Together AI To Establish New HQ In San Francisco Design District

Together AI leased 150,000 SF at 2 Henry Adams St. for a new San Francisco headquarters, expanding well beyond its original 3,000 SF footprint and later 30,000 SF space at 251 Rhode Island St. The company recently closed a $305M Series B at a $3.3B post-money valuation and is reportedly considering another roughly $1B funding round. The article underscores continued strength in Bay Area AI office demand, with OpenAI also taking 202,000 SF in Richmond and AI firms now occupying 8.75M SF in San Francisco.

Analysis

This is less a one-off leasing headline than evidence that AI capex is becoming a durable demand engine for urban office and specialized industrial real estate. The second-order winner is not just the landlord base; it is the ecosystem of buildout vendors, power/cooling contractors, network providers, and local service firms that get pulled in as these tenants scale from “startup space” to quasi-institutional campuses. That tends to re-rate best-in-class San Francisco Class-B assets with power, floorplate flexibility, and transit access, while leaving commodity offices without these attributes structurally impaired.

The more interesting signal is the power-intensity bifurcation inside AI. Traditional model training and research teams are office-light, but robotics and physical-AI workflows pull in industrial-like power and occupancy requirements, which makes proximity to the city less about prestige and more about talent-density plus utility density. That should extend leasing demand beyond Mission Bay/Downtown into fringe submarkets with better amperage, and it creates an embedded option value for owners who can upgrade electrical infrastructure over 12-24 months.

The market is likely underestimating how much of this is a forward indicator for the Bay Area’s office absorption cycle rather than just a headline about one tenant. If AI fundraising stays open, rent pressure in the few functional buildings should intensify quickly, but the setup is fragile: a funding market freeze, slower model monetization, or a shift toward remote-first AI operating norms could stall absorption within 1-2 quarters. The contrarian risk is that investors extrapolate current demand into every office asset, when only a narrow set of power-capable, design-district-adjacent properties can actually capture the flow.

Key Decisions for Investors

• Long select San Francisco office REIT/owners with power-capable, boutique industrial-adjacent assets; pair against broader office exposure. 6-12 month horizon, looking for 10-15% relative outperformance as AI tenant demand concentrates in scarce inventory.
• Initiate a barbell pair: long industrial REITs with power-advantaged Bay Area exposure, short commodity office names lacking upgrade optionality. Risk/reward favors a 2:1 payoff if leasing spreads widen over the next 2-3 quarters.
• Buy call spreads on local electrical infrastructure / data-center-enabling supply chain names for 6-9 months. The thesis is not generic AI spend, but rising demand for electrical upgrades, switchgear, and cooling retrofits as office-to-AI hybrid requirements increase.
• For public REITs with Bay Area footprints, buy dips only on assets with clear power and transit advantages; avoid chasing broad office beta. The trade should be sized as a relative-value position, not a directional macro bet.
• Watch private funding data for AI names over the next 30-60 days; if the rumored mega-round closes, add to long exposure in scarce San Francisco power-constrained assets, as financing would extend leasing demand into 2026.