San Francisco's Office Market Found a Pulse in 2025. What's In Store for 2026?

For most of the post-pandemic stretch, San Francisco’s office story has been simple: too much space, not enough certainty. In 2025, that story finally started to bend.

Preliminary CBRE data points to 10.2 million square feet of office leasing signed in 2025 (including renewals), the city’s highest total since 2019. A striking share of that activity came from AI. Roughly a quarter of leased space involved AI companies, and AI accounted for more than 80% of newly leased space, according to CBRE’s early read.

That does not mean San Francisco’s office market is fixed. A 33.5% vacancy rate is still historically extreme. But vacancy moved in the right direction, slipping from 34.4% in Q3 to 33.5% later in the year, and CBRE described the annual decline as the largest since 2011.

The important shift is not that the market is healthy again. It is that the market is now directional, and for the first time in years, the direction is up.

What changed in 2025: demand became real again

A “recovery” in office is often just a set of renewals that pads totals without truly improving street-level fundamentals. 2025 was different because the narrative had identifiable engines, and they signed meaningful deals.

Two examples captured the tone:

• Sierra AI took about 257,000 square feet at 185 Berry Street, the biggest new lease of the year in the CBRE tally.

• Nvidia leased around 45,000 square feet at Mission Rock for its first San Francisco office, a signing that matters as much for signaling as for the square footage.

And there is more behind the scenes. CBRE’s reporting also noted AI companies are pursuing another 2.8 million square feet, roughly a third of all active demand.

Our hypothesis: this is not a broad office comeback, it is a “prime-only” rebound

Here is the pattern we think is emerging. San Francisco is not healing evenly. It is splitting into two markets:

1. Trophy and highly upgraded buildings that fit the current buyer of office space (AI or adjacent tech, top professional services, well-capitalized firms)

2. Everything else, where vacancy, discounting, and uncertainty linger

CBRE’s own market commentary highlights that while overall vacancy remains very high, the best space is far tighter. The Chronicle report framed it this way: more than a third of SF office is vacant, but only about 15% of trophy space is vacant, which is why developers and owners are watching the top of the market first.

If you believe this bifurcation is real, it explains a lot:

• Why leases cluster in places like Mission Bay, China Basin/South Beach, and newer SoMa nodes where buildings are infrastructure-ready.

• Why some corridors still feel empty even while headline leasing improves.

• Why owners of commodity buildings increasingly view conversion, recapitalization, or “sell the entitlement” strategies as the only rational path.

A prime-led recovery can still be economically meaningful, but it does not magically “lift all buildings.” It mostly lifts the spaces that the market would build today if it could start from scratch.

Why AI companies are behaving differently than the last tech boom

During the 2010s, a lot of tech hiring was mass hiring. Headcount grew fast, teams became large, and office footprints followed.

In this cycle, AI companies can be revenue-strong without being headcount-heavy. The Chronicle reporting makes this point directly, noting that the job market has not surged like the last boom, partly because AI firms often run leaner.

So why lease at all? Our read is that AI leasing is driven by a few practical realities:

• Security and privacy constraints: Enterprise AI work often wants controlled environments, not purely distributed setups.

• Speed and coordination: The “iteration loop” is faster when core teams are physically together, especially for product, go-to-market, and partnerships.

• Talent density: SF still concentrates a specific mix of builders, investors, and customers. AI firms treat that density as a strategic asset.

• Brand and recruiting: A real office in a high-signal location is still a magnet for senior talent and partners.

RTO is still weak, but policy is shifting

One reason the office market has struggled is simple: usage never came back. The Chronicle report puts the RTO rate around 45% of pre-pandemic levels, and it has been stubborn.

What changed in 2025 is that companies appear to be moving away from fully remote as a default. Not necessarily five days a week, but toward hybrid norms that require real space again.

Two anecdotes from the same CBRE and Chronicle coverage matter here: Brex and Coinbase, both of which had pulled away from SF during the pandemic era, signed new leases to return. When companies start re-anchoring, even selectively, it changes how the next wave of companies perceives the city.

The “2026 setup”: more demand, but also real constraints

If 2025 was the inflection year, 2026 is the proof year.

The good news: CBRE’s tech insights view is that there is “real growth” behind some of these AI tenants, and that ongoing AI investment plus improved tourism could support sustained positive job growth. The risk: the recovery can still be derailed by a few structural problems.

1. Transit finances and commuting friction

If BART and Muni face steep fiscal constraints, service cuts or instability can directly suppress office utilization. Even a strong leasing year does not fully matter if commuting becomes unreliable.

2. Housing affordability

The city can sign leases without creating enough housing, but it cannot scale a durable jobs recovery without it. In practice, housing cost is the “tax” on hiring locally.

3. A high vacancy rate that will take years to work down

Even after the 2025 improvement, 33.5% vacancy is a large overhang. If this remains a prime-only rebound, the vacancy reduction could be uneven, leaving older buildings stuck.

What we will be watching next (and what it would mean)

If you want to know whether 2025 was a one-off or the start of a cycle, these are the signals that matter most:

1. Net absorption, not just leasing volume
   Leasing totals can be inflated by renewals. Net absorption tells you if the city is truly filling space. CBRE’s Q3 snapshot showed positive net absorption at that point, which is a constructive sign.

2. The “2.8M sf pipeline” converting into signed leases
   If AI companies are truly seeking ~2.8M sf, watch how much becomes real deals, and in what submarkets.

3. Trophy vacancy tightening further
   If trophy space tightens meaningfully from already lower levels, it can trigger two downstream effects: rent stabilization at the top end and selective new construction planning.

4. Conversions and “pivot projects”
   San Francisco’s next phase may involve fewer pure office starts and more mixed-use, conversions, or office-to-residential pivots in certain districts. If the prime market tightens while commodity space remains stranded, conversion economics improve.

Bottom line

San Francisco did not “solve” its office market in 2025. But it did something that matters more: it demonstrated that demand can return, and that AI-driven growth can translate into real occupancy, not just headlines.

Our bet is that the city is heading into a multi-year unwind where the best buildings stabilize first, and the rest of the inventory is forced into reinvention, recapitalization, or conversion. If AI hiring and revenue growth continue, and if the city avoids self-inflicted wounds on transit, permitting, and public safety, 2026 can look like a real continuation of the 2025 momentum.