Serve Robotics reported third-quarter revenue of $2.8 million, up 313% year-on-year, and on the same day opened the software and data stack behind its fleets to other robot makers under the name Physical AI Services. The earnings are in a 5 November 2026 investor release; the product launch is in a companion IR announcement. Both are company claims for the period ended 30 September 2026.
Recurring revenue was 69% of the quarter’s total. Cash and marketable securities stood at $198.0 million. Daily Active Robots averaged 917, versus 312 a year earlier. Serve said it has deployed more than 2,000 robots, added an Australian city in October, and is raising full-year 2026 revenue guidance to $12 million from a prior $9–10 million range, while cutting non-GAAP operating-expense guidance to $150 million.
Co-founder and chief executive Ali Kashani tied the two stories together: demand for the robots is growing, unit economics are improving, and Physical AI Services is “a significant new opportunity” to sell the layer that keeps fleets connected and learning.
What Physical AI Services actually sells
The product note names three pieces. Serve Link connects robots to off-board intelligence over networks including Wi-Fi, LTE, 5G and satellite. Serve Data Curator is meant to surface rare, high-value field examples so models improve without reviewing every mile of video. A foundation-model data catalog draws on Serve’s own sidewalk deployments — terrain and infrastructure data — for training in complex human environments.
Bedrock Robotics, which builds autonomy for heavy construction equipment, is named as a launch customer using Serve Link. Serve says the same services already power thousands of vehicles across industries including construction, mining and autonomous trucking as well as its own delivery fleet — a company claim, not an audited third-party census.
Earnings texture without over-reading the fleet
Fleet services were $2.6 million of the $2.8 million quarter; software services were $0.3 million. Gross loss remained large on a GAAP basis because cost of revenues still dwarfs sales, but management pointed to sequential fleet-margin improvement and lower operating expenses. Healthcare automation — via the Diligent Robotics acquisition — added multi-year hospital contract extensions, Serve said. Branding revenue grew nearly 80% sequentially on Serve’s figures, another reminder that “robot company” revenue can include media and software lines beside delivery fees.
North American industrial robot orders are rising on a broader A3 survey TSN covered for Q3 robot order growth, while foundation-model robotics is moving into cloud previews such as Gemini Robotics ER 2 on Google Cloud. Public-markets robotics also includes humanoid SPAC paths such as Agility’s listing track. Serve’s pitch is narrower: sidewalk and indoor service robots plus a B2B layer for other fleets.
Why selling the stack changes the story
Delivery robots alone are a volume-and-margin grind. Licensing connectivity, curation and real-world catalogs is an attempt to monetise the painful parts of scaling autonomy without every customer rebuilding them. That only works if Link and Curator stay reliable outside Serve’s own routes, and if Bedrock-style customers renew. Construction and mining fleets have different duty cycles and safety cases than pizza sidewalks; a shared connectivity layer is plausible, but the IR note does not publish SLAs or attach rates.
The $198 million cash pile buys time to prove the thesis. The raised $12 million revenue guide still implies a small absolute top line against a $150 million non-GAAP opex plan, so Physical AI Services has to show up in software-services growth before the story shifts from “fleet scaling” to “platform”. Kashani’s framing — that other robot makers should not rebuild the systems around the machine — is the product pitch; sequential fleet-margin improvement is the operations proof Serve wants investors to weigh alongside it.
UK and other non-US readers should treat commercial availability of Physical AI Services as a US-company offering described in the IR note. Local regulatory permission for sidewalk robots, and hospital indoor-robot rules after the Diligent acquisition, are separate questions Serve does not settle here. Australia is named as a new city launch, not a full regulatory dossier.
Watch next quarter’s mix of fleet versus software services, Bedrock and any further named Physical AI customers, and whether Daily Active Robots keep climbing without margin relapse.
Sources
Primary:
- Serve Robotics Announces Third Quarter 2026 Results, 5 November 2026
- Serve Robotics Launches Physical AI Services, 5 November 2026
- Serve Form 8-K, 5 November 2026
Related TSN:
- Google Brings Gemini Robotics ER 2 to Businesses on Google Cloud
- Agility Robotics–Churchill Capital XI S-4 Declared Effective
- North American Robot Orders Rise in Q3, A3 Says
Related stories
- Google Brings Gemini Robotics ER 2 to Businesses on Google Cloud
- Agility Robotics–Churchill Capital XI S-4 Declared Effective; Vote Set for 2 November
- North American Robot Orders Rise 24.8% in Q3 to 10,989, A3 Says
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
