Office demand in 2026 is not dead. It is getting selective.
San Francisco makes that clear. While the city still struggles with elevated vacancy and the long shadow of remote work, one tenant group is moving decisively in the opposite direction. AI companies leased nearly one million square feet in the first half of 2025 alone.
This is not a broad-based recovery. It is a quality-driven one.
AI firms are not absorbing space indiscriminately. They are choosing very specific buildings, moving quickly, and paying for infrastructure that actually works. For landlords who understand what they want, this shift is creating real pricing power and shorter vacancy cycles.
The Numbers Tell a More Important Story Than Vacancy Rates
On the surface, one million square feet sounds like a footnote in a market this large. In reality, it signals a structural change.
AI tenants are not filling leftovers. They are competing for buildings that meet strict criteria, often bidding against each other for a limited subset of inventory. The result is a bifurcated market where average statistics look weak, but the right assets are quietly stabilizing and even improving.
The demand is narrow but deep.
AI companies tend to want larger floor plates, immediate availability, and buildings that can support serious infrastructure loads. If a property does not meet those needs, it may as well not exist to them. If it does, it suddenly sits in a very different leasing universe.
Infrastructure Is Now the Rent Driver
For decades, tenant improvements focused on finishes, layout, and cosmetic upgrades. AI tenants have shifted the conversation entirely.
Power density is now one of the first questions asked. Many AI firms require three to five times the electrical capacity of a traditional office tenant. That impacts transformers, panels, risers, backup systems, and HVAC capacity to handle heat loads that standard office buildings were never designed for.
Security has also moved from a nice-to-have to a baseline requirement. AI companies work with sensitive data and proprietary models. They expect controlled access, robust visitor management, and in some cases dedicated secure areas within the building.
Connectivity is equally non-negotiable. Redundant fiber, low latency connections, and sufficient building backbone capacity matter far more than how nice the lobby looks. Buildings that cannot support these requirements are filtered out early.
For landlords, this means capital decisions now directly determine tenant relevance.
Speed to Occupancy Is a Competitive Weapon
One of the most overlooked advantages in attracting AI tenants is speed.
Many AI companies cannot wait twelve to eighteen months for a custom buildout. They often need space in three to six months, sometimes faster. Buildings that can deliver power ready, connected, and secure space quickly win deals even if they are not perfect in every other respect.
Landlords who pre invest in AI ready infrastructure, spec suites, and experienced contractor relationships consistently outperform those who wait for tenant driven upgrades. Speed reduces friction, shortens downtime, and justifies premium rents.
In this segment, time is value.
What This Means for Landlords and Investors
The rise of AI tenants reinforces a broader reality in commercial real estate. Generalist strategies are losing ground. Specialization is winning.
Not every building should chase AI tenants. But assets near talent pools, transit, and innovation corridors may justify targeted investments that reposition them into a high demand niche.
Understanding the local AI ecosystem matters. Training heavy operations, application focused teams, and research groups all have different space and infrastructure profiles. The landlords who succeed are not guessing. They are building intentionally for a defined tenant type.
The Takeaway
If you own quality office space near talent and transit, prioritize power capacity, flexible floor plates, secure infrastructure, and fast delivery. Aesthetics still matter, but they no longer lead the decision.
The AI tenant trend is not really about artificial intelligence. It is about serving capital intensive, high growth businesses with precision.
Landlords who adapt will see stronger leasing, better tenants, and improved returns. Those who do not will continue competing in a shrinking pool.