How is AI sector growth reshaping office demand, and which markets and building types are actually capturing the incremental demand?

Synthesized the NYC AI leasing report, the national CoStar leasing snapshot, the NYC talent and migration notes, Anthropic's San Francisco expansion, and the Newmark AI office-employment frame. Read against AI Infrastructure and Office Demand 2026 so this page could stay on office demand selection rather than the broader infrastructure bridge.

AI is not a broad office recovery thesis. It is a demand accelerator for a small number of trophy, urban-campus, and mixed-use office nodes that already had strong talent and quality moats, while the first-order AI real estate demand remains compute infrastructure rather than office space.

The current source stack supports four practical conclusions:

1. the national office rebound is real but still selective
2. AI office demand is concentrated in true agglomeration markets
3. AI firms are helping set rents at the top of the market, not across the average building stock
4. the secondary-market beneficiaries are character-rich mixed-use districts, not generic suburban Class A

That means AI should be used as a selection filter inside office underwriting, not as a blanket argument that the sector is back.

The page should therefore be read beside AI Infrastructure and Office Demand 2026: office captures the second-order, cluster-specific expression of AI growth; data centers, power, powered land, and supply-chain facilities capture the larger first-order physical footprint.

New York remains the cleanest proof point.

The strongest facts in the stack are:

• AI leasing pace in NYC running at double 2025's rate
• Q1 2026 volume already at half of full-year 2025
• average AI lease size rising from 16,600 SF to 34,500 SF
• rent prints of $320/SF and $327.50/SF at the top end

That is not just startup noise. It shows mature, well-funded AI tenants taking meaningful blocks in trophy product and paying record rents to do it.

San Francisco is supported more by repeated expansion and agglomeration logic than by a clean rent comp set, but the Anthropic expansion still matters because it reinforces the same pattern as New York: firms choose the deepest talent and research ecosystems, not the cheapest space.

The May 2026 Anthropic / 330 Hudson report adds a New York version of the same campus-aggregation pattern. Because the Bisnow item described a reported pending full-building lease rather than a closed lease, treat it as demand-pipeline evidence rather than a final lease comp. Its underwriting value is still meaningful: the search range and building scale show that AI tenants are now looking for 250,000+ SF blocks in Manhattan's talent-rich Midtown South / Hudson Square ecosystem. See Source: Anthropic 330 Hudson Full-Building Lease 2026.

Boston belongs in the same top-tier bucket more because of its talent stack and quality-product logic than because the current source set gives a Boston-specific AI leasing comp as strong as New York's.

The useful non-gateway rule is that AI-adjacent or tech-quality tenants in secondary cities still choose walkable, design-forward, mixed-use environments. They do not validate broad suburban office recovery.

This is why the right read-through from places like Neuhoff is "quality and identity matter at every market tier," not "AI lifts the whole Sun Belt."

Newmark's broader AI office-employment framing is what keeps this page disciplined. Office-using employment growth is still modest, and AI can displace some automatable functions even while helping better space. So the right implication is sharper quality bifurcation, not broad demand rescue.

The May 2026 ConnectCRE / Cushman & Wakefield summary reinforces that discipline: AI can lift productivity before it lifts hiring or revenue, so some occupier demand may lag even as tenants demand more flexible, higher-quality, technology-enabled environments. The NAREIM / Juniper Square survey and Henry AI Q&A add a workflow layer: AI adoption is real, but it is still more visible in internal process speed, pre-pitch work, data governance, and diligence response times than in broad office absorption. See Source: Digging Beneath the Confusion: The Future of Artificial Intelligence and Commercial Real Estate, Source: Survey: Institutional Real Estate Has AI Readiness Problem, and Source: Henry AI Q&A - Automation for CRE Deals.

Commercial Observer's 2026 State of Office Forum recap keeps the same balance: speakers described AI-linked leasing momentum in New York, but also warned about tech headcount, rapid winner-take-most shifts, and utilization uncertainty. Use it as qualitative risk-balanced support, not as a lease-volume dataset. See Source: AI Is in the Office in New York 2026.

