Ultra Warehouse Robots Raise $62M for Monthly Rentals

A Brooklyn robotics startup is betting that the robots which pay off first will not look anything like the ones in the movies. Ultra Warehouse Robots are the dullest-looking machines in the industry, and that is precisely the point. Ultra, a company that builds dual-arm packing robots and rents them to warehouses by the month, announced on October 9, 2026 that it has raised a $50 million Series A led by Framework Ventures, according to Fortune's Term Sheet reporting as covered by AI Weekly.

The new round brings Ultra's disclosed funding to $62 million. It follows an earlier $12 million seed round that Y Combinator and NextView Ventures led, with Y Combinator participating again in the Series A, according to Fortune's account of the company's announcement.

Ultra's model is a deliberate break from the industry's standard playbook. Instead of selling expensive machines to warehouse operators, the company charges an upfront integration fee to install its robots and then bills a recurring monthly fee for hardware and software support. Ultra CEO Jon Miller Schwartz told Fortune that traction has been strong enough for the company to raise its prices.

The machine doing the work is a stationary dual-arm robot Ultra calls Operator, or OP1. According to reporting by Humanoids Daily, the robot stands on locking caster wheels so it can be repositioned between workstations, works within a five-by-five-foot footprint, and handles payloads of up to 10 pounds per arm on a standard 120-volt outlet. Its listed jobs are packing, sorting, and kitting.

Ultra says its fleet has packed more than 500,000 orders across customer sites in New York, Georgia, New Jersey, and Texas. At Highline Commerce in Brooklyn, the company's robots handle up to 30 percent of fulfillment volume, according to AI Weekly's coverage of the announcement.

Brains rented from a $5.6 billion AI lab

The most interesting part of the announcement may be who supplies the intelligence. Ultra builds and installs the robots, but it relies on Physical Intelligence to supply the AI models that let each unit learn and improve in response to a given customer's setup, according to Fortune.

Physical Intelligence is the robot-AI company founded by a team out of Google DeepMind, carrying a $5.6 billion valuation. According to AI Weekly, the firm announced a deepening tie-up with Ultra alongside the funding, with Physical Intelligence's policy stack sitting behind what AI Weekly describes as a learned fleet.

The division of labor is clean. Ultra owns the hardware business and the warehouse relationships. Physical Intelligence supplies the perception and manipulation models. And Physical Intelligence gets something it cannot fake in a lab: hours of real orders involving real, messy items moving through actual fulfillment workflows, according to Bad Signal's analysis of the deal.

That structure also lets Ultra sidestep the hardest part of robotics. Building a foundation model for manipulation costs a fortune. Renting one lets a young hardware company spend its capital on manufacturing, installation, and service instead of competing with well-funded AI labs on model training.

A counterweight to the humanoid hype

Schwartz is pitching the round as proof that unglamorous robots outperform famous ones. Fixed industrial systems are having more real-world impact than humanoid robots, which receive most of the attention, he told Fortune. It is a pointed argument. Humanoid builders are prone to demo-day stumbles and slow real-world deployment, while a box-packer bolted to a warehouse floor quietly stuffs padded mailers at 3 a.m.

The bet on recurring revenue matters too. Selling a robot means asking a warehouse operator to sign a big upfront check. Renting one by the month, like a forklift, removes that barrier, according to Bad Signal. Ultra carries the hardware risk instead: if machines break or sit idle, the cost lands on the company's books rather than the customer's.

There are real caveats. Ultra has not disclosed revenue figures, so the size of the actual business is unknown. The 500,000-order count is the company's own number, and the reporting puts no dollar figure on revenue and no count on installed robots, according to AI Weekly. Physical Intelligence remains pre-revenue with no public pricing, so the depth of that partnership is structured as individual contracts, according to Tech Holler's review of the announcement.

Still, the pattern rhymes with the broader robotics capital cycle. Readers may recall my earlier report on Mecka's $60M Series B to teach robots human motion, which chased the same thesis: real-world deployment as the moat. The difference is that Ultra is monetizing deployments now, through service fees, rather than building toward a platform later.

Why the service model matters for embodied AI

The robotics industry has a deployment problem that pure software does not. AI agents in the browser can scale to millions of users overnight. A robot has to be manufactured, shipped, installed, maintained, and eventually repaired. That is why the robots-as-a-service model keeps attracting capital: it converts a capital-expenditure sale into an operating-expense subscription, which most logistics companies prefer. That subscription math is the Ultra Warehouse Robots insight in one line: renting a robot turns a one-time sale into a compounding relationship.

It also changes who bears the technical risk. When the vendor owns the hardware and the AI keeps improving through fleet data, each installation makes the next one cheaper and more reliable. That flywheel is the promise of the tie-up with Physical Intelligence. Each Ultra warehouse feeds deployment data back into models that improve every robot in the fleet, and Ultra has indicated that approach is already letting the company raise prices, according to Fortune.

The security dimension is worth watching as well. As I reported earlier this month, Rein Security raised $25M for AI agent runtime security, and fleets of rented, cloud-connected robots will need exactly that kind of tooling once they scale. A robot that learns from every site it touches is also a robot that needs hardened, monitored AI runtimes.

Whether a boring box-packer rented by the month can out-earn the humanoid hype cycle is the $62 million question. Ultra's answer is that the next wave of AI money landing anywhere physical will reward the machine that ships boxes on a subscription. Investors at Framework Ventures, at least, are willing to find out.