Ultra raises $62M to scale warehouse robots and deepen Physical Intelligence partnership

Ultra, a Brooklyn-based robotics startup founded in 2024, announced on October 9, 2026 that it has raised $62 million and is deepening its partnership with Physical Intelligence, the San Francisco firm that supplies the artificial intelligence running inside its warehouse machines. The money came in two parts: a $50 million Series A led by Framework Ventures, with Y Combinator participating, and an earlier $12 million seed round led by Y Combinator and Next View.

The company does not sell robots. It leases them, under a monthly 'robots as a service' arrangement that asks warehouse operators for an up-front integration fee and then recurring payments for hardware and software support. Its machines are dual-arm, stationary units called Operator, or OP1, that handle packing, sorting and kitting, the job of bundling several separate items into a single shippable package.

The most striking thing about Ultra is what it is not. It does not build humanoids. While much of the robotics industry and most of the venture capital press attention has gone to bipedal machines that walk, Ultra builds fixed industrial systems meant to drop into warehouse floor space that already exists, doing repetitive fulfillment work that people currently do by hand.

That pitch has now attracted enough customer demand for the company to raise prices, a rare thing for a young hardware business. Ultra says its robots have packed more than 500,000 orders across warehouses in the United States, with deployments in New York, Georgia, New Jersey and Texas. The company frames the round as a bet that the unglamorous robot format is the one that actually ships.

Key Facts

Fortune reported on October 9, 2026 that Ultra announced the $62 million raise on Friday and was deepening a partnership with Physical Intelligence, which Fortune described as a 'robot brains' artificial intelligence firm. The Term Sheet item, written by finance editor Jeff John Roberts and published at 6:33 AM Eastern time, noted that Ultra bets on non-humanoid warehouse robots rather than bipeds.

Crypto Briefing reported on October 9, 2026 that Ultra rents its robots to warehouses the way a contractor might rent a forklift. The outlet put the financing at a $50 million Series A led by Framework Ventures, with Y Combinator participating, plus an earlier $12 million seed round led by Y Combinator and Next View. It noted that Ultra went from its 2024 founding to a $50 million Series A in roughly two years.

TokenPost reported on October 9, 2026 that Ultra's systems handle packing, sorting and kitting, can be installed in hours rather than weeks or months, and are deployed in New York, Georgia, New Jersey and Texas. The same report gave the Operator hardware its specifications: a 5-by-5-foot footprint, a work area of up to 10 by 10 feet, a vertical reach of up to 10 feet, and arms that can each carry up to 10 pounds.

The artificial intelligence side of the partnership has its own numbers. Physical Intelligence, also founded in 2024, reached a $5.6 billion valuation after a $600 million Series B in November 2025, and its π0.6 model reportedly reached 96.4 percent autonomy during full shifts in real warehouse settings, according to deployment data released in February 2026. TokenPost reported that the autonomy figure covers an eight-hour shift shown in a continuous-deployment video.

Customer evidence is limited but specific. Crypto Briefing reported on October 9, 2026 that one deployment, at Highline Commerce in Brooklyn, handles up to 30 percent of that company's fulfillment volume. Ultra says its robots have packed more than 500,000 orders at United States sites, and the company says the model has worked well enough that it has been able to raise prices on customers.

Analysis

The bigger picture here is that the funding market for robotics has quietly split into two very different businesses, and the money is starting to follow the one that looks less exciting. Humanoid robots generate the videos, the conference keynotes and the cultural conversation. Fixed industrial arms generate invoices. Ultra chief executive and co-founder Jon Miller Schwartz made exactly that argument, comparing the situation to the way 1980s fictional robots overshadowed the industrial robotic arms that were already doing real work in factories. He also said the industry is about five years away from seeing this sort of scaling on the humanoid side, which is a notably patient framing from someone whose own company is scaling right now.

The structure of Ultra's business explains why investors were willing to fund it. Because customers pay an integration fee plus a monthly subscription, Ultra carries the hardware on its own balance sheet, which is capital intensive and risky, but it also builds a recurring revenue line that software investors understand. The 'body and brains' split does the rest of the work. Ultra builds and installs the metal; Physical Intelligence supplies the vision-language-action models that let a machine see a shelf, decide what to grab and place it into a box. That lets each company stay narrow. Ultra does not need to staff a frontier AI research lab, and Physical Intelligence does not need to run a manufacturing and field service operation.

What this really means is that the practical bottleneck in warehouse automation is no longer whether a robot can pick an object, but whether a robot can keep picking objects for eight straight hours without a human standing behind it. The 96.4 percent autonomy figure that Physical Intelligence reported for π0.6 during real shifts is the number that matters most in this deal, because every point of remaining autonomy is a person a warehouse still has to pay. A machine that needs intervention every few minutes is a demo. A machine that runs a full shift is a product.

The pricing power Ultra claims is arguably the strongest signal in the announcement. Robotics companies that lease hardware usually compete on cost against cheap labor, and that race runs only one direction. A startup that can raise prices is telling the market that its units are not competing with a minimum wage worker but with an unfilled shift, a missed shipping deadline or a peak season that a warehouse cannot staff. That is a different and much better argument, and it is the one that made a Series A possible roughly two years after founding.

Why It Matters

Ultra's round matters because it tests whether the 'robots as a service' model can survive contact with real balance sheets. Leasing robots means Ultra funds the hardware, insures it, maintains it and hopes the monthly fees outrun the depreciation. If customer demand holds and pricing holds, the model produces the kind of predictable revenue that public markets reward. If a warehouse chain decides to buy its own arms instead, Ultra is left holding expensive inventory.

It also matters for the wider AI capital cycle. Physical Intelligence's $5.6 billion valuation after a $600 million Series B in November 2025 is a bet on foundation models for physical action, a category with far less commercial proof than language models. Ultra is one of the clearest places where that bet gets tested in the real economy, because the π0.6 model is not being evaluated on a benchmark suite but on live customer floors in Brooklyn, New Jersey, Georgia and Texas.

Finally, the deal matters for how the industry talks about itself. Crypto Briefing reported on October 9, 2026 that investor appetite continues for robotics firms that can show revenue rather than just research, and that framing is likely to harden from here. If Ultra succeeds while humanoids remain roughly five years from this kind of scaling, the lesson for founders will be that the highest-value robot is often the one that never moves its base.

Next Up

The immediate question is deployment. Ultra says installations take hours rather than weeks or months, which means the constraint on growth is sales and support capacity rather than engineering. Expect the Series A money to go toward field operations, more Operator units and additional customer sites, with Highline Commerce in Brooklyn serving as the reference account the company will show to other fulfillment operators.

The second question is the partnership itself. Physical Intelligence's π0.6 model already runs on Ultra robots in live customer warehouses, and the next milestone to watch is whether autonomy climbs past the 96.4 percent reported in deployment data released in February 2026, and whether that improvement arrives through a new model release. For now, the two companies have a clean division of labor, one that lets a startup founded in 2024 in Brooklyn sell a robot without owning the intelligence inside it.