the extension itself works in our OPT and STEM OPT guide.

What we are deliberately not publishing

A ranking. We have ordered programmes by one measurable variable. That is not the same as saying one university is better than another, and we are not saying it.
Admission rates or GRE cut-offs for these master's programmes. The admission rate in the Scorecard institution file is an undergraduate figure. Applying it to graduate admissions would be wrong, so we have left it out.
Cost of attendance. The tuition column is published out-of-state tuition and fees. It is not your total cost — it excludes living costs, health insurance and fees that vary by programme. Treat it as a comparison between rows, not as a budget.
Sponsorship rates by university. No federal file links a graduate's university to whether their employer later sponsored them.

How to actually use this

Start from the debt column, not the earnings column. Earnings vary with the job market. Debt is fixed the day you sign, and you repay it in dollars while possibly on a temporary visa.
Check the cohort size before you trust any median.
Confirm the CIP code against the DHS STEM list before accepting.
Look up your exact programme on the College Scorecard site rather than relying on this table. We show twelve rows; the file has tens of thousands.
Treat a missing programme as unknown, not as bad.

If your plan after the degree is to work in the US, the mechanism that gets you from an F-1 to a job is the part worth understanding next — that is cap-gap, and then the H-1B cap itself.

Which US Master's Actually Pays: What the Federal Earnings Data Shows