More than half a million orders packed, and the robots doing the work haven’t needed to walk a step. Ultra Robotics is a Brooklyn startup founded in 2024 that leases its OP1 packing robots to US warehouses, with Physical Intelligence providing the AI. The company has raised $62 million, including a $50 million Series A led by Framework Ventures announced on October 9, 2026.
If you’re expecting a robot with legs wandering between shelves, this is something different. Ultra puts two robotic arms at a packing station and gets them handling outgoing orders. Here are the company basics before we get into the machinery.
| Key fact | Details |
|---|---|
| Founded and headquarters | Established in 2024 and based in Brooklyn, New York |
| CEO | Jon Miller Schwartz |
| Total funding | $62 million across seed and Series A rounds |
| Series A lead | Framework Ventures |
| Seed leads | Y Combinator and NextView |
| Robot | OP1 Operator, a stationary robot with two arms |
| AI partner | Physical Intelligence |
| Orders packed | More than 500,000, according to Ultra |
| Valuation and revenue | Neither amount has been made public |
What Ultra Robotics builds
Ultra builds robots for repetitive industrial work, starting with the packing jobs that get online purchases ready for delivery. Its OP1 Operator fits into an existing workstation, where two arms handle goods and packaging while the base stays put. To understand what you could put it to work on, start with its physical limits and current tasks.
You see, OP1 isn’t a humanoid. Workers roll its unpowered base into position, lock the wheels and connect it to power and the network. You don’t have batteries to recharge or swap, and installation doesn’t require permanent infrastructure.
The aim is useful work at the station you already have. That means reaching for items, handling mailers and getting parcels ready to leave.
OP1 Operator specs at a glance
Wondering whether one would fit beside your existing packing equipment? These are Ultra’s published specifications, with its output claim listed separately from the physical measurements.
| Specification | Published detail |
|---|---|
| Floor space | 5 × 5 feet |
| Working area | Up to 10 × 10 feet |
| Vertical reach | From floor level to 10 feet |
| Lifting capacity | Up to 10 pounds on each arm |
| Electricity | Standard 120V outlet with a peak draw of 10A |
| Network | Dedicated Ethernet connection at 10 Mbps |
| Base | Unpowered wheels with locks |
| Operating time | Designed for operation up to 24 hours a day, seven days a week |
| Claimed output | Up to 1,000 items daily for each robot |
That final number needs a little care. It is a company claim, so you shouldn’t read it as a guaranteed daily result for your own orders. The same goes for round-the-clock operation, which describes operating capability rather than a promise of uninterrupted autonomous work.
Warehouse jobs it handles now
The useful question is what happens once goods reach the station. Across the documented installations, OP1 handles several parts of order preparation.
- Packing. Places goods into packaging, then carries out sealing and labelling.
- Sorting. Separates mixed incoming items into bins containing a single item type.
- Kitting. Groups several items together in one parcel.
- Polybagging. Handles bulk bagging work using an automatic bagging machine.
- Mailer packing. Works with padded and paper mailers in different sizes.
- Warehouse software. At Highline Commerce, uses the warehouse management system’s screen for scanning, finding orders and producing labels.
- Dropped items. Reaches down to retrieve goods that fall onto the floor.
How much money Ultra Robotics has raised
Ultra has raised $62 million in total, but that money came through two rounds. If you’ve seen both $50 million and $62 million in coverage, you’re looking at the latest round and the cumulative amount. Here’s how the funding adds up.
| Funding round | Amount | Investors identified |
|---|---|---|
| Seed | $12 million | Co-led by Y Combinator and NextView |
| Series A announced October 9, 2026 | $50 million | Led by Framework Ventures, with Y Combinator participating |
| Total | $62 million | Combined funding from both rounds |
Ultra plans to put the money into engineering and getting more robots into customer facilities. Its own announcement ties the Series A to expanding the robot fleet.
What does that tell you about the business? It has raised capital and is already earning revenue, but you can’t calculate its commercial performance from the funding announcement alone. Ultra hasn’t released its revenue amount or valuation.
Schwartz has said customer demand allowed the company to charge more. That gives you another reported sign of demand, although the actual prices remain private.
The founders behind Ultra
Ultra joined Y Combinator’s Summer 2024 batch and is based in Brooklyn’s Industry City. Several founders had already worked together on 3D printing businesses before starting this company, while their backgrounds also include machine learning and robotics. Here’s who handles the main leadership roles.
