Entry-Level AI Startup Jobs in August 2026: 325 Openings, and Where They Are
By Fast AI Startup Jobs
The argument about young workers and AI got loud this summer. Researchers tracking employment data report a 13% to 16% relative decline in jobs held by 22-to-25-year-olds in the occupations most exposed to AI. Other analysts look at the same frozen bottom rung and argue the cause is cyclical: a hiring market that stopped moving after a rate cycle, not a model that learned to do junior work.
We can't settle that. What we can do is check one corner of the market where we have unusually good visibility, which is the AI startups themselves. If AI were creating a new class of junior work anywhere, this is where it would show up first.
It hasn't. As of today, 325 of the 28,414 open roles in our database carry an entry-level title. That's 1.1% of everything on the board, and where those 325 sit turns out to be more interesting than the number itself.
What the 325 actually are
We counted a posting as entry-level if the title says so: intern, internship, new grad, entry level, early career, campus. It's a strict test. It misses roles that are junior in practice but titled like any other opening, so treat 325 as a floor rather than a census. It doesn't make many mistakes in the other direction, though. A title that says "intern" is an intern.
| Type | Postings | Share of 325 |
|---|---|---|
| Internship | 255 | 78.5% |
| New-grad / entry-level | 70 | 21.5% |
Four out of five entry-level postings are internships. That leaves 70 permanent junior roles spread across 1,824 tracked companies, one for every 26 companies on our list. Both cuts are live boards if you want to work them directly: internships and new-grad roles.
The split says more than the total does. An internship is a company buying a twelve-week option on you with no obligation to exercise it. A new-grad requisition is salary, a desk, and somebody's mentoring hours committed to a person who won't ship anything meaningful for a quarter. AI startups are doing a lot of the first and very little of the second, and the 3.6-to-1 ratio between them is the clearest signal in this dataset about how much risk they're willing to take on unproven people.
Almost none of it is remote
Seventeen of the 325 entry-level postings are remote. That's 5.2%. Across all 28,414 postings the remote share is 13.0%, so the entry rung is less than half as remote as the market it sits inside.
| Cut | Remote share |
|---|---|
| Entry-level postings | 5.2% |
| All postings | 13.0% |
This is the finding we'd hand to a graduating senior first, because it quietly kills a lot of search strategies. If you're filtering for remote and entry-level at the same time, your entire addressable market at AI startups right now is seventeen jobs.
The reason is apprenticeship, not policy. Companies hire juniors where somebody can watch them work, correct them in the moment, and absorb the cost of the first few months. That's hard over Slack and easy across a desk. Our analysis of remote roles at AI startups found the overall remote share is already thinner than the discourse suggests, and the entry-level slice is where it thins out most.
The door is widest at Seed
We looked at what share of each funding stage's postings are entry-level. The pattern runs almost cleanly downhill.
| Funding stage | Entry-level | Total postings | Entry-level share |
|---|---|---|---|
| Pre-Seed | 0 | 72 | 0.00% |
| Seed | 34 | 1,272 | 2.67% |
| Series A | 70 | 4,597 | 1.52% |
| Series B | 53 | 5,103 | 1.04% |
| Series C | 64 | 4,712 | 1.36% |
| Series D | 29 | 3,176 | 0.91% |
| Growth | 6 | 1,065 | 0.56% |
Seed companies post entry-level roles at roughly five times the rate of growth-stage ones. Series C breaks the slope, sitting above Series B, and three of the heaviest entry-level hirers in our data happen to be Series C hardware companies. We'll come back to them.
Read this table twice, though, because the best rate and the most jobs are not at the same stage. Seed has the friendliest odds and 34 openings. Series A has 70, more than any other stage, at a still-healthy 1.52%. If you're optimizing for the chance that any given application lands, aim at Seed. If you're optimizing for how many applications you can send at all, Series A is where the volume is. Pre-Seed is a rounding error: 72 open postings, none of them entry-level, which is what you'd expect from teams of four people who need every hire to arrive already knowing how to do the job.
Who's hiring juniors right now
Twelve companies account for a meaningful chunk of the 325. The interesting column isn't the count, it's the last one.
| Company | Stage | Category | Entry roles | Total open roles | Entry share |
|---|---|---|---|---|---|
| Anduril | Series G+ | Defense Tech | 18 | 978 | 1.8% |
| Etched | Series C | AI Infrastructure | 17 | 102 | 16.7% |
| Aircall | Series D | Enterprise SaaS | 11 | 71 | 15.5% |
| Base Power | Series C | Climate Tech | 10 | 162 | 6.2% |
| Simular | Series A | AI Agents | 10 | 22 | 45.5% |
| Netic | Series B | Enterprise SaaS | 7 | 21 | 33.3% |
| Bending Spoons | Post-IPO | Consumer | 6 | 36 | 16.7% |
| Figure AI | Series C | Robotics | 6 | 118 | 5.1% |
| Blackstar | Seed | Hardware & Robotics | 5 | 14 | 35.7% |
| Meshy | Series A | Generative AI | 5 | 22 | 22.7% |
| Flexion Robotics | Series A | Robotics | 5 | 14 | 35.7% |
| Notion | Private equity | Enterprise SaaS | 5 | 108 | 4.6% |
Anduril leads on raw count with 18 entry-level openings, and that's 1.8% of its 978 postings. It's a big company running a normal-sized campus program. Simular is the opposite trade: 10 of its 22 open roles are entry-level, which means nearly half of everything that company is hiring for right now is junior. Blackstar and Flexion Robotics are both hiring 14 people total and five of them are early-career.
