Robot Data Collection Faces IPO Scrutiny and Business Model Pressure
TMTPost reports that China established 106 humanoid-robot data-collection centers in under two years, most jointly built by public and private entities. It describes a model in which companies sell robots to collection centers, which use them to gather data that the companies later buy back for model training. Mech-Mind founder Shao Tianlan and other industry participants alleged that some firms used such transactions to inflate revenue; the report did not independently establish the allegations. Since the second quarter, IPO inquiries for robot companies on the STAR Market and Beijing Stock Exchange have reportedly focused on revenue authenticity, related-party transactions and sustainable gross margins. One collector cited by the report said data prices fell from more than 200 yuan per hour last year to more than 100 yuan this year. The report also describes a shift from robot teleoperation toward bodyless collection, and presents the potential effect on demand for robot-based collection as conditional rather than demonstrated.
The report says investors are now asking who generates revenue, whether equipment is actually used and whether customers will continue buying. It also identifies difficulty recovering the substantial costs of robot-based collection as data prices fall and describes bodyless collection as a possible alternative, highlighting risks to the model’s revenue quality and commercial viability.
（本文作者为 硅碳变量，钛媒体经授权发布）
文｜硅碳变量，作者｜胡一刀
一边是机器人公司为了融资和上市制造订单，一边是不断降价、层层转包的数据，一位数据采集商无奈表示，“这门生意明年会是什么光景，还很难说”。
这两年，具身智能如火如荼，但不同于大语言模型，互联网有海量的数据可以采集，机器人此前基本没有现实数据，由此，数据采集和交易这一生意突然变得火热起来。可是，伴随着新一轮的上市潮，机器人的数据采集正在引发越来越多的质疑，前不久梅卡曼德创始人邵天兰的公开发声，更是直接把部分机器人企业借助数采中心包装虚假营收的乱象摆在了台面上。
一时间，数采不再是产业基建，反而成了关联交易、报表注水的“代名词”。
更令从业者不安的是，人形机器人的订单开始集体挤水分，明星企业的估值泡沫似乎也被戳破了。一旦市场动荡，数采这门生意可能更不好做了。