A holder who keeps ownership and licenses non-exclusively sells the same dataset again and again; a made-to-order sale ends with the first buyer
Haitian Ruisheng (SpeechOcean) sold usage licences on 395 of its own off-the-shelf datasets in 2020, 576 licences in all, from a catalogue of 762; its made-to-order data passes to the client and cannot be resold. In 2020 those owned datasets earned 45.48 percent of its revenue. A Brazilian data supplier to Defined.ai told Reuters he passes 20 to 30 percent of deal amounts to the owners of the photos, podcasts and medical data he sources.
Dated 5 Apr 2024Source checked 16 Oct 2026Page out of reach at the last check, 16 Oct 2026Basis: indicative
“销售量（个数） 395 293 183 销售量（授权次数） 576 414 265”
Haitian Ruisheng IPO prospectus (招股说明书), training data products, checked 9 Aug 2021
Everything read from this source (8)Archived copy, 16 Oct 2026
Haitian Ruisheng IPO prospectus (招股说明书), business model, checked 9 Aug 2021
Everything read from this source (8)Archived copy, 16 Oct 2026
Reuters, Inside Big Tech's underground race to buy AI training data (Katie Paul and Anna Tong), checked 5 Apr 2024
- Off-the-shelf speech recognition training datasets (wake words, read sentences, dialogues; Chinese and foreign languages), usage licences sold by Haitian Ruisheng in 2020: 198,951 hours
- Off-the-shelf speech synthesis (TTS) training datasets, usage licences sold by Haitian Ruisheng in 2020: 1,029 hours
- Off-the-shelf speech recognition training datasets (wake words, read sentences, dialogues; Chinese and foreign languages), usage licences sold by Haitian Ruisheng in 2019: 136,411 hours
- Off-the-shelf speech synthesis (TTS) training datasets, usage licences sold by Haitian Ruisheng in 2019: 274 hours
- Off-the-shelf speech recognition training datasets (wake words, read sentences, dialogues; Chinese and foreign languages), usage licences sold by Haitian Ruisheng in 2018: 115,636 hours
- Off-the-shelf speech synthesis (TTS) training datasets, usage licences sold by Haitian Ruisheng in 2018: 96 hours
It is shown on its own, not counted into a figure.
Cited in this guide