A Teachers’ Pension Now Backs EliseAI at $4 Billion

EliseAI closed a $350 million round at a $4 billion valuation on September 29, 2026. Andreessen Horowitz and Bessemer Venture Partners led, and Ontario Teachers’ Pension Plan joined as a new investor.

CEO Minna Song wrote the same day that the New York company’s tools already reach 6.5 million U.S. apartment units and 30 million unique renters. The people writing the checks are funds. The people on the line are tenants and the offices that used to answer them.

Ontario Teachers Now Sits Behind the Leasing Phone

Song said existing backers wanted to co-lead after a year up close with the company. Sapphire Ventures and Navitas Capital joined again. The round was all primary capital, she said, with no early holders selling stock in the deal.

Ontario Teachers’ is the new name in the cap table. The pension, which invests for 346,000 working members and pensioners, reported $303.2 billion in net assets as of June 30, 2026, and a 9.5% net return for the first half of the year. Its Teachers’ Venture Growth sleeve held $25.9 billion across 45 companies, a book that already includes Anthropic, Databricks, and SpaceX.

Four days before the EliseAI announcement, the same pension said it was putting money into Danish rental housing through a joint venture with H Capital. It now holds a stake in apartments abroad and in the American software that answers leasing and maintenance calls. Song has said a16z and Bessemer have now funded EliseAI four times since 2023.

ELISEAI’S LAST THREE ROUNDS

The new money sits about 13 months after the Series E. Song said it will pay for product work and for more engineering, deployment, and sales staff across North America, plus a second engineering hub in San Francisco beside the New York headquarters.

Six and a Half Million Units, One Vendor

Song founded EliseAI in 2017 with Tony Stoyanov, who is the company’s chief technology officer. She has described a stretch working at a real estate firm, before any code, as a hunt for the one bottleneck that made housing expensive to run. The first two years, she has said, brought in no venture money.

The bottleneck she named was the phone-and-inbox grind of leasing, renewals, maintenance, and collections. On announcement day she put hard counts on how far that work has spread.

ELISEAI AT SCALE

• Units live: Song wrote that 6.5 million units are on the platform, which she called one in five U.S. multifamily apartments.
• Unique renters: She put unique renters reached at 30 million.
• Recurring revenue: The company said it passed $200 million in annual recurring revenue in June 2026, the fifth straight year that figure doubled.
• Call volume: Song has said the housing and health-care businesses together handle about 5 million calls a month.

Company materials also describe training on more than 300 million resident and prospect interactions and more than 450 product and feature releases a year. Those are EliseAI’s figures, and they are how the firm argues it is no longer a leasing chatbot sitting beside a property-management system.

The renter does not buy this software. The operator does. That is why a pension that also buys buildings has a cleaner fit than the affordability language around the round first suggests.

Apollo Does the Work the Office Used to Do

EliseAI unveiled Apollo on September 3, 2026, and now sells it as one agent for property management. The pitch is a single teammate that can take any action already available in the Elise platform, from the leasing desk to the executive suite, without a new setup.

Song used a blunt contrast in the launch statement.

Every property team already has a dozen dashboards and reports telling them what happened. Apollo tells you what to do next, and then does it.

Minna Song, cofounder and CEO, Apollo launch statement

The same release walked through jobs that used to eat a shift. A leasing agent can ask Apollo to reschedule every tour for a colleague who is out sick. A maintenance tech can ask which units keep throwing the same work orders. Because the agent inherits each user’s existing permissions, it only sees and does what that person is already cleared to do. When it does not know something, the company says, it is built to say so instead of guessing.

WHAT THE AGENT IS SOLD TO FINISH

• Leasing desk: Inquiries, tours, follow-ups, and messages to everyone who walked a unit the day before.
• Resident line: After-hours questions, renewal outreach, and delinquency contact that used to sit on a coordinator’s queue.
• Maintenance: Work-order assignment, recurring-issue flags, and faster unit turns that feed the next lease.
• Office stack: Reports, settings questions, and tasks that previously meant hopping across dashboards.

Kate Scharff, owner of First Montgomery Group, said in that launch that Apollo is “the smartest AI I know” at pulling data, asking about settings, and pointing to revenue the office might be missing. Song’s longer note the morning of the round went further: in-house models for jobs frontier systems still miss, smart-lock tours that need wayfinding through a real building, and affordable-housing files that can run 200 pages with no room for error.

