한국어
ArchiveCase2020–present

Servi: a serving robot from an engineer who ran a tofu restaurant, now an LG subsidiary

Servi is a restaurant serving robot made by Bear Robotics in the US. It spread to Japanese family restaurants and Korean eateries, and in 2025 the company became a subsidiary of LG Electronics.

MakerBear Robotics (Redwood City, California, founded 2017). CEO John Ha
LaunchFirst mass-production model in 2020. Japan announced 28 September 2020; from January 2021 on three-year rental at ¥99,800 a month (excl. tax)
Size44.5×43×104.6 cm, two trays and a bus tub. 10 kg per tray, 30 kg total, up to 1.2 m/s (2020 Japanese materials give 1,051 mm and 35 kg)
SettingRestaurants, corporate cafeterias, senior care, casinos; now also hospitals, hotels and airport lounges
ReachMore than 16,000 robots shipped since 2017 (June 2026, all models)
Funding$32M Series A led by SoftBank (2020), $81M Series B (2022). LG Electronics took 21% in 2024 and 51% in January 2025
NowServi, Servi Mini, Servi Plus and Servi Q on sale. LG’s commercial robot business is being merged into Bear Robotics
Ended byNot applicable. Still on sale in 2026.
On sale; LG Electronics subsidiary

A Google engineer who ran a tofu restaurant

Bear Robotics was founded in California in 2017 by John Ha, a former Google engineer. While at Google he opened a tofu restaurant on the side. The work turned out to be harder than he expected, and he left Google to build robots for the tasks that could be automated. After a first robot called Penny, Bear launched its first mass-production model, Servi, in 2020. It stands just over a metre tall with two trays and a bus tub, and uses lidar and cameras to carry food to tables and bring back empty dishes. The indoor serving robot Dilly that Baemin showed at CES in 2019 had also been developed by Bear Robotics, a company Baemin had invested in.

From Japanese family restaurants to an LG subsidiary

In September 2020 SoftBank Robotics announced that it would offer Servi in Japan from January 2021 on a three-year rental at ¥99,800 a month, excluding tax. It said more than ten restaurant companies had already ordered 1,000 units. Seven & i Food Systems, which runs Denny’s in Japan, decided to adopt it after a trial, and the operator of Yakiniku King planned to put it in about 300 restaurants. Bear raised $81 million in 2022 and launched the larger Servi Plus in 2023. LG Electronics bought 21% for $60 million in 2024 and raised its stake to 51% in January 2025, taking control. LG said it would merge its CLOi commercial robot business into Bear Robotics. In June 2026 the company said it had shipped more than 16,000 robots since 2017 and agreed to buy the UK humanoid maker Kinisi.

Help for staff, not a replacement

Bear says Servi is designed to assist staff, not replace them. In 2022 Ha said: “Having started my own restaurant years ago, I learned firsthand just how hard things could get.” SoftBank said that in Japanese trials hall staff doubled the time they spent with customers and lunch-time table turnover rose 20%. These are company figures. The US Mexican restaurant chain Abuelo’s piloted one Servi at a single location in 2022. This research found no independent study or staff reaction specific to Servi.

A robot focused on one job stayed

Servi does not chat or show a face. It carries food and clears dishes. Like BellaBot, it found its place by taking on part of restaurant staff’s work. The striking part of this record is the company’s path. A start-up founded in 2017 took over a conglomerate’s commercial robot business eight years later. LG had serving robots of its own, but it folded that business into Bear Robotics.

How narrow should a robot’s job be for it to stay in use on site?

What we haven’t confirmed

  • How many Servi units have been sold or are in use (the 16,000 figure covers all Bear models)
  • US purchase price and monthly rental fee
  • When and how much Baemin invested in Bear Robotics, and whether all of Baemin’s later serving robots were Bear products
  • The actual start of sales in Japan and the number of units placed there