ChainCatcher report: Venture capitalist Tomasz Tunguz wrote that the AI infrastructure narrative resembles a slow relay race, with bottlenecks sequentially propagating from GPUs to memory, CPUs, and storage—each stage freezing the supply chain of the next for years and locking in higher baseline costs. In early 2023, the GPU surge drove H100 rental rates above $9 per hour and reduced server shipments by 22%. Subsequently, manufacturers shifted capacity toward HBM, causing enterprise SSD prices to rise 80% in a single quarter and DRAM prices to increase over 60%. By the end of 2025, agent workloads will push the CPU-to-GPU ratio in servers to approximately 1:1, raising the average server CPU price by 27% year-over-year; in 2026, nearline HDD annual capacity will be fully sold out. Data center construction costs have risen to approximately $20 billion per gigawatt, with long-lead-time equipment such as transformers and turbines already booked through 2029 to 2031. Tunguz calls this the bullwhip effect in hardware: multi-year manufacturing lags amplify downstream demand shocks upstream; when pressure eases at one bottleneck, it is delayed in propagating to the next. Transformers expanded for 2027–2028 delivery, NAND wafer fabs, and turbine production lines may face overcapacity risks.
Tomasz Tunguz: AI Infrastructure Faces 'Whiplash Effect' as Bottlenecks Drive Costs
ChaincatcherShare
Summary
Tomasz Tunguz warns that AI infrastructure is facing a "whiplash effect," with bottlenecks in GPUs, memory, CPUs, and storage driving up costs. The Fear and Greed Index for AI hardware is near all-time highs. In Q1 2023, H100 GPU prices rose above $9/hour, and server shipments declined by 22%. The shift to HBM caused SSD prices to surge 80% in a single quarter. DRAM prices increased over 60%. CPU-to-GPU ratios are projected to reach 1:1 by late 2025, with server CPU prices up 27% year-over-year. Nearline HDD capacity is fully booked through 2026. Data center costs have reached $2 billion per gigawatt. Orders for transformers and turbines are scheduled through 2029–2031. Altcoins to watch may benefit from AI-driven shifts in demand. Overcapacity risks in NAND and turbines could emerge by 2027–2028 if bottlenecks persist.
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information.
Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.