Computer Science Unemployment Rate in 2026: What Four New Studies Mean for International CS Grads on OPT

Four new 2026 studies disagree on how hard AI is hitting new CS grads, and on OPT the disagreement matters less than how you spend your unemployment days.

Part of our guide to OPT unemployment days.

There is no single official computer science unemployment rate for 2026, but the best current data lines up like this. The New York Fed puts the overall recent-graduate unemployment rate at about 5.6% for 2026:Q2, with 42% underemployment. CS looks worse at the entry point: a September 2026 Census paper found the most AI-exposed majors (CS, information systems, software) lost 5 percentage points of initial employment probability after ChatGPT. Yet an NBER paper found no AI-linked spike in recent-grad unemployment this summer. Both can be true. If you are an international CS grad on OPT, the gap between them matters, because your 90-day unemployment limit keeps running whichever headline is right.

Status line, as of 2026-10-14. In effect: the 90-day OPT and 150-day STEM OPT unemployment limits; the wage-weighted H-1B selection rule (effective 2026-02-27, used for FY2027). Proposed only: OPT fees of $70,000 initial / $30,000 subsequent (NPRM published 2026-10-08, comments close 2026-11-09; the school would be the legal payer and may pass the cost on). Blocked: the Duration-of-Status rule (enjoined 2026-09-14, DHS has appealed; the 60-day F-1 grace period still applies) and the $100,000 H-1B fee (Proclamation 11069 extended the restriction to 2027-09-21, but two court orders block the fee). None of these change the job-market data below.

The computer science unemployment rate in 2026, by source

Nobody publishes a CS unemployment rate for international grads specifically. There are four new studies plus the NY Fed's recurring data series, and each measures something different. As of October 2026:

The NY Fed's by-major table is the right place to look up the current computer science line yourself. We aren't printing a CS-specific number here because it gets revised, and a stale figure is worse than none.

How the four studies fit together

At first glance they contradict each other. They mostly don't, because they measure different stages of a career.

Fewer doors, not more people pushed out

The Stanford and Indeed findings describe the same mechanism from two angles. Stanford says young workers in AI-exposed roles are below trend mainly because companies hired fewer of them, not because existing workers were let go. Indeed shows the posting side: in the most exposed occupations, the share of postings aimed at entry-level candidates fell from 29% to 10% while senior postings rose from 22% to 47%. So employers are still hiring, but they are asking for experience.

We cover the "automate vs augment" distinction behind the Stanford data in more depth in our breakdown of the Stanford entry-level hiring study.

Why unemployment didn't spike anyway

The NBER paper looked for an AI-driven jump in recent-grad unemployment in summer 2026 and didn't find one. If hiring of 22-25-year-olds is below trend, where did those graduates go? The Census paper gives part of the answer: for the most AI-exposed majors, initial earnings fell 13%, and about half of that drop came from graduates shifting into lower-wage sectors. Many took jobs, just not the jobs they trained for. That fits the NY Fed's 42% underemployment figure. The NBER authors also found that occupational remote-work availability correlated with 2026 unemployment, a reminder that AI is not the only force in play.

What this means in plain terms

• The computer science graduates unemployment problem is mostly a first-job problem. The hardest part is the first role, and that is exactly the stage you are at.
• The AI impact on entry-level jobs shows up in postings and pay before it shows up in unemployment rates. A low national rate can coexist with a brutal new-grad search.
• It is not economy-wide displacement. The Stanford authors say so directly and call their numbers descriptive, not causal.

Why OPT makes the same market harder

A US citizen with a CS degree who spends eight months job hunting has a bad year. An F-1 graduate on OPT who does the same can run out of permitted unemployment time.

That third row matters. The Census finding that part of the earnings drop came from shifts into lower-wage sectors describes a fallback that works for citizens. For you, a job outside your field may not count as qualifying OPT employment, so the days keep running. Whether a specific role qualifies is a question for your DSO, not a recruiter and not a blog post.

For a day-by-day look at how the count works, see how STEM OPT unemployment days are counted, and for tactics, our guide to beating the OPT 90-day unemployment clock.

Is AI taking entry-level jobs for CS grads specifically?

Partly, and in a specific way. The research supports three claims and does not support a fourth.

1. Supported: postings in AI-exposed fields have shifted sharply toward senior candidates (Indeed).
2. Supported: young workers in those occupations are employed below trend because of reduced hiring (Stanford).
3. Supported: the most AI-exposed majors, CS among them, lost 5 percentage points of first-job employment and took a recession-sized hit to first-job pay after ChatGPT (Census).
4. Not supported: that AI caused a broad spike in recent-graduate unemployment in 2026 (NBER found none), or that it is displacing workers economy-wide (Stanford's own authors say it isn't).

So "AI taking entry-level jobs" is better read as "AI is raising the experience bar for the first job," a problem you can partly work on by building and showing experience.

