Tomasz Tunguz Warns AI Infrastructure "Bullwhip" Dynamics Are Cementing Higher Costs
AI Market Summary
Tomasz Tunguz argues AI infrastructure is experiencing a "bullwhip" cycle where sequential bottlenecks (GPUs, HBM/DRAM, SSDs, CPUs, HDDs, and grid equipment) lock in higher baseline costs and extend lead times. The narrative implies persistent capex pressure for data centers and shifting profit pools across the supply chain. It also flags potential overcapacity risk later in the cycle (e.g., transformers, NAND fabs, turbines) once constraints ease.
Impact level
● Medium
Affected assets
NCSKNVDA2USD/USDT-2.25%
AI Insight · NCSKNVDA2USD/USDTAI Insight
● Neutral
Trade now
⚠️ AI-generated insights are based on news content and are provided for informational purposes only. They do not constitute investment advice or represent the views of BingX. Investing involves risk. Please trade responsibly.
ChainCatcher reports that venture capitalist Tomasz Tunguz said the buildout of AI infrastructure is unfolding like a slow relay race, with constraints moving in sequence from GPUs to memory, CPUs, and storage. Each chokepoint can stall the next link in the supply chain for years, embedding a higher cost floor across the stack.
Tunguz noted that the early-2023 GPU crunch pushed Nvidia H100 rental rates above $9 an hour and cut server shipments by 22%. As suppliers redirected capacity toward high-bandwidth memory (HBM), prices surged elsewhere: enterprise SSD prices jumped 80% in a single quarter, while DRAM prices rose more than 60%.
Looking ahead, he expects agent-style workloads to drive the CPU-to-GPU ratio in servers to roughly 1:1 by the end of 2025, lifting average server CPU prices 27% year over year. In 2026, he wrote, annual nearline HDD capacity is likely to be fully sold out.
On the facilities side, data center construction costs have climbed to about $20 billion per gigawatt. Long-lead items including transformers and turbines are already booked out through 2029 to 2031.
Tunguz described the pattern as a hardware bullwhip effect: multiyear manufacturing lead times amplify downstream demand shocks upstream, and when one constraint eases, relief reaches the next bottleneck only with a delay. He added that capacity expansions for transformers slated for 2027–2028 delivery, NAND wafer fabs, and turbine production lines could ultimately face overcapacity risk.