- Jon Miller Schwartz, founder and CEO. His previous work includes Arena AI, Voodoo Manufacturing, Body Labs and Layer By Layer. He studied mechanical engineering.
- Max Friefeld, co-founder and COO. He co-founded Layer By Layer and Voodoo Manufacturing and later led teams at Boston Consulting Group. His education at Harvey Mudd covered computer and electrical engineering.
- Oliver Ortlieb, co-founder and CTO. He previously held the CTO role at Voodoo Manufacturing and co-founded Layer By Layer. He has several 3D printing patents and a computer science degree from Harvey Mudd.
- Chetan Parthiban, co-founder and chief scientist. He studied mathematics and robotics at the University of Pennsylvania and was among Arena’s earliest machine learning scientists.
How leasing an Ultra robot works
You pay to use Ultra’s robots through a robots-as-a-service arrangement, with an initial integration charge followed by monthly fees. That lets you bring the hardware into your warehouse without paying a large purchase cost upfront. Here’s how the process is described.
- Pay the integration fee. There is an upfront charge for bringing the system into your operation. Ultra hasn’t published the amount.
- Get the station ready. OP1 needs its working space, a suitable wall outlet and Ethernet. Ultra says installation can take hours.
- Pay the monthly service fee. The recurring payment covers the hardware and software support.
- Use the three-month trial. Ultra advertises a three-month trial period, although its website doesn’t list pricing for it.
- Add robots when you’re ready. The offer allows you to expand after trying the system in your operation.
There’s one distinction you’ll want to remember. Installation within hours doesn’t mean reliable output arrives just as quickly. At Highline Commerce, getting from introduction to dependable order filling took about three months.
Highline’s billing also changed from a charge per order to a monthly service fee intended to encourage more use. So if you’re considering a lease, ask Ultra for its current charges and trial terms. With no public price list and reported price rises, you can’t build a dependable budget from the announcement alone.
Where Physical Intelligence comes in
Ultra builds and installs the hardware, while Physical Intelligence supplies the AI that helps it handle each customer’s packing work. That gives you two companies behind the service, each responsible for a different part. Here’s how their work connects.
Physical Intelligence was founded by a team with Google DeepMind experience and has a reported valuation of $5.6 billion. Its software also serves other robot makers, so Ultra isn’t its only hardware partner. It’s part of the rise of physical AI that we’ve been following all year.
For Ultra, the partnership brings AI that can learn from warehouse tasks. Physical Intelligence receives data from those tasks to help improve its models. As robots encounter more packing situations, those experiences can feed into later training.
You may have seen the 96.4% autonomy figure. It comes from a partner demonstration of the π0.6 model packing orders throughout one full working period, whose length Physical Intelligence didn’t specify. It isn’t a guaranteed result for every Ultra robot or customer.
Why people still watch the robots remotely
Does that mean people can leave the robots entirely alone? No. Human operators still help correct errors, and those corrections provide training material for future models.
Schwartz sees remote human control as a way to develop greater autonomy. He has argued that simulation still struggles to reproduce physical manipulation accurately enough. For you as a customer, the useful distinction is between a robot doing more work independently and one that never needs assistance. Training data is valuable enough that some startups pay people to train robots.
Inside a real deployment at Highline Commerce
Highline Commerce gives you a customer example beyond the product specifications. The Brooklyn logistics provider has used three OP1 systems under a service arrangement, with Ultra robots handling up to 30% of its orders. Here’s what that installation has involved.
Getting dependable output took about three months after the robots arrived. During that period, reported autonomy improved from roughly 70% into the 90% range.
That’s progress, but it also tells you why installation time and operating readiness are different measurements. Plugging in the machine was only part of the work.
- Three robots in Brooklyn. Highline put multiple OP1 units to work and nicknamed one “Fleetwood Pack.”
- Screen-based operation. The robot uses Highline’s warehouse software through its interface, handling scans, order searches and labels without custom software integration.
- A monthly fee. Billing changed from payment per order to a recurring service charge intended to encourage heavier use.
- A dedicated employee. Highline hired a robot manager to support the operation.
- Further training. Work has included teaching the robots multi-item orders, custom mailers and apparel orders with full-page inserts.
- More capacity planned. Highline ordered additional robots ahead of peak season.
There’s a practical lesson here. Highline CEO Richard Hurley has identified keeping goods supplied to the robot as a major operating challenge. You still need the surrounding workflow to keep pace.
