Who's Actually Hiring AI Engineers in 2026
Open any company's careers page in 2026 and you will be told, somewhere above the fold, that they are an AI company. The payroll software is AI-powered. The dog-walking app is AI-native. Your bank is, apparently, an AI lab that happens to hold your money.
So I did the boring thing and checked. I went through all 1,077 tech companies on Remoet and asked two separate questions. One: how many describe themselves with the word "AI"? Two: how many actually run a machine-learning stack (PyTorch, TensorFlow, an LLM, something a real ML engineer would touch on a Tuesday)?
The gap between those two numbers is the whole story.
83% say AI. 31% have the stack.
890 of the 1,077 companies (that's 83%) have "AI" somewhere in how they describe themselves or tag their work. Nearly five in six.
337 of them, 31%, actually have a machine-learning technology in their stack. (I counted any of PyTorch, TensorFlow, an LLM, deep learning, Hugging Face, Keras, or JAX. A low bar, on purpose.)
So of the 890 companies waving the AI flag, fewer than four in ten have anything under the hood. The word has detached from the work. "AI company" in 2026 means roughly what "digital transformation" meant in 2016: a thing you put in the headline because everyone else did.
And it shows up in the jobs, not just the branding. Of the 15,000-odd open roles across all these companies, only about 1,100 are actual AI or ML jobs (machine learning, data science, applied research, that kind of thing). One in fourteen. Everyone markets AI. Almost nobody is hiring for it at volume.
For you, the person actually looking for an AI or ML job, that's the trap. Search "AI companies" and you get 890 results, most of which want a regular backend engineer and slapped "AI" on the job description because marketing asked. The 337 are the ones where the work is real.
The companies building AI as the product
Start with the obvious tier: the companies where AI is the thing they sell. These are the ones hiring the most right now, by open roles:
OpenAI, 324 open roles (ChatGPT, GPT, Sora)
Relativity, 191 (AI for legal e-discovery)
Databricks, 157 (the data and AI platform)
Anthropic, 58 (Claude)
Harvey, 57 (legal AI)
Cursor, 45 (the AI code editor)
Scale AI, 44 (training data and evals)
Mistral AI, 38 (open-weight frontier models)
Cohere, 32 (enterprise LLMs)
Synthesia, 31 (AI video)
Behind them the same tier keeps going: Together AI (28), Suno (27, AI music), Deepgram (26, voice AI), Cognition (13, the Devin people), Writer (12), Perplexity (9).
If you want to work on models themselves, this is the list. But notice something before you pour all your applications into it: with a couple of exceptions, these are not big companies. The frontier-lab tier is genuinely small, and the seats skew senior. Which is exactly why the next part matters more than most job seekers realize.
The plot twist: the biggest ML employers aren't AI companies
Here's the part I didn't expect when I started counting.
Once you filter to "has a real ML stack," the list stops being dominated by AI labs. It's everyone else. Machine learning quietly became a horizontal (a thing every category now does in-house), and the companies with the deepest ML hiring are often the ones that never call themselves an AI company at all.
The pattern, by category:
Fintech and payments. Stripe (113 roles), Robinhood (64), Adyen (60), Ramp (35), Block (35), Affirm (31), Chime (20). Fraud, risk, and underwriting are ML problems, and these teams have been at it for years.
Developer tools and infrastructure. MongoDB (100), Datadog (73), Snowflake (70), GitLab (58), Cloudflare (38).
Consumer health and wearables. Oura (51), WHOOP (49), Eight Sleep (23), Neko Health (9). Each one is an ML company wearing a hardware costume.
Marketing and CRM. Klaviyo (50), Braze (20), Attentive (11), HubSpot (10).
None of these will headline a piece about AGI. All of them are hiring people to build and ship machine learning in production, today, often with more open seats than the frontier labs and less of a stampede at the door.
If your goal is to do ML work rather than to work at an AI company (and for most people those are different goals), this is where a lot of the actual jobs are.
How to find your slice without reading 337 company pages
The catch, same as always: this only helps if you don't have to open 337 tabs to use it. You will quit around tab nine. So would I.
That grind is the part Remoet takes off your plate. Connect whatever AI agent you already use (Claude, Cursor, whatever's open) to Remoet over MCP. Then ask it for the cut you actually want:
"Find companies with a PyTorch or LLM stack that have open mid-level roles."
"Which fintech companies are hiring ML engineers, and what's the rest of their stack?"
"Show me AI-native companies using Python and Rust."
"Star the ones that also do computer vision."
It reads each company's real, tag-level stack, not the marketing copy, and hands you the shortlist. Star the ten or fifteen that fit, and from then on you get one email a week when those companies post something new. You read the ones worth reading. The agent does the soul-deadening middle.
Where this data is soft
I'd rather you trust these numbers than have me oversell them, so here's where they bend.
The "has AI" count is generous. It catches anything with "AI" in the name, description, or tags, which sweeps up plenty of companies that merely mention it. That's the point, I'm measuring the claim, not the substance. The ML-stack count is the stricter number, and it's the one I'd hang my hat on.
The stacks themselves are auto-detected from job postings and company data, so a company doing real ML that never wrote it down reads as a false negative. The true number of ML builders is probably a little higher than 337, not lower.
The "one in fourteen roles is AI/ML" figure is detected from job titles, so it catches the obvious ones (machine learning, data scientist, applied scientist, computer vision, MLOps) and misses an AI role hiding under a generic "Software Engineer" title. Treat it as a floor, not a ceiling.
And the company role counts are total open roles, not ML roles specifically. OpenAI's 324 includes recruiters and salespeople, not 324 research scientists. Use the count as a signal of "this company is hiring hard," then go check the actual roles. (Counts are also distinct titles, not seats: one role open in five cities shows up once, so the real number of openings runs a bit higher.)
Last thing, it's a snapshot, late June 2026. Next month's list will have moved, which is the whole reason tracking companies beats bookmarking a page.
The takeaway holds either way. Almost everyone says AI now. About a third can back it up. If you're looking for ML work, the trick isn't finding the companies that say it, it's finding the ones that ship it, and a lot of those aren't the ones you'd think. If you go dig into your own slice and find one I missed, or something I got plain wrong, come tell me. I want to know.
