Data from venture capital firm a16z reveals an intriguing trend in the AI market, highlighting a Jevons paradox. Despite falling prices of AI tokens, which represent the cost of accessing AI services, the rental prices for H100 GPUs—a key hardware component for AI workloads—have either remained steady or increased, according to The Decoder.
This phenomenon suggests that cheaper AI tokens are driving demand for AI services faster than the cost reductions in hardware can keep up. Jensen Huang, CEO of NVIDIA whose H100 GPUs are widely used in AI, is indirectly part of this dynamic where hardware costs do not decline in tandem with token prices.
For Japanese investors and technology firms, this disparity between AI service affordability and hardware pricing signals a complex landscape for AI adoption and infrastructure investment, emphasizing the need to monitor both software token markets and underlying hardware costs carefully.