Sridhar Vembu warns that the current AI investment boom may lead to a financial bust, as surging memory chip prices and over-exposure threaten to distort the wider economy, echoing past tech bubbles.
Sridhar Vembu has warned that the surge in artificial intelligence spending is spilling far beyond software and into the wider economy, pushing up costs for phones, laptops and the infrastructure that supports them. In a post on X on Tuesday, the Zoho co-founder said the current wave of investment has been financed by what he called a post-pandemic credit bubble, and argued that it is distorting demand across sectors ranging from power generation and transformers to backup generators, cooling systems, memory chips, CPUs and GPUs.
Vembu compared the pattern with the telecom investment boom of the late 1990s, when a real and transformative technology still left much of the equipment industry badly damaged after the crash. He said the same danger now exists in AI: the technology itself may prove durable, but the financing around it could still unwind sharply. The point, he argued, is not to reject AI, but to avoid being over-exposed to the speculative build-out surrounding it.
His remarks come as Wall Street and the technology sector continue to pour money into AI infrastructure. Axios reported this month that Goldman Sachs economists estimate AI investment will reach about $600 billion in 2026, equal to 2% of US GDP and 15% of equipment investment, while also raising borrowing costs more broadly for companies outside the sector. Nvidia, meanwhile, has outlined plans with six major financial firms to mobilise more than $500 billion for AI data centre funding, a move that has been read both as evidence of deep investor confidence and as a sign of how dependent the industry has become on ever-larger pools of capital.
Vembu also highlighted memory inflation as a direct business problem. He said prices for memory chips had risen 500% in 12 months and were now around ten times their lowest level, making it harder for Zoho to avoid passing on costs. Market data reported by Tom’s Hardware suggests the rise in DRAM and NAND prices is still continuing, although more slowly, as consumer demand weakens while AI-related demand keeps supply tight. Vembu added that software development itself may need to change, because many programming languages were designed on the assumption that memory was effectively free. In his view, that assumption no longer holds, and efficiency will matter more for both general software and AI systems.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
