PRESENTED BY
Digit 5 in Agility’s official launch imagery. Early access is expected in the first half of 2027. Image: Agility Robotics.
Agility unveiled Digit 5 today, putting a new machine behind the commercial promises that will follow it into its planned SPAC listing. Skild AI says it has passed $100 million in annual recurring revenue just ten months after its first commercial deployment. Physical AI is building a business case in customer facilities.
That shift runs through this week’s reporting: industrial buyers are helping shape Europe’s humanoids, while Reward AI is trying to make dexterous skills cheaper to teach. The question connecting them is how reliably a robot can do useful work, shift after shift.
IN THIS EDITION
TOP STORIES
Agility’s redesigned humanoid adds a 50 lb payload, new legs and interchangeable grippers. The company lists 90 minutes of runtime and nine-minute charging, alongside a safety architecture intended to let Digit work closer to people. These are preliminary specifications: some safety features remain in development. Early access is expected in the first half of 2027, with general availability targeted by year-end and EU and UK deployment planned.
Why it matters: With a proposed $2.5 billion SPAC merger ahead, this is a consequential product for Agility. Its filing tied a $300 million related-party commitment to 1,000 Digit v5 robots, subject to milestones and specifications. The new timetable moves early access beyond the filing’s earlier late-2026 expectation. Turning that commitment into working customer fleets is the test now.
Skild AI co-founders Deepak Pathak and Abhinav Gupta. The company reports more than 60 enterprise customers.
Skild says it has crossed $100 million in annual recurring revenue, with more than 60 enterprise customers spanning electronics assembly, food services, warehousing and inspection. Named customers include Foxconn, Mitsui and Sumitomo Wiring Systems. At G10 Fulfillment, a customer executive reports throughput rising from around 50 lines per hour manually to 130–140 with Skild’s systems.
Why it matters: This is a substantial commercial signal for physical AI software across several robot types. ARR describes recurring revenue at an annualized rate; it is not $100 million already booked over a full year. Skild’s accompanying argument is just as relevant: a robot must keep pace with the production line as well as complete the task. Paying deployments make those weaknesses hard to hide.
FEATURE OF THE WEEK · 25 MIN READ
The Humanoids of Europe: Built by Their Customers.
Europe’s humanoid industry is becoming easier to see if you start with the buyers. Our feature maps seven manufacturers across five countries: NEURA, Agile Robots, Hexagon, Humanoid, Wandercraft, Generative Bionics and UMA. Industrial groups including Schaeffler, Bosch and Renault appear repeatedly as customers, suppliers, manufacturers and investors.
Why it matters: Those relationships can give a robot company factory access, components and a route to deployment from the outset. They also create dependence on a small group of powerful buyers. Schaeffler’s commitments across several competing platforms show both sides of that arrangement. This is the longer read to save if you want to understand how Europe’s market is being built.
PRESENTED BY ROBOSTRATEGY
Most of the companies we usually cover in this newsletter are still private, which puts them out of reach for public-market investors. RoboStrategy, Inc. ($BOT) is built to give you access to several of them.
Investing involves risk, including possible loss of principal. Read the prospectus before investing.
Reward AI’s OM-1 demonstrations span different robot bodies. Image: Reward AI.
Reward AI has introduced OM-1, a robot policy it says learns from direct human manipulation data without teleoperation or on-robot experience in pretraining. Its wearable Omnibody Hand records movement, vision, touch and proximity. The company claims transfer across tabletop arms, industrial manipulators and humanoids, and learning of new long-horizon skills from under 30 minutes of human demonstrations.
Why it matters: Collecting useful training data is still one of robotics’ hardest scaling problems. Reward’s approach could let people teach skills through natural movements while capturing the contact information ordinary video misses. The next evidence to look for is sustained performance beyond the launch demonstrations, across unfamiliar tasks and customer environments.
UBTECH has commissioned its Liuzhou plant, targeting annual capacity above 10,000 industrial humanoids and wheeled embodied platforms. The 14,000-square-meter facility uses a flexible assembly layout designed around a ten-minute production interval. Wheeled Cruzr manipulators, autonomous forklifts and other logistics robots supply parts and move material around the floor.
Why it matters: Manufacturing capacity is becoming more concrete. Capacity still needs to translate into shipments, and shipments into useful work. The Liuzhou plant makes the next question sharper: how quickly can customers absorb the output and keep those robots productive?
THE RUNDOWN
Maven Robotics — Raised a $100M Series A for wheeled robots tackling mixed palletizing. Maven reports fleets of up to eight robots working 16-hour shifts at 99% or higher uptime. The round was led by RoboStrategy, this newsletter’s sponsor. More →
Unitree G1+ — A moving neck, stronger motors and a $15,000 base price before tax and shipping. Listed battery runtime remains around two hours. More →
Open robot models — Unitree released its six-billion-parameter UnifoLM-WLA-1.0; AGIBOT followed with GE-Act 2.0, a world-action model trained for manipulation.
Nucleus — Customer feedback pushed Nucleus II onto wheels, with a claimed 10 kg payload per arm and five to nine hours of operation. More →
Sharpa Wave — Our hardware guide examines the tactile hand used in NVIDIA’s reference humanoid and Google DeepMind’s demonstrations. Read the guide →
Munich’s “Kyle” — Nikolai Ensslen’s stealth team showed its walking development platform. A first-generation product reveal is targeted for early 2027; delivery timing remains undisclosed. More →
XPENG — He Xiaopeng reflects on six years of IRON development, public failures and the move toward automotive-scale manufacturing. More →
Agile Robots — CEO Zhaopeng Chen sees a market eventually ten times larger than automotive, while emphasizing the work needed to capture factory operators’ expertise. More →
This edition of Humanoids Daily was presented by RoboStrategy.
The most useful robotics metric may be the one a shift manager checks: how much work got done today? This week brought more evidence that customers are starting to shape the answer.
Seeing something before I do? Just hit reply — that's where a lot of these stories start.
P.A
Humanoids Daily
Know someone following the robotics industry? Forward them this briefing—it helps Humanoids Daily reach more people.
Received this from a friend? Subscribe to the free weekly briefing.