Beckmann

Dossier · Ongoing

Memory prices drive up AI costs

From Samsung's HBM4 surcharge in July to Nvidia's server price hike in August 2026: how scarce memory is making the AI supply chain more expensive.

Since mid-2026, a single component has been pushing up costs across the entire AI industry: memory. With data centers, smartphone makers and PC manufacturers competing for the same production capacity, the three major suppliers Samsung, SK Hynix and Micron have regained their bargaining power and are passing higher prices on to their customers.

This dossier tracks how those surcharges move through the supply chain step by step – from memory manufacturers through chip and server makers to cloud providers and their business customers. The decisive question remains open: where in the chain do the extra costs come to rest, and when will they become visible to companies in the price of AI services?

Timeline

  1. Samsung Raises AI Memory Prices by Up to 20 Percent

    Tight production capacity for HBM4 chips is pushing AI memory prices higher, and analysts expect a sharp increase by 2027.

  2. Nvidia raises prices for AI servers by more than 15 percent

    Sharply rising memory chip prices are pushing up Nvidia's AI server prices from 2027 – major customers have already been notified.