The U.S. Cities facing the deepest AI-related job cuts are also the ones pulling in the most space from AI firms. That counterintuitive overlap is the headline of new JLL research, and it reframes a story that has mostly been told in broad strokes.
Working alongside MIT's Center for Real Estate, JLL finds that artificial intelligence is sorting U.S. Property markets into winners and laggards. The same metros that carry high AI-displacement risk are simultaneously absorbing the bulk of AI-tenant leasing, and the gap between adaptive and non-adaptive cities is widening.
What happened
- U.S. Tech employment fell 1.5% in early 2026, while AI-company office leasing demand continued to rebound.
- San Francisco: nearly 30% of total leasing since 2025 has come from AI firms.
- AI exposure risk and AI leasing demand are concentrated in the same markets.
- Adaptive capacity, not exposure level, is the leading indicator of real estate performance, per JLL.
- Research conducted in partnership with MIT's Sloan School of Management and Center for Real Estate.
Background and the numbers
The simple read on AI and real estate has been one-directional: AI takes jobs, demand softens, rents fall. JLL's data complicates that read by showing that the metros facing the highest AI-driven job dislocation are simultaneously pulling the largest share of AI-company leasing. Demand is not disappearing; it is relocating.
San Francisco is the clearest case in point. Nearly 30% of total leasing there since 2025 has come from AI companies, according to JLL, while the metro carries among the highest exposures to AI-driven job displacement in the U.S. The local talent pool, infrastructure, and stock of quality real estate is what is pulling in that demand.
The general principle follows JLL's framing: a market's ability to adapt and redeploy its workforce matters more for real estate performance than raw exposure. Markets with the right mix of talent, infrastructure, and quality real estate will capture the demand; markets without that mix will absorb the displacement without the offset.
| Indicator | Adaptive markets (e.g. San Francisco) | Non-adaptive markets |
|---|---|---|
| AI-tenant leasing share | ~30% of total leasing since 2025 | Low, single-digit share |
| AI job-displacement exposure | High | High |
| Net office demand effect | Positive | Negative |
| Winning factor | Adaptive capacity | None |
Our analysis
This is a useful correction to the simple "AI kills office" narrative. Office demand is not monolithic; it splits between firms whose workers are being displaced and firms whose workers are being hired. The latter group is concentrated in a handful of adaptive metros, and that concentration is the real story.
For landlords in those adaptive markets, the read-across is direct. AI-tenant demand is pulling rents and pushing concessions tighter in the right submarkets. For landlords in markets without that mix, the same data is grim: AI-driven displacement is a headwind without the offsetting demand. Different playbook. The market will bounce.
For investors, the differentiation cuts across asset classes. Industrial and logistics benefit indirectly from the build-out of AI infrastructure, including data centers and power-hungry grid upgrades. Retail in adaptive markets benefits from the spending power of AI-hire cohorts. Office in non-adaptive markets is the most direct loser, and the gap between adaptive and non-adaptive office is widening quarter by quarter.
The practical filter for underwriting is straightforward. Ask whether a metro has the talent, infrastructure, and quality real estate base to attract AI-tenant demand. If yes, the AI-displacement risk is a manageable side effect. If no, the displacement risk is the whole story.
Worked example
Say you are a landlord deciding between two office acquisitions: a 200,000-square-foot vintage asset in San Francisco's SoMa submarket, and a similar asset in a Midwest metro with limited AI-tenant pipeline. JLL's data says SoMa can absorb AI-tenant demand even as the city carries displacement risk; the Midwest asset has neither. Underwrite to the differential. The SoMa asset can command AI-tenant lease renewal at premium rents; the Midwest asset relies on legacy demand. The cap rate should reflect that.
What to watch next
Bottom line: AI demand is concentrated, not diffuse, and the concentration is widening. Watch three signals into the next JLL update. First, the share of AI-tenant leasing in adaptive markets outside San Francisco, including Austin, Boston, and Seattle. Second, the pace of overall tech employment in those metros versus non-adaptive markets. Third, the spread between asking rents in adaptive AI-tenant submarkets and the broader market. Together, those will tell you whether adaptive markets are widening their lead or whether non-adaptive markets are starting to absorb some of the demand.
Key takeaways
- AI firms drove nearly 30% of San Francisco leasing since 2025, even as AI job-displacement risk runs high.
- Adaptive capacity, meaning the mix of talent, infrastructure, and quality real estate, is the leading indicator of real estate performance, per JLL.
- The AI-versus-office narrative is binary; the real story is a widening gap between adaptive and non-adaptive markets.
- Underwriting should price adaptive capacity into cap rates, not just tenant credit.
Frequently asked questions
What share of San Francisco leasing has come from AI companies?
Nearly 30% of total leasing in San Francisco since 2025 has come from AI companies, according to JLL's research.
How did U.S. Tech employment move in early 2026?
Overall U.S. Tech employment declined 1.5% in early 2026, even as office leasing demand from AI companies continued to rebound.
What does JLL say matters most for real estate performance in the AI era?
JLL says a market's adaptive capacity, the mix of talent, infrastructure, and quality real estate, matters more than the level of AI exposure.
Who conducted the research with JLL?
JLL conducted the research in partnership with MIT's Sloan School of Management and Center for Real Estate.