The May 2026 podcast synthesis adds a second AI layer: AI is showing up in CRE as workflow, underwriting, capital-raising, and data-governance infrastructure before it shows up as broad office absorption. source-podcast-jf-4237-ai-in-commercial-real-estate-practical-ai-use-cases-and-the-rtcf-framewo-05d5059311ec13ccc2ad812b|Best Ever on practical AI use cases, source-podcast-jf-4222-preparing-your-business-for-ai-agents-crm-systems-and-data-strategy-with-1cd829ea842f87a7674966c9|Best Ever on AI agents and data strategy, source-podcast-how-family-offices-use-ai-data-to-evaluate-deals-investor-panel-0d1d05004f7d70461cc59a6c|Family Office panel on AI/data deal evaluation, source-podcast-episode-two-how-ai-brings-business-value-to-cre-490fb3ec5691cc55a87c986a|BOMA Suburban Chicago on AI business value, and source-podcast-the-growing-power-of-trust-in-an-ai-world-062a347071487b110ca1c264|Family Office Club on trust in an AI world all support the same caution: AI-enabled CRE firms may need better data rooms, faster diligence, and higher-trust investor communication before they need more office square footage.

The transcript / rich-episode upgrade pass sharpens the BOMA item. Its preserved Substack page frames AI's business value around building operations, predictive maintenance, energy optimization, smart-building technology, and operating efficiency, with Bridge Industrial, SITE Technologies, and Prologis innovation / operations voices. That supports the page's boundary condition: AI is currently easier to underwrite as an operations and infrastructure capability than as broad incremental office demand. If AI changes office leasing, the stronger case is through tenants demanding smarter, higher-quality, technology-enabled environments, not through a simple headcount multiplier.

The winning office expression of AI demand is:

• trophy office in real agglomeration markets
• clusterable urban product near deep talent pools
• character-rich mixed-use office in a limited number of secondary markets

The losing use of the theme is using AI headlines to justify generic suburban or midtier urban office recovery.

The AFIRE source still matters here because global capital prefers the safest and most legible parts of U.S. CRE. AI demand and cross-border capital can reinforce one another at the very top of the office stack.

• Trophy-office underwriting in New York, San Francisco, and similar agglomeration markets
• Mixed-use office selection in secondary cities with real design and talent advantages
• Distinguishing true AI demand from generic office-recovery marketing

• Commodity suburban office
• Weak midtier urban stock
• Any model assuming AI adoption translates into broad office demand growth
• Any model treating enterprise AI workflow adoption as immediate leasing absorption

• Absolute NYC AI leasing totals rather than only relative growth stats
• Better San Francisco rent and leasing evidence tied directly to AI tenants
• Whether more AI firms disclose flexible expansion rights or unusual lease structures
• Whether any secondary-market mixed-use districts produce a cleaner rent premium comp

• The NYC AI leasing data is strong but still partly relative rather than fully volumetric.
• San Francisco remains more asserted than measured in the current source stack.
• Boston is included as a logical companion market but not yet with a comparably strong AI-specific lease comp.
• Secondary-market examples remain thinner than the gateway-market evidence.

• Source: AI Tenants Double Their NYC Leasing Activity Year Over Year
• Source: U.S. Office Leasing Has Its Best Quarter in Nearly Eight Years
• Source: NYC Undisputed Champ in Attracting Tech and Finance Talent — Commercial Observer
• Source: Anthropic Grows San Francisco Footprint Yet Again
• Source: AFIRE's Gunnar Branson: U.S. CRE Continues to Be Seen as Safest Investment
• Source: Understanding the AI-Office Space Connection
• Source: JLL Research — NYC Wins Migration Competition for High-Value Talent Despite Broad Population Outflows
• Source: Anthropic 330 Hudson Full-Building Lease 2026
• Source: Survey: Institutional Real Estate Has AI Readiness Problem
• Source: Henry AI Q&A - Automation for CRE Deals
• Source: Digging Beneath the Confusion: The Future of Artificial Intelligence and Commercial Real Estate
• Source: AI Is in the Office in New York 2026

• AI Infrastructure and Office Demand 2026
• Office Bifurcation
• New York Office Capital Markets and Talent Concentration 2026
• Office Debt Markets 2026
• Distressed Office Price Discovery 2026
• Office Hub
• Analyses Hub
• United States

• Adds a small AI / marketing-tech Manhattan office lease as tenant-depth evidence in the AI-office watchlist. See source-hightouch-275-seventh-avenue-lease-2026. Caveat: Small-block lease; not a rent comp without economics.