OpenAI's Sam Altman Makes A Big U-Turn On Creating Low-Cost AI Models: Know Why

Last Updated:February 05, 2025, 12:18 IST

OpenAI CEO is visiting India this week and he was understandably asked about his previous views on making AI models with less money.

Altman was part of an event in Delhi this week where he backtracked on his earlier comments.

OpenAI is facing a tough fight from China-based DeepSeek and the company’s chief Sam Altman now feels that you can build large language models (LLMs) at low-cost which goes against his previous comment of needing big money and power to make the AI models. But the emergence of DeepSeek has seen Altman change his views and at the event in Delhi.

Altman is visiting India this week and was part of a public event on Wednesday in New Delhi, talking to IT Minister Ashwini Vaishnaw. These views are a big U-Turn from Altman who, during his last visit in 2023, claimed building large AI models cannot be done cheaply and you need big funds to power these computing systems.

He was quizzed about building powerful AI models with a small team and budget of around $10 million, to which Altman said, “It’s totally hopeless to compete with us on training foundation models."

The big U-turn from Altman also tells you the pressure OpenAI is facing to become profitable and ensure that its investors feel their money is being put to good use. He even highlighted the role of India in the AI battle and how building new models have become less complex. “India is a very important market for AI. It is our second biggest market. We are now in a world where we have made incredible progress with distillation. Models are still not cheap, but they are doable. India should be a leader there of course," he said on the sidelines of the event.

It is intriguing to see Altman backtrack on his views from two years back, which shows the pace of AI advancement in the last 24 months and how these US companies are being shown up by startups from China like DeepSeek and Alibaba among others.

He has even talked about being impressed by the DeepSeek R1 model which is claimed to have been developed using less powerful Nvidia GPUs and a leaner team. It is not the first time when someone has made a big change in their opinions and Altman seems willing to admit that his judgement was premature and the future of AI will surely see more twists and turns to get people talking.

News tech OpenAI's Sam Altman Makes A Big U-Turn On Creating Low-Cost AI Models: Know Why