Every AI company says they're "flexible." Most have a "remote-friendly" bullet point on their careers page. But when you look at actual job listings, employee reviews, and office attendance expectations, a very different picture emerges.
We maintain culture profiles for 118 AI and tech companies. For each one, we evaluate remote policies using strict criteria: does the company have mandatory office days? Are the majority of roles listed as remote? Do employee reviews confirm remote actually works there? Do remote workers have equal career opportunities? Only companies that pass all four tests earn the "remote-first" designation in our directory.
The result: just 23 out of 118 companies — 19% — are genuinely remote-first. The other 81% require some form of in-person presence, whether that's hybrid (2-3 days/week), "flexible" (meaning in-office with occasional remote days), or fully on-site.
The 23 Companies That Are Actually Remote
These companies don't just allow remote work — they're architecturally built around it. No mandatory office days. Async-first communication. Remote workers at every level of leadership. Most were founded as distributed companies or made a permanent structural commitment to remote during or after 2020.
Other remote-first companies in our directory: ElevenLabs, n8n, Pinecone, Weaviate, Chainguard, Twilio, Webflow, Runway, Descript, Apollo.io, and Poolside.
Why 81% of AI Companies Aren't Remote
The AI industry, paradoxically, is one of the least remote-friendly sectors in tech. Three structural forces explain this:
1. GPU clusters require physical proximity
Training frontier AI models requires massive GPU clusters. The teams building and maintaining these systems — infrastructure engineers, ML researchers doing pre-training, systems engineers — often work in or near data centers. Companies like OpenAI, Anthropic, and DeepMind have strong in-office cultures partly because the research requires tight collaboration around hardware resources that don't exist in the cloud.
2. Research velocity favors co-location
AI research moves fast. Papers get published weekly. Competitive dynamics between labs create intense pressure to iterate quickly. Many AI research leaders believe that co-located teams can maintain higher research velocity through spontaneous whiteboard sessions, rapid pair programming, and real-time debugging of training runs. Whether this belief is accurate is debatable — but it shapes hiring decisions.
3. Well-funded companies can afford offices
The AI companies with the most funding (OpenAI at $10B+, Anthropic at $7B+, DeepMind backed by Google) can afford premium office space in SF, London, and NYC. They use offices as a recruiting tool — beautiful spaces with catered meals and on-site amenities. Companies with less funding are more likely to go remote out of financial necessity.
Remote vs. Office: What the Work-Life Balance Data Shows
One question matters more than the philosophical debate: do remote-first companies actually produce better employee experiences? Our data says yes — decisively.
Across our 118-company dataset:
- Remote-first companies average 4.0/5 work-life balance on employee reviews
- Office-first/hybrid companies average 3.4/5 work-life balance
- The gap is 0.6 points — statistically significant across 23 vs 95 companies
The remote-first companies with the highest WLB scores: Grafana Labs (4.3), PostHog (4.5), Linear (4.8), Supabase (4.5). These companies don't just offer remote work — they've built cultures that structurally protect work-life boundaries.
The office-first companies with the lowest WLB scores: Cohere (2.5), Scale AI (2.8), several others below 3.0. These tend to be fast-growing companies where in-office culture correlates with longer hours and always-on expectations.
The "Remote-Friendly" Trap
The most misleading category is "remote-friendly" — companies that list some remote roles but aren't structurally remote. Here's what that typically means in practice:
- Most roles require 2-3 days/week in a specific office
- Remote workers exist but are at a career disadvantage (missed promotions, less visibility)
- Meetings default to synchronous, scheduled during office-hours in one timezone
- Leadership and executives are almost entirely in-office
- The "remote" option can be revoked with a policy change (as Amazon demonstrated in 2024)
Companies we've explicitly declined to label as "remote" despite their marketing: Stripe ("remote role available" but primarily SF/Seattle), Databricks (hybrid), Figma (hybrid), OpenAI (SF-centric). All of these have some remote employees — but the default is in-person and the culture reflects that.
What This Means for Job Seekers
If remote work is genuinely important to you — not just "nice to have" but a requirement — here's how to evaluate companies honestly:
- Look at actual job listings. If 80%+ of roles specify a city name with no remote option, the company is not remote regardless of what the careers page says.
- Check where leadership sits. If the CEO, CTO, and VPs are all in one office, remote workers will always be second-class citizens.
- Ask about promotion rates. In your interview, ask: "What percentage of your director+ leaders are fully remote?" The answer reveals everything.
- Read employee reviews for "remote" mentions. If remote employees write reviews saying they feel disconnected or out of the loop, that's the real signal.
- Prefer founded-remote over converted-remote. Companies that were always remote (Grafana Labs, GitLab, Automattic) have deeper distributed DNA than companies that converted during COVID.
Browse Remote-First Jobs
Filter for companies that are genuinely remote-first — not just "remote-friendly."
Remote Jobs → Remote-First Companies →The Outlook: Will More AI Companies Go Remote?
Probably not in the near term. The AI arms race creates competitive pressure that favors co-location. As long as labs are racing to train frontier models, they'll want teams physically together. The macro trend in 2026 shows 77% of new job postings across all industries are fully on-site, and AI is no exception.
But here's the counter-signal: the companies building AI infrastructure tools — the picks-and-shovels layer — are increasingly remote. Hugging Face, Pinecone, Weaviate, n8n, Supabase — these companies don't need GPU clusters. They build software products that serve developers. And for software companies, the economic and talent advantages of remote work are overwhelming: access to global talent, lower burn rate, and employees who rate their experience 0.6 points higher on work-life balance.
The AI industry will likely bifurcate: frontier research labs will stay office-first (the OpenAI/Anthropic/DeepMind model), while AI application companies and developer tools companies will increasingly go remote (the Grafana Labs/Hugging Face model). If you want to do frontier research, expect to be in an office. If you want to build AI-powered products, remote options are abundant and growing.