Article

Major technology companies, including Microsoft, Nvidia, and Apple, are investing hundreds of billions of dollars annually in AI infrastructure, primarily for data centers and chips, with capital expenditures reaching $400 billion in the past year. Experts warn of a potential AI bubble, which could threaten the global economy if it bursts.

The lifespan of chips used for AI training is a concern, with specialized GPUs estimated to last 18 to 36 months under heavy workloads, compared to 5 to 7 years for traditional CPUs. Nvidia states its CUDA system allows for software updates to extend chip life.

Despite increasing demand for generative AI from individual users, corporate customers are still figuring out how to monetize the technology. Concerns over the sustainability of long-term AI demand persist, especially as many companies have not seen improved financial results post-implementation. The investment landscape is complicated by the need for continuous infrastructure upgrades amid these uncertainties.