The AI boom is creating a massive wave of investment across the technology industry, but Zoho founder Sridhar Vembu believes the financial frenzy around AI could end badly for many companies. Vembu has described the current AI spending cycle as a “credit bubble,” arguing that huge investments in data centres and supporting infrastructure are pushing up costs across the technology supply chain. He also warned that the fact that AI technology is real and useful does not mean every company investing heavily in it will survive.
“The AI credit boom will meet the same fate, and I actually do believe the technology will achieve its full potential,” Vembu wrote on X.
Vembu compared today’s AI boom with the telecom investment bubble of the late 1990s. He pointed out that the internet and high-speed optical fibre eventually became widely available, including in remote areas, but that did not save many telecom equipment companies from going bankrupt when the bubble burst in the early 2000s.
According to Vembu, AI is creating similar distortions in areas ranging from electricity generation and transformers to cooling equipment, backup power systems, memory, CPUs and GPUs. His argument is not that AI will fail. Instead, he believes companies need to separate the long-term potential of the technology from the financial risks of investing too aggressively in it.
“Staying positive on the tech while being careful to not lose one’s shirt in the bubble unleashed by boosters of the same tech – that is the careful balancing act needed today,” he said.
AI is already hurting businesses
Vembu’s warning comes as the cost of running technology businesses is rising. In another post, he talked about a sharp increase in memory prices and said higher AI token costs were making business difficult for Zoho.
“Memory prices, along with AI token prices, have made business very difficult. We have held back from raising prices but it is becoming hard,” he wrote.
The pressure is not limited to smaller companies. Even major technology firms are beginning to question whether every AI-related expense can deliver a clear business return. Uber COO Andrew Macdonald recently acknowledged that the company does not always have an easy way to justify some of its AI investments. Microsoft, meanwhile, has reportedly scaled back certain external AI coding subscriptions, including Anthropic’s Claude Code, as companies look more closely at the costs of using AI at scale. Duolingo has also dropped an internal policy that connected employee performance reviews with AI usage after workers reportedly raised concerns about being encouraged to use AI even when it was not necessarily useful for their work.
Vembu also argued that the memory crunch could force programmers to rethink how software is built. For decades, programming languages have largely operated on the assumption that memory is relatively cheap and plentiful. Vembu believes that assumption no longer works and called for more memory-efficient programming languages and smarter compilers.
Zoho has already made clear that it will continue developing AI capabilities without joining the industry’s infrastructure spending race. Vembu said the company is focusing on areas such as data curation, reinforcement learning and compiler technology rather than chasing the massive investment cycle.
– Ends
Leave a Reply