What the salary data says for your H-1B math

NACE's Winter 2026 projections put average starting salaries at $81,535 for CS bachelor's grads (NACE) and $94,212 for CS master's grads (NACE). Those are averages, and the Census paper's 13% earnings drop for exposed majors is a warning that many offers land below them.

Salary matters more for you than for a citizen classmate, because the wage-weighted H-1B selection rule (in effect since 2026-02-27 and used for FY2027) gives registrations at higher DOL wage levels more chances in the lottery. How a specific offer maps to a wage level depends on the occupation code and work location, which your employer's immigration counsel determines. Our explainer on the wage-weighted lottery for new grads walks through the mechanism, and the MS computer science H-1B sponsorship guide covers which employer types sponsor CS master's grads.

A 30-60-90 plan built for this market

Entry-level postings are scarce, employers want proof of experience, and your unemployment days are fixed. Build around all three.

1. Before your OPT start date: confirm your EAD validity dates and ask your DSO how unemployment days are tracked in SEVIS for your record. Set a calendar countdown.
2. Days 1-30: narrow your target list to employers with a sponsorship history, using public LCA data on dol.gov. Mass applications run straight into the shrinking entry-level share; targeted ones sidestep part of it.
3. Days 1-30, in parallel: convert class and capstone work into evidence an employer reads as experience: a deployed project, a measurable result, a code sample a reviewer can run.
4. Days 31-60: shift effort toward referrals and direct outreach to hiring managers. When postings skew senior, the junior roles that exist are often filled before or without a public posting.
5. Days 31-60: widen titles, not fields. QA/SDET, data engineering, solutions engineering and support engineering can be directly related to a CS degree. Confirm with your DSO that a specific role relates to your major before accepting it.
6. Days 61-90: if you are approaching the limit, talk to your DSO before day 90, not after, about how your record stands and what options exist. Questions about H-1B or other visa paths go to a licensed immigration attorney.

Common mistakes

• Treating the national unemployment rate as your odds. A 5.6% recent-grad rate tells you nothing about your OPT clock. Your constraint is days, not percentages.
• Mixing up study vintages. The Stanford figure cited here is the August 2026 revision (about 19% below trend); the July 2025 vintage put the gap at about 15%. Quote one vintage, with its no-displacement caveat.
• Reading "below trend" as "being fired." Stanford and Indeed describe reduced hiring. That means your problem is getting in the door, so spend effort there.
• Taking an unrelated job to "stop the clock." Work that is not directly related to your degree may not count as OPT employment. Ask your DSO before you accept.
• Assuming proposed rules are law. The OPT fee is proposed, not in effect. The $100,000 H-1B fee is blocked. Don't decline an offer or change plans based on a headline; check the status date.
• Waiting for the market to recover. The NBER and Census papers measure different things and nobody has a recovery date. Your unemployment limit does not wait for researchers to agree.

Frequently asked questions

What is the computer science unemployment rate for recent grads in 2026?

The New York Fed puts the overall recent-graduate unemployment rate at about 5.6% for 2026 Q2, with underemployment at 42%. It also publishes a major-by-major table, which is the place to check the current computer science figure. Census research adds that the most AI-exposed majors, including CS, saw a 5-percentage-point drop in initial employment probability after ChatGPT.

Is AI taking entry-level computer science jobs?

The evidence points to fewer entry-level openings rather than mass job losses. Stanford found employment for ages 22-25 in highly AI-exposed occupations about 19% below trend, driven by reduced hiring, and its authors say this shows no economy-wide displacement. An NBER paper found no summer-2026 spike in recent-graduate unemployment linked to AI.

How many unemployment days do I get on OPT?

Post-completion OPT allows up to 90 days of unemployment. If you receive the 24-month STEM OPT extension, the limit becomes 150 days in aggregate across both periods. Your DSO is the authority on how your specific days are counted in SEVIS; SEVP's Study in the States help hub explains the STEM OPT side.

Does a high computer science unemployment rate change my OPT rules?

No. National labor statistics do not change your 90-day or 150-day limit, your reporting duties, or the requirement that OPT work relate to your degree. A slow market only makes those fixed limits harder to work within, so plan your search around them early.

How much do CS new grads earn in 2026?

NACE Winter 2026 projections put average starting salaries at $81,535 for CS bachelor's graduates and $94,212 for CS master's graduates. Those are averages across employers, not offers you can count on, and Census research found initial earnings for the most AI-exposed majors fell 13% after ChatGPT.

This post explains how the research and the OPT rules work. It is not legal advice. Questions about your SEVIS record or whether a job qualifies go to your DSO; questions about H-1B or green-card strategy go to a licensed immigration attorney.

If you want help turning a tight CS market and a fixed OPT clock into a focused search plan, talk to F1Jobs. We run a paid job-search service, so weigh that; the plan above does not depend on working with us.