What does OpenAI's reported workforce expansion mean for office investors? It is a demand signal to investigate, not a formula for converting employees into leased space. In March 2026, CNBC reported the Financial Times' hiring story. Reuters also relayed that report: OpenAI planned to increase its workforce from about 4,500 to 8,000 by year-end. That was a reported plan, not confirmation that the hiring target had been reached. Office-market evidence needs to be assessed separately. For the broader tooling context, see our guide to AI tools for commercial real estate.
Source update: This article preserves the March 2026 hiring story as a dated report. Market figures below identify their own periods. Unsupported vacancy, leasing-share and AI-specific rent-premium claims from the original version have been removed or replaced with attributed market evidence.
Key Takeaways
- The reported hiring target was 8,000 employees by the end of 2026. It does not establish where those employees would work or how much additional office space they would need.
- Reported office commitments, occupied space and future expansion options are different measures. Do not combine them into a square-feet-per-employee forecast.
- Compare office-market evidence by period and property segment before applying it to an individual building.
- CBRE's 62% San Francisco figure measures AI companies' share of tech-tenant demand, measured in square feet, not AI's share of all completed leases.
- Underwrite the building, executed lease and tenant credit separately. The cited research does not isolate a rent premium caused by AI tenancy.
OpenAI's Hiring Plan and Office Footprint Are Separate Facts
The workforce report implied roughly 3,500 planned additional roles, mainly across product development, engineering, research and sales. Reuters attributed it to the Financial Times and noted that Reuters could not immediately verify the report. Neither this hiring target nor a company valuation supplies a building-level occupancy schedule, a credit guarantee or evidence of achieved headcount.
On the property side, The San Francisco Standard's April 7 office survey estimated OpenAI's San Francisco footprint at 1.2 million square feet. Its account identified space around Uber's former Third Street headquarters, 550 Terry Francois Boulevard and 1800 Owens Street in Mission Bay. The account included space expected to be occupied later, so the estimate should not be read as fully occupied space on the report date. This is an attributed newsroom estimate, not a company-certified portfolio total.
For a prospective lease, request the actual premises, delivery schedule, current occupancy and any rights over additional space. A global hiring plan and a city-level footprint cannot establish future demand for a particular property. This distinction removes the need for an assumed square-feet-per-employee ratio.
What the Office-Market Evidence Shows
Cushman and Wakefield's Q1 2026 San Francisco MarketBeat reported 31.6% vacancy, down from 34.3% a year earlier, and nearly 900,000 square feet of net absorption. It also identified OpenAI's 282,000-square-foot sublease at 1800 Owens Street. These are period-specific observations, not a forecast that every San Francisco office will recover.
CBRE's May 2026 Tech Gateway Office Markets report put AI companies' share of tech-tenant demand, measured in square feet in San Francisco at 62%. It recorded 10.6 million square feet of AI leasing in San Francisco and 10.4 million in Silicon Valley from 2019 through Q1 2026, together 66% of the six largest AI markets. CBRE's leasing measure includes its defined AI-company sample; it is not a tally of all office occupiers.
An earlier CBRE brief from May 2025 projected up to 16 million square feet of San Francisco AI-company leasing through 2030. That is a dated forecast, not space already leased or a commitment by OpenAI. The brief also explains that earlier technology cycles ended with companies giving back surplus space. A demand scenario belongs alongside a downside scenario.
The distinction for underwriting is between regional activity and property-specific evidence. A city can attract expanding AI occupiers while an individual building remains poorly located, costly to adapt or unsuitable for the tenant's plans. Review recent comparable transactions and current competing availability for the asset itself.
What to Ask a Prospective AI Office Tenant
The following is a diligence checklist, not a claim that every AI company needs the same building:
- Actual use: Establish whether the premises will house ordinary office work, laboratories or computing equipment. Have qualified engineers assess the electrical and cooling needs of the proposed use.
- Connectivity: Ask the tenant to specify bandwidth, redundancy and service requirements; confirm what the building can deliver and who pays for upgrades.
- Layout and occupancy: Reconcile the proposed seating plan, collaboration areas and occupancy phases with the space being leased.
- Expansion rights: Price options, rights of first offer and reserved space separately from the initial committed premises.
- Location: Test the site's recruiting, transit and operating requirements with the tenant rather than assuming that any address in an AI cluster will qualify.
The AI Consulting Network can help CRE teams structure and review these information requests. Engineering capacity, legal obligations and tenant credit still require the appropriate professional review.
Financial Implications for Office Investors
Separate Building Quality From Tenant Industry
Cushman and Wakefield's Q1 2026 report gives annual gross asking rent of $69.16 per square foot overall and $110.15 for Class A Tier 1, a reported 42.8% premium over the broader CBD market. This is a building-quality comparison, not an AI-tenant premium or effective rent after concessions.
Use comparable executed leases, concessions, tenant-improvement obligations and downtime when forming a rent assumption. An AI company signing in a premium building does not by itself establish that its industry caused the premium.
Read the Initial Lease Term and Options Separately
A large footprint does not establish a long, firm lease commitment. Review the executed initial term, commencement conditions, termination rights, renewal options and any guarantees. An option controlled by the tenant should not be treated as binding future occupancy. Do not assume that AI companies generally sign longer leases than traditional startups; the sources cited here do not establish that comparison.
Underwrite Credit and Concentration
Request financial and contractual evidence relevant to the entity signing the lease. A valuation headline or hiring announcement does not substitute for payment capacity or recourse. Where the evidence is incomplete, discuss deposits, letters of credit, guarantees and other protections with the deal's advisers.
Model what happens if the tenant does not renew, exercises a termination right or needs less space. Include downtime, reletting costs and building adaptation costs. A building concentrated in one tenant requires a different downside analysis from a diversified rent roll; an AI label does not remove that exposure.
Choose the Submarket and the Building, Not Just the Theme
CBRE's evidence gives investors a reason to investigate San Francisco and Silicon Valley, with Manhattan and Boston also present in its AI-leasing comparison. It does not establish that every property in those markets is an attractive purchase. Use the regional evidence to narrow the search, then evaluate the building's basis, competing supply, tenant fit and lease economics.
For a portfolio review, keep three columns separate: observed market evidence, tenant commitments supported by documents, and assumptions still requiring validation. This makes it harder for a credible hiring headline to become an unsupported rent or valuation assumption. CRE investors looking for help organizing that analysis can contact Avi Hacker, J.D. at The AI Consulting Network.
Frequently Asked Questions
Q: Did OpenAI confirm that it reached 8,000 employees?
A: The March story described a year-end target reported by the Financial Times and relayed by other outlets. This article does not establish that the target was achieved.
Q: How should investors interpret the reported San Francisco footprint?
A: Treat the Standard's April estimate as attributed portfolio reporting that included later occupancy. For a transaction, verify the particular premises and binding lease commitments rather than using that estimate as an occupancy forecast.
Q: How should a buyer use regional market data?
A: Use it to frame questions. Test the proposed purchase against comparable executed leases, building condition, tenant requirements and a downside model. Regional data cannot replace property-level diligence.
Q: Do buildings with AI tenants automatically command higher rents?
A: The cited reports support differences by building quality and market segment. They do not isolate a rent premium caused by the tenant being an AI company. Underwrite the space and lease economics directly.
Q: What is the key leasing risk?
A: Distinguish a hiring plan from binding rent obligations. Review the signing entity, initial term, options, concessions and recourse, then test the cost of vacancy and reletting if the tenant leaves.