It is starting to look like big tech’s dreams of affordable robots are hot air and hype.
Nvidia chief executive Jensen Huang said in January last year, “The ChatGPT moment for general robotics is just around the corner,” and in January this year, he said the ChatGPT moment for physical AI was “nearly here”.
According to a UBS report, ChatGPT became the fastest-growing consumer app in history, reaching 100 million monthly active users within two months. Meanwhile, Huang’s robot revolution is moving like an asthmatic ant with a heavy load of shopping.
While it is true that robotics has moved on in the past year, with self-driving cars inching forward, the retreats are just as telling. US supermarket chain Kroger said in November it would shut three of its eight robotic warehouses that pick and dispatch customer orders.
At the same time, Kroger expanded relationships with gig-economy outfits Instacart and DoorDash, which marshal self-employed individuals as “personal shoppers”. When the robots get expensive, the spreadsheet quietly hires more people on precarious terms.
Dropping generative AI into a business can be as easy as paying a monthly subscription to OpenAI or Anthropic. Robotics is different because you are buying planning, integration, downtime risk and a warehouse full of pricey kit.
Kroger’s sites, powered by technology from UK company Ocado, need strong utilisation to justify the spend. Grocery retail runs on thin margins, so a bad demand forecast hurts more than a bruised ego.
STIQ warehouse automation expert Tom Andersson said: “It’s a bit more like a factory set-up: it needs to have minimum throughput to make money.”
STIQ warehouse automation expert Tom Andersson said: “In the end, you need to have a really good business case for why you do automation, and when you do those business cases. Sometimes these projects can take three years in the planning — if your forecast is wrong at that point, it will be tricky.”
Automation is not dead in groceries; it is just picky about geography and maths. Ocado’s model looks better in the densely populated UK, while Dutch online grocer Picnic has made progress with a similar approach.
Plenty of supermarkets have automated further up the supply chain, even if they keep the flashy robots out of the shop floor. Walmart has grown revenue by more than $150bn over five years while headcount has dipped slightly, yet it still employs more than two million people.
Robot fans now want humanoids to do more work, partly because human-shaped machines can fit into workplaces built for humans. That makes deployments sound gradual and flexible, rather than a single giant bet that can sour fast.
Even if humanoids hit human-level capability this year, as Huang predicts, businesses still have to worry about safety when robots work alongside humans. A hallucinating chatbot can wreck your numbers, while a malfunctioning humanoid can wreck someone’s ribs.
Then there is stamina, because batteries are still the villain of physical AI. University of Wisconsin-Madison associate professor of mechanical engineering James Pikul noted Boston Dynamics’ robot Spot can work for about 90 minutes before recharging, while humans in warehouses routinely grind through 10-hour shifts with a couple of breaks.