She posted the round herself that morning, with the unit counts and the operator results attached.

https://t.co/TJMNlmukMB

— Minna Song (@minnasong) September 29, 2026

The hottest AI trade in this corner of the market is less admin, not a smarter general chatbot. Apollo is built to empty the queue. The open question is whose day gets lighter, and whose job is simply not refilled.

The Savings Show Up as Landlord Income First

Song’s public case is that housing cannot get cheaper for tenants unless it gets cheaper to run. She wrote that a typical U.S. apartment now costs more than $600 a month just to operate, on processes that have barely moved since on-prem software arrived in the early 1980s, and that EliseAI’s stack already touches 58% of a typical operator’s profit-and-loss statement.

The proof she offered with the new valuation is not a rent cut. It is occupancy, renewals, and net operating income.

WHO THE ANNOUNCED GAINS ACCRUE TO

Congressional Research Service estimates for 2024 put severe housing cost burden at 67.3% of renter households earning under $30,000. Among all renters, 49.4% were cost-burdened, or 22.7 million households, and 26.2% were severely burdened, paying more than half their income for housing.

Song argues AI can reverse the long drift that made TVs cheap and apartments dear, because the model does the work rather than sitting on a shelf. That only holds if operators spend the extra occupancy and the extra NOI on lower rents or more units, instead of on margins that have been chewed up by insurance and labor. The term sheet does not require that pass-through. The customer is the landlord.

What Binding Decisions Still Require a Person?

Song has been careful on the legal edge of this product. Anything that ends in a binding decision, she has said, still goes through a person: approving or denying an application, sending a formal notice, signing off on a lease term. Fair housing calls are in that bucket. Apollo will draft a response, flag the relevant policy, and queue it. It does not get the final say.

She has also said the agent only ever acts inside the permissions of the person using it, and that the higher the stakes, the more a human checkpoint is built in. That is the renter’s remaining protection in a stack that already drafts the message, books the tour, and dispatches the work order.

It is also a map of which jobs shrink. Drafting, routing, after-hours coverage, and first-pass screening are the layers Apollo is sold to swallow. Judgment, legal sign-off, and the human touch Song still mentions are what stay. Property firms that have talked publicly about the older Elise tools have described payroll savings from leaving vacated leasing seats empty rather than from mass layoffs. That pattern, if it holds, is how 6.5 million units get cheaper to run without a pink-slip headline.

The Same Playbook Is Running in Exam Rooms

EliseAI took the housing workflow into specialty physician groups after it had already wired leasing offices. The health-care line automates the patient path from the first inbound call through referrals, scheduling, insurance checks, chart prep, and follow-up. Song has said the company is one of OpenAI’s larger deployments at scale, used to test new voice tools quickly.

KFF polling has found that 41% of U.S. adults carry some form of medical debt, including balances on cards, in collections, or owed to family. Song groups housing and health care together as more than 40% of what the average American household spends, and treats both as places where making delivery cheaper is the only way to make the bill cheaper.

The paying customer is again the operator, a specialty practice under labor strain, not the patient on hold. Sameer Dholakia, the Bessemer partner who is joining EliseAI’s board with this round, called the housing work a mix of AI research and a detailed grasp of how properties actually run, and said that mix has produced “deep customer love.” The customers in that sentence are the firms that buy the seats.

Hiring Across Six Cities With No IPO Date

Headquarters is a 109,000-square-foot office in Manhattan’s former Tiffany & Co. building on Fifth Avenue, occupied in the summer of 2026. Song said the new capital will open a second engineering hub in San Francisco and add people in New York, Boston, Chicago, Austin, and Toronto.

On September 24, 2026, five days before the round, EliseAI said it was teaming with DOOR so renters can unlock units themselves for self-guided tours and some maintenance visits. That is the physical end of the same idea Apollo represents on the screen: fewer people in the office, more of the building running on permissions, keys, and queued human sign-off.

Asked about a public offering, Song would not set a clock. She said the company is trying to build the best business it can, with no specific timelines or definitive outcomes for every stakeholder in the cap table.