Other customers and sites
Highline isn’t the only name attached to Ultra’s machines. Manifest has provided a customer testimonial, and Ultra describes installations across several US locations. Here’s where you can see evidence of its activity.
- Manifest. Senior operations manager Colter Lutz describes Ultra taking routine order processing off the team’s hands.
- New York. A robot alternates between bulk polybagging with an automatic bagger and packing padded mailers using items supplied in totes.
- Atlanta. Robots pack and seal padded and paper mailers across different product types and package sizes.
- Carlstadt, New Jersey. Ultra reports that a delivered unit began packing within hours of being unpacked.
- Austin, Texas. The company shows six OP1 robots operating alongside one another.
Ultra has targeted 100 deployments by the end of 2026. Treat that as a goal, since the available reporting doesn’t establish that it has already reached it.
Ultra versus humanoids and other warehouse robots
Warehouse robots don’t all do the same job, even when they appear in the same funding coverage. OP1 works at a station, while other systems use different designs for logistics tasks. Use these distinctions when comparing the names.
| Robot or company | Main distinction |
|---|---|
| Ultra OP1 | Two arms handle goods and packaging from a stationary workstation |
| Exotec | Uses robots capable of climbing storage racks |
| Amazon’s fleet | Operates robots across its US facilities, with no single equivalent machine specified here |
| Atoms | Develops logistics robots without a humanoid form |
| Humanoid robots | Use a humanlike form, unlike OP1’s stationary design |
Schwartz expects humanoids to reach comparable commercial scale in roughly five years. That is his prediction, rather than a confirmed timetable you can plan around. For a humanoid example from China, see BYD’s Xiao Di humanoid.
What Ultra has not disclosed yet
You can find funding totals, hardware specifications and customer examples, but several commercial details remain private. Those missing figures matter if you’re considering the service for your own warehouse. Keep these questions on your list.
- Robot pricing. There is no public amount for the integration charge or monthly service.
- Revenue. Ultra reports earning revenue without releasing a figure.
- Valuation. The company hasn’t disclosed its own valuation.
- Installed fleet. Reporting doesn’t provide a confirmed current robot count.
- Independently verified output. Published throughput claims don’t establish what you would achieve with your order mix.
- AI dependence. The service relies on models supplied by Physical Intelligence, rather than models Ultra owns.
What to watch next
Ultra’s next test is whether more installations can turn into dependable daily packing capacity. Customer results will help you assess that more clearly than another funding headline. These are the developments worth following.
- Deployment progress. Watch for confirmation of how close Ultra gets to its year-end target of 100 installations.
- Peak-season performance. Look for results from customers adding robots for heavier order volumes.
- Additional tasks. Follow progress on multi-item parcels, apparel and more demanding packaging requirements.
- AI updates. Watch whether later Physical Intelligence models reduce the need for human help.
- Recruitment. Y Combinator’s company page lists 21 openings, giving you another indicator of Ultra’s expansion plans.
If you’re considering a trial, start with your own order types and ask Ultra how it would measure success at your station. You’ll have a more useful conversation with clear expectations for output, human assistance and cost.
FAQs
Ultra makes OP1 Operator, a robot with two arms that performs tasks such as packing, sorting and kitting. You place it at a workstation, where it handles goods without travelling around the warehouse.
Ultra has raised $62 million through a $12 million seed round and a $50 million Series A. Y Combinator and NextView co-led the seed financing, while Framework Ventures led the Series A with Y Combinator participating.
Ultra’s documented offer is a lease, and the supplied information doesn’t describe an outright purchase option. You pay an initial integration charge followed by monthly fees for the hardware and software support.
Physical Intelligence supplies the AI used by Ultra’s robots to perform and learn packing tasks. Data from customer operations then helps train future models, while Ultra handles the physical robots and their installation.
No, OP1 is a stationary industrial robot with two arms rather than a humanoid. You can roll its base into place, but it doesn’t walk between workstations.
Highline Commerce is a documented user, and Manifest has supplied a customer testimonial. You can also find Ultra’s descriptions of installations in New York, Atlanta, Carlstadt and Austin, although those location listings don’t identify every customer.
Ultra was founded in 2024 and is based in Brooklyn’s Industry City in New York. If you’re checking its startup background, it also participated in Y Combinator’s Summer 2024 batch.