Small companies with high entry shares are worth more of your time than the household names with more openings. At Anduril your application lands in a pile sorted by a recruiting team with a process. At a fourteen-person robotics company, a founder reads it.
Notice what most of this list has in common. Etched builds transformer chips. Base Power builds home batteries. Figure AI, Blackstar, and Flexion build robots. When the product is physical, the work is physical, and someone has to be standing next to the hardware. That's the same force behind the 5.2% remote number, and it's the strongest current running through this month's entry-level data: junior hiring is concentrating where the job requires a body in a building.
The map
Entry-level openings cluster harder than you'd guess.
| City | Entry-level postings |
|---|---|
| San Francisco | 48 |
| San Jose | 24 |
| Austin | 14 |
| Palo Alto | 10 |
| Atlanta | 10 |
| New York City | 8 |
| Costa Mesa | 8 |
| Berlin | 7 |
San Francisco, San Jose, and Palo Alto hold 82 of the 325 between them. A quarter of every entry-level AI startup job in our database sits inside a twenty-minute drive of the others. San Francisco alone has six times as many entry-level postings as New York City.
Costa Mesa is on the list because Anduril is headquartered there, so those eight are effectively one employer's campus program showing up as a city. Austin at 14 and Atlanta at 10 are the ones we didn't expect, and Berlin is the only non-US entry in the top eight. Combined, these eight cities cover 129 of the 325 postings. The other 196 are scattered thin enough that no single city carries more than a handful.
Which brings the relocation question into focus. If you're early-career, remote is nearly closed to you, and the densest concentration of open doors is in three adjacent Bay Area cities. That's an expensive answer, and it's the one the data gives.
Three months later, the door is the same width
In May we ran a version of this analysis and found 458 entry-level roles out of 35,450 listings, or 1.3%. This month's number is 1.1%. Do not read that as a decline.
| Analysis | Method | Entry-level | Denominator | Share |
|---|---|---|---|---|
| May 2026 | ATS seniority labels | 458 | 35,450 raw listings | 1.3% |
| August 2026 | job-title matching | 325 | 28,414 deduplicated postings | 1.1% |
The two numbers were produced by different instruments. May used the seniority label companies attach in their applicant tracking system, counted against raw listings, which double-count a single role posted to several locations. August used title text against postings deduplicated by job ID. Either change alone would move the result by more than the gap between 1.3% and 1.1%, so subtracting one from the other tells you about our method, not about the market.
Here's the honest version. Measured two different ways, three months apart, the entry door at AI startups stays around one percent. That's the finding. A stable, very small number is a more useful thing to know than a fake trend line, and it's a reasonable data point to set next to the national argument we opened with. At the companies actually building AI, the bottom rung was already this narrow in May, before this summer's round of commentary about whether AI is closing it.
If you're the one applying
Weight the search toward Seed and Series A. Series A carries the most openings in absolute terms and Seed carries the best odds per application, and both beat Series D and growth-stage by a wide margin.
Assume the job is on-site and plan around that rather than filtering it away. Seventeen remote entry-level postings is not a search strategy.
Treat internships as the main channel, because at 255 of 325 that's what they are. And spend disproportionate time on the small companies with high entry shares in the table above, where a junior hire is a real fraction of the team rather than a line item in a campus program.
Methodology
The data comes from the Fast AI Startup Jobs database as of August 3, 2026: 28,414 open postings across 1,824 tracked AI startups, deduplicated by job ID so that one role advertised in three cities counts once.
A posting was classified as entry-level if its title contains intern, internship, new grad, entry level, early career, or campus. This is a title-text method, not a seniority-label method. It's precise about what it catches and it undercounts roles that are junior in substance but conventionally titled, so 325 is a lower bound on genuinely early-career openings.
Funding stage comes from company profiles in the same database, which covers 97% of tracked companies. The seven stages in the stage table account for 256 of the 325 entry-level postings; the remaining 69 sit at Series E and later, post-IPO, or private-equity-backed companies that we did not break out separately. Remote classification uses the location field on each posting.
Two limitations carry over from our earlier work. Our coverage comes from major ATS platforms, so companies that post only to their own site or hire through founder networks are invisible here, which skews the dataset toward companies structured enough to run an ATS. And city counts reflect the location written on the posting, so a role listed in a company's headquarters city may be worked elsewhere.
Data sourced from fastaijobs.com as of August 3, 2026. We analyzed 28,414 deduplicated open roles across 1,824 tracked AI startups; funding data covers 97% of companies.
Related: We Analyzed 35,000 AI Startup Job Listings. Only 1.3% Are Entry-Level., our May analysis using ATS seniority labels, and the reality of remote AI startup jobs, which explains why the 13.0% remote baseline is itself smaller than most job seekers expect.