BENGALURU: IBM chairman and CEO Arvind Krishna observed that the AI race is focused on whether bigger models mean better outcomes. Yet, AI systems do not inherently need to consume extensive resources or require substantial investments.
“Following IBM’s earnings announcement earlier today, I had a lot of conversation and without fail, DeepSeek came up…. For too long, the AI race has been a game of scale where bigger models meant better outcomes. But there is no law of physics that dictates AI models must remain big and expensive. The cost of training and inference is just another technology challenge to be solved,” he wrote in a LinkedIn post last week. “We’ve seen this play out before. In the early days of computing, storage and processing power were prohibitively expensive,” Krishna wrote.
Yet, Krishna said through technological advancements and economies of scale, these costs plummeted. “AI will follow the same path. At IBM, we’ve been on this journey for years. We’ve been steadfast in our belief that efficient, fit-for-purpose models can deliver exceptional value and that is key to innovation. Our work has already led to up to 30-fold reductions in inference costs. This is promising for businesses. Technology becomes truly transformative when it becomes more affordable. As more companies embrace this shift, we’ll see an AI landscape that is both more powerful and accessible,” he said.
Last week, Chinese AI startup DeepSeek’s dramatic impact on tech stocks worldwide generated a huge interest among users. DeepSeek is founded by Liang Wenfeng, who previously founded a hedge fund. What makes its achievement noteworthy is its ability to make a ChatGPT-class language model at a fraction of ChatGPT’s cost. DeepSeek’s R1 AI model is said to match or even beat the likes of ChatGPT and Google’s Gemini on multiple parameters.
Krishna’s LinkedIn post came after IBM’s Q4 results. The tech firm reported a revenue of $62.8 billion for the 2024 fiscal year, up 3% in constant currency. Its GenAI business contributed more than $5 billion since its inception and grew by $2 billion compared to the previous quarter. IBM follows a calendar year.
“Our AI portfolio is tailored to meet the diverse needs of enterprise clients, enabling them to leverage a mix of models, IBMs, their own, open models from Hugging Face, Meta and Mistral. IBM’s Granite models designed for specific purposes are 90% more cost-efficient than larger alternatives,” said Krishna in a recent earnings call.
