Business Shanghai's Yuanshi Technology Raises Nearly RMB 300 Million for AI Virtual Cells, Two Weeks After the Last Chinese Record Fell Security GAO Finds No Widely Accepted Test for Generative AI Cyber Skills and Hands Congress 16 Options, Including Human Sign-Off on Consequential Actions News Microsoft Turns Three USB4-Cabled PCs Into One AI Machine: Qwen3-235B Runs at Up to 22 Tokens a Second Across 384GB of Ordinary Memory Business Dell, HPE, Lenovo and Supermicro Are Demanding Up to 45% More for Signed AI Server Deals, and Trading Firms Are Paying Opinion OpenAI Scores Its Own Models on 99 Drug-Chemistry Problems, and Compound Selection Is Where They Stall Coding Assistants AWS Puts Distributed KV Caching Behind an Agent Skill: Up to 40% Cheaper Inference and 80% Faster First Token AI Agents Tencent and King's College London Build an Agent Memory Test From 80,299 Real Group-Chat Messages News Tesla Starts FSD v15 With a Generative Reality View and a 25x Intervention Claim It Has Not Published Business Shanghai's Yuanshi Technology Raises Nearly RMB 300 Million for AI Virtual Cells, Two Weeks After the Last Chinese Record Fell Security GAO Finds No Widely Accepted Test for Generative AI Cyber Skills and Hands Congress 16 Options, Including Human Sign-Off on Consequential Actions News Microsoft Turns Three USB4-Cabled PCs Into One AI Machine: Qwen3-235B Runs at Up to 22 Tokens a Second Across 384GB of Ordinary Memory Business Dell, HPE, Lenovo and Supermicro Are Demanding Up to 45% More for Signed AI Server Deals, and Trading Firms Are Paying Opinion OpenAI Scores Its Own Models on 99 Drug-Chemistry Problems, and Compound Selection Is Where They Stall Coding Assistants AWS Puts Distributed KV Caching Behind an Agent Skill: Up to 40% Cheaper Inference and 80% Faster First Token AI Agents Tencent and King's College London Build an Agent Memory Test From 80,299 Real Group-Chat Messages News Tesla Starts FSD v15 With a Generative Reality View and a 25x Intervention Claim It Has Not Published

Dell, HPE, Lenovo and Supermicro Are Demanding Up to 45% More for Signed AI Server Deals, and Trading Firms Are Paying

Dell, HPE, Lenovo and Supermicro Are Demanding Up to 45% More for Signed AI Server Deals, and Trading Firms Are Paying

The Financial Times reports Dell, HPE, Lenovo and Supermicro are asking 25-45% more at delivery than contracted prices for Nvidia AI servers, citing memory and circuit-board shortages. Market maker XTX says Dell sought an extra $62-70 million on a $549 million fixed-price order it had paid $5 million to lock.

Dell, HPE, Lenovo and Supermicro are asking customers to pay 25 to 45 percent more for Nvidia-based AI servers than the prices fixed in contracts they signed months earlier, the Financial Times reported, as a memory and circuit-board shortage shifts pricing power decisively toward suppliers.

The clearest test case is playing out in an Irish court. Market maker XTX Markets says it agreed a fixed price of $549 million for a batch of AI servers and paid an additional $5 million specifically to secure price certainty. According to the FT, Dell later told XTX the order would need another $62 million to $70 million to cover component-cost increases. XTX filed suit in Ireland in June; the claims remain contested.

The squeeze behind the repricing is severe. Research from JPMorgan cited by the paper puts memory cost inflation at more than 400 percent since early 2024. One person familiar with the orders told the FT that GPU-server prices rose at least 20 to 30 percent, while CPU-node prices tripled or more, because memory makes up a much larger share of a CPU machine's bill of materials.

HPE has turned the practice into an explicit contract term. Chief executive Antonio Neri told investors the company had "amended our quoting terms with a right to reprice existing orders for commodity cost increases between quoting and shipment," and said he expects elevated memory and storage prices to persist well into 2027. Neri has previously said memory now exceeds half the materials cost of a server. A global circuit-board shortage is reported to run into 2028.

The buyers absorbing the increases are not the usual cloud hyperscalers. The FT named quantitative trading firms Jane Street, Optiver and Hudson River Trading, which have been building large private data centers to run their own AI infrastructure, among those affected. The paper reported that firms have largely accepted the higher bills because scarce compute is still worth more than the surcharge.

Whether the suppliers are contractually entitled to reopen signed deals remains unclear. Buyers did not confirm their terms, and the XTX case may become the reference dispute: if an agreement described as fixed-price, with a premium explicitly paid for certainty, can still be reopened, then price certainty in AI infrastructure contracts means something very different than it did a year ago.

The episode also marks a second-order shift in how compute gets procured. If post-signing repricing becomes routine, memory, board and accelerator cost volatility will increasingly be written explicitly into AI hardware contracts rather than absorbed by vendors, turning server procurement into a component-hedging problem. For buyers outside the largest cloud platforms, the AI buildout is becoming a financing and negotiation problem as much as an engineering one.

Comments (0)

Log in to join the discussion

Log In

No comments yet