---
title: "A Robot Company Raised US$70M. Its Machine Made 2,156 Drinks in One Run, the Company Says"
date: 2026-11-19
category: Humanoids & Robotics
site: NeuroAI
canonical: https://neuroai.site/a/na-robot-unix-ai-prea
language: en
---

# A Robot Company Raised US$70M. Its Machine Made 2,156 Drinks in One Run, the Company Says

> UniX AI's Pre-A round is less about a humanoid (人形机器人) demo than about an unglamorous metric: robots that keep working after the cameras leave.

At a JD Seven Fresh Coffee store, a machine takes an order, grinds, pours and hands over a cup. There is no operator in the background. It ran for twenty-four hours straight, and the company says it will do it again tomorrow.

The company behind it just raised enough money to stop proving it can make one cup and start proving it can make ten thousand.

On November 9, UniX AI (优理奇), a Suzhou-based embodied AI robotics firm, announced a Pre-A funding round of RMB 500 million (≈ US$70M / HK$550M). Existing backers Xienuo Investment and Yonghua Capital added more, while Greenland Group, Keli Sensing Technology and several other industrial investors came in new. The coffee machine is not a promotion. It is the closest thing the company has to a sales receipt.

## Key takeaways

- UniX AI says it has shipped more than 100 robots a month since late 2025, with a peak above 300 a month, and that several customers have placed orders for more than 100 units.

- Its robots have been deployed in building security, hotel chains including Hilton, Ji Hotel and Home Inn, and retail. In one 24-hour run at a JD Seven Fresh Coffee store, a robot made 2,156 drinks without an error — a Guinness World Record, according to the company.

- Greenland Group, a property developer that was a customer, joined the round as a strategic investor. The two plan to work together on security, hospitality and elderly-care deployments.

- The money goes into algorithms, engineering and scaling — not into a new robot reveal. That is a telling shift for a sector still judged largely on videos.

## The customer became the investor

The most interesting name on the investor list is not a venture fund. Greenland Group buys and operates space — office towers, hotels, residential compounds. It is the kind of company that would normally sign a purchase order for security robots. Instead it bought a stake.

UniX AI says the two will jointly build solutions for security, hospitality and elderly care, linking robot hardware, AI models and operating data. For the robotics industry this is a small but real change in how deals are structured. A customer that takes equity has a reason to keep the machines running and to feed back the messy details that never appear in a demo: a lifting lobby door, a guest who changes his mind mid-sentence, a floor that has just been mopped.

The rest of the round follows the same logic. Keli Sensing brings sensor supply-chain experience, a Hong Kong University of Science and Technology-backed fund brings academic ties, and a local industrial fund from Yulin brings a route into regional deployments. This is capital organized around places where robots can actually be put to work.

## What "shipping 300 a month" means

Robot funding announcements usually lead with a dramatic number and trail off before the delivery figures. Here the delivery figures are the story. The company says more than 100 units went out every month from late 2025, peaking above 300, with positive cash collection and gross margin. Those are self-reported numbers, but they describe a different business from a lab that has built twelve prototypes.

The hardware behind them is unglamorous. Wanda and Panther are wheeled, dual-arm platforms rather than walking humanoids (人形机器人). The company says it develops its own joint modules and arms. On the software side it lists UniFlex for learning manipulation from demonstration, UniTouch for combining vision and touch, and UniCortex for planning long task sequences. A tactile-model paper, UniTac, was accepted at ECCV 2026.

That stack explains the 24-hour run the company is promoting. Making 2,156 drinks in a day is not a feat of general intelligence. It is a feat of repeatability — getting the same grip, the same cup position and the same timing right, thousands of times, without a human stepping in.

## Why the medals matter less than the job list

UniX AI also entered three Panther robots in the 2026 World Humanoid Robot Games, taking one gold, one silver and one bronze, including a second consecutive hospitality title and a first-attempt medal in a firefighting task. Competition results make good footage. They are also a weak proxy for commercial value: a course is a controlled environment, and a hotel lobby is not.

The stronger signal is the customer list. Security robots have been sold to property groups and to firms including Certis and Inspur. Hotel robots are running at Hilton, Ji Hotel and Home Inn brands. These are not research partners. They are organizations that have to justify a line item, and they tend to renew only when a machine survives a shift without generating complaints.

For ordinary readers, the shift is subtle but important. The embodied AI sector has spent two years being sold as a race to build a walking, talking machine. The investors in this round are betting on something quieter: machines on wheels that do one repetitive job reliably, in buildings that already exist, with a person still somewhere in the loop.

## Honest limitations

This article draws on Science and Technology Daily, the Suzhou Industrial Park administration, investment-media reports and an English-language industry outlet, all describing the same funding announcement. The delivery numbers, the 2,156-drink record, the positive margin and the investor roles are company claims repeated by media; I have not seen audited financials, the Guinness certificate or customer contracts.

The currency conversion is approximate and depends on the exchange rate used; some English coverage puts the round at US$75 million rather than US$70 million. I have used the yuan figure as the anchor and kept the dollar figure as a rough guide. Valuation was not disclosed.

The company says it has shipped over 100 units a month, but a shipment is not the same as a paid, renewed deployment, and "positive gross margin" is not profitability. Nothing in the sources breaks down revenue by customer, so concentration risk cannot be assessed. I have also not independently checked the competition results beyond the consistent reporting of a one-gold, one-silver, one-bronze finish.

## What readers can do now

- **If you work in facilities, hospitality or retail, ask vendors for a paid pilot with a renewal clause.** A robot that completes a shift and a robot that a customer orders again are different products.

- **If you are following the sector, watch the ratio of delivery news to demo news.** Shipments, renewals and error rates will tell you more about the industry's health than another podium finish.

- **If you use one of these machines, report the failures.** The awkward handoff — the cup placed two centimetres off, the door that will not open — is exactly the data that makes the next version useful.

The coffee robot did not impress anyone by walking or talking. It impressed someone by showing up for every shift. In a business built on spectacle, that may be the most conventional measure of progress there is.

---

Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-robot-unix-ai-prea
Free to quote with attribution and a link to the original.
