At 11:40 on a weekday, a basement kitchen in a Chinese office district produces two hundred portions of twice-cooked pork. Nobody is holding a wok. A drum of oil, a drum of meat, a drum of leeks and a machine the size of a vending unit do the work in ninety-second cycles, and the only human in the room is wiping the floor.
The food is fine. The argument was never really about whether a machine can brown pork. It was about what happens when the same chain opens its four hundredth branch and the pork tastes different there.
On 1 November 2026, that argument got a referee.
Key takeaways
- GB/T 46718-2025, Commercial intelligent stir-frying machine (《商用智能炒菜机》), was published on 31 October 2025 and came into force on 1 November 2026.
- It is a recommended national standard (GB/T), not a mandatory one — it sets technical requirements and test methods, it does not ban anything.
- The market it governs grew fast: one industry report put 2025 value at 7.72 billion yuan (≈ US$1.09B / HK$8.5B), up 68% year on year, forecasting 13.12 billion yuan (≈ US$1.85B) for 2026.
- The economics driving adoption are labour, not novelty: a case cited by one vendor puts two wok chefs at 22,000 yuan a month (≈ US$3,100) replaced by one operator at 6,000 yuan (≈ US$845).
- The standard deliberately excludes home units and continuous industrial fryers — this is a rulebook for restaurant kitchens.
What the document actually covers
The standard was published by the State Administration for Market Regulation and the national standardisation body, with the China National Light Industry Council as the responsible authority.
Its scope is narrow and telling. It applies to commercial intelligent cooking equipment that is mainly used for stir-frying, with single-phase rated voltage not exceeding 250 V (480 V for other configurations) and a single heating unit rated at no more than 35 kW. Where a machine runs on gas or oil instead, its electrical parts still have to comply.
It explicitly does not apply to home stir-fry machines, to continuous-production industrial frying equipment, or to equipment used in corrosive or explosive atmospheres.
That boundary is the point. The people writing this document were not thinking about a gadget on a kitchen counter; they were thinking about a 35 kW appliance bolted into a restaurant that serves strangers, and about what "safe" and "consistent" have to mean before an insurer will touch it.
Why the market got here before the rules did
For most of the last five years, commercial stir-fry robots were sold on two promises: replace a scarce wok chef, and make the four hundredth branch taste like the first. Both promises are hard to verify from the outside, which is exactly the environment in which bad products thrive.
The demand side is real. A 2025 survey by the China Hotel Association found 78% of catering companies facing a staffing gap above 30% in peak season, with hiring cycles of 15 to 20 days — and wok chefs, who take years to train, are the hardest gap to fill.
Vendors responded with a familiar Chinese industrial pattern: a fast race to volume, then pressure for rules. One vendor, Xianglu Robotics (橡鹿机器人), founded in 2021, said that by October 2026 it had served more than 3,000 restaurant brands across more than 13,000 stores in close to 350 Chinese cities and over 20 overseas countries and regions, with nearly 600 million cumulative servings and close to one million digitised recipes. In April 2026 it closed a 300-million-yuan (≈ US$42M / HK$330M) funding round led by a Beijing Economic-Technological Development Area industrial upgrading fund.
Company-disclosed engineering details show where the work went: solid seasoning dosed to about ±0.5 grams, temperature control claimed at ±0.5 °C, and a visual model that watches the wok through global-shutter industrial cameras to catch the failure modes a recipe cannot predict — too much water in the mapo tofu, meat that went in half-frozen.
The part that decides whether this holds
A standard does not make a machine cook better. What it does is move the argument from "trust me" to "test it".
For a chain operator, the practical change is that procurement can now specify a test method instead of a demonstration. That matters most for the boring properties: how stable the output temperature is across a full service, what the machine does when a heating element degrades, how it behaves after eight hours of continuous use.
It also reframes the numbers vendors like to quote. A claimed payback of five to seven months for a fast-food model, or a fivefold speed-up in a breakfast chain case, only means something against a defined duty cycle. Without one, "output" is whatever the machine managed on the day of the demo.
The standard is also a signal about maturity in a less flattering way. Home stir-fry machines, once pitched as the consumer future, were excluded from it, and one industry estimate put home units at just 5.4% of 2025 sales by value. The commercial kitchen, not the family kitchen, is where the money is.
According to Li An, chief scientist at BrainNet (脑机网), the shift worth tracking is that cooking robots stopped being sold as appliances and started being sold as a reproducibility guarantee — and guarantees are what standards exist to price.
Honest limitations
This article relies on official standard records, state-affiliated media and industry research reports. Market-size figures vary wildly between sources: one report puts 2025 Chinese commercial stir-fry robot market value at 7.72 billion yuan (≈ US$1.09B), while another industry association's estimate is around 3.7 billion yuan (≈ US$520M) for a broadly similar category. The gap is large enough that these should be read as directional, not measured.
Vendor figures — store counts, servings, recipe libraries, payback periods, temperature and dosing precision — are company-disclosed and were not independently audited here. Payback maths depends heavily on local wages, rent and menu, and a five-to-seven-month figure from a fast-food format does not transfer to a full-service restaurant.
Not covered here: how quickly compliance testing capacity will exist, what enforcement looks like for a voluntary standard, and how the export market — more than 600,000 overseas Chinese restaurants are often cited — will treat a Chinese GB/T mark.
What readers can do now
- If you run or supply a restaurant, stop evaluating these machines on a tasting. Ask for the test report under GB/T 46718-2025 and read the duty-cycle conditions, then price the machine against a full service rather than a single dish.
- If you are considering one for a small kitchen, rent before you buy. Some vendors have offered leasing at roughly 3,000 to 5,000 yuan a month (≈ US$420–700), which is a cheaper way to learn whether your menu suits the machine than a purchase.
- If you are a chef or kitchen worker, the useful skill to build is menu engineering for machines: recipe standardisation, yield control and equipment maintenance. That is the work the standard makes measurable.