The CEO of IBM, Arvind Krishna, is questioning the substantial investments being made by major technology companies into artificial intelligence (AI) infrastructure. In an interview with the German business publication “Handelsblatt”, Krishna indicated a more cautious approach, stating IBM would not be committing to such large-scale expenditures.
The rationale behind this perspective stems from recent record-breaking investments in data centers and GPU semiconductors, vital for powering AI services similar to OpenAI’s ChatGPT and Google’s Gemini. Alphabet, Google’s parent company, is reportedly planning an investment of $75 billion this year alone.
Krishna outlined IBM’s preference for renting necessary infrastructure, emphasizing the need for flexibility as the AI landscape rapidly evolves. He predicted that the costs associated with “inference” – the application of AI models – could decrease to just one percent of current levels within five years.
IBM faces particular scrutiny regarding its AI strategy. Once considered a pioneer in the field with its Watson supercomputer, the company missed the wave of excitement surrounding large language models. Furthermore, the burgeoning cloud data center business has largely benefited competitors like Microsoft and Amazon Web Services (AWS).
Rather than pursuing a broad-spectrum approach, Krishna is advocating for a niche strategy focused on smaller, specialized AI models. He highlighted IBM’s investment in numerous, less resource-intensive models, capable of running on less powerful computing systems. While offering competitive models through partnerships, Krishna asserted that an estimated 99 percent of enterprise data remains untouched by existing large language models, indicating considerable untapped potential.