AI Infrastructure Costs Rise Due to Supply Chain Bottlenecks

Venture capitalist Tomasz Tunguz has highlighted the ongoing challenges in AI infrastructure, likening the situation to a slow relay race where bottlenecks in one area impact subsequent components. This sequential transmission of issues has resulted in increased costs across the supply chain, particularly affecting GPUs, memory, CPUs, and storage.

At the start of 2023, a significant spike in GPU demand led to rental prices for H100 units exceeding $9 per hour, while server shipments dropped by 22%. In response, manufacturers redirected their production capacity towards high-bandwidth memory (HBM), which contributed to an 80% increase in enterprise SSD prices and over a 60% rise in DRAM prices in the following quarter.

Looking ahead, Tunguz predicts that by the end of 2025, the workload of intelligent agents will necessitate a CPU to GPU ratio of approximately 1:1, resulting in a projected 27% year-on-year increase in server CPU prices. Additionally, the construction costs for data centers have surged to around $20 billion per gigawatt, with orders for essential long-cycle equipment, such as transformers and turbines, extending into 2029 and beyond. Tunguz refers to this phenomenon as the 'long whip effect,' where manufacturing delays exacerbate demand shocks, causing further complications in the supply chain.

© 2026 KLEA News. All Rights Reserved. This article is provided for informational purposes only. It is not offered or intended to be used as legal, tax, investment, financial, or other advice.

Source: KLEA News