Data is the new oil. Is the well running dry?
The internet is no longer the best source of information for training algorithms. AI’s raw material is getting scarce, but it may be a competitive advantage for some companies.
“Data exists and has always existed. We are always creating more data. But in the past, to create Large Language Models, many companies simply went to the internet”, recalls Daan Baldewijns, VP of marketplace at Defined.ai, which sells datasets for training those models. Today’s reality, however, is very different.
The next chapters of generative AI face multiple challenges, such as growing needs for energy and computing capacity. At the same time, the abundance of information that allowed us to get this far is giving way to a scarcity of good-quality open data, which could hamper future progress in these areas.
AI translation from the Portuguese. First published in ECO on 31 January 2026.