Things changed. Widespread remote work shifted the dynamic. Companies realized they could hire talent anywhere. Engineers saw their money go further outside California. Competition for talent became national, not local.
Looking for technical talent?
RFS specializes in technical recruiting — placing software engineers, ML engineers, and product leaders at high-growth startups.
Work with us → Browse open rolesSan Francisco still offers density. Of startups. Of VC money. Of a certain kind of network. But the strict salary separation erodes. It's a supply and demand issue. Supply of talent expanded. Demand for SF office presence decreased.
Over the last 30 days, we tracked 200 software engineer roles. The median base salary across these roles was $190K. The 25th percentile sat at $160K. The 75th percentile reached $242K. These numbers include both SF and remote positions.
Here are our projections for median base salaries in 2026. This focuses on AI-native startups.
| Role (2026 Median Base Salary) | San Francisco (Projected) | Remote (Projected) | SF Premium (%) |
| Staff Software Engineer | $265,000 | $245,000 | 8.2% |
| Senior Software Engineer | $215,000 | $200,000 | 7.5% |
| Machine Learning Engineer | $235,000 | $220,000 | 6.8% |
| AI/ML Research Scientist | $280,000 | $270,000 | 3.7% |
| Data Engineer (AI Focus) | $205,000 | $190,000 | 7.9% |
The premium exists. It's just smaller. For highly specialized roles, like AI/ML Research Scientists, the premium is negligible. Talent is scarce. Location matters less.
Companies like Drweng often pay at the higher end. Their compensation structures are competitive. Verily operates similarly. Perplexity, Thinking Machines Lab, Clera also pay well. These companies attract top talent. They know what it takes. They pay it.
These aren't always strict SF premiums. Sometimes, they're just high-paying companies. They pay everyone well. Regardless of location. But if they require office presence, the de facto SF premium holds. Because few choose SF without higher pay.
Consider the specifics of the role. A hardware AI engineer might still see a premium for SF. Lab access. Specialized equipment. Collaboration requirements. A pure software AI engineer? Less so. The work is portable.
The biggest driver: talent availability. There isn't an infinite supply of engineers. Especially not in AI. Companies must compete. Regardless of where the engineer lives.
This means a remote Staff Software Engineer for an AI startup might make $245K. Their SF counterpart earns $265K. A $20K difference. Substantial, but less than historical gaps. The $20K difference disappears quickly. High SF rent. Local taxes. Everything adds up. The true economic benefit of the SF premium dwindles.
These companies often demand a physical presence. Or hybrid models. If they're in SF, they pay SF rates. They must. Otherwise, no one shows up. Or the quality of talent drops. This is where the premium persists most strongly. It's tied to an explicit office expectation.
But even these companies face pressure. Talent pipelines dry up. If only local talent is available, hiring slows. The choice becomes stark. Pay up for SF. Or open to remote. Most choose the latter. Eventually.
Companies use this in their calculations. They know engineers compare. Not just base salary. But lifestyle. Savings potential. The remote option offers a compelling value proposition. Engineers consider it.
For AI startups, this is critical. They need to attract engineers. From anywhere. Focusing only on SF talent limits their pool. It also increases their burn rate. When every dollar counts. Especially for early-stage companies.
Evaluate total compensation. Factor in equity. Bonus structures. Benefits. Then, consider cost of living. Your personal preferences. Do you want to be in the office? Or fully remote?
The market is competitive. Companies still pay well. But the location premium is mostly an anachronism. For general software engineering. For AI engineering, it's almost gone. Except for niche cases.
Tell us about your open roles and we'll start sourcing within 48 hours.