Now Playing
Data Center Tariffs Bite, Apple M6 Debuts & BIS Targets Freight
Audio is available on Spreaker — see link below.
Phase 2 tariffs have stripped exemptions from AI data center infrastructure, rewriting hyperscaler capex models overnight. Plus: Apple's M6 and M5 Ultra chip architectures, China's domestic-chip inference claims, Intel's $2.1B foundry loss, and a landmark BIS freight-forwarder enforcement action.
Audio is available on Spreaker — see link below.
The Trump administration has removed the tariff exemptions protecting AI data center infrastructure, and the economics of US hyperscaler buildout just got significantly harder to model. Phase 1, announced in January, was a deliberate structural choice.
While tariff risk clouds hyperscaler spending, Apple moved in a different direction entirely. The M6 chip brings two-nanometer process technology to Mac for the first time, with a dual sixteen-core Neural Engine that doubles prior on-device AI compute.
Out of China, Zhipu AI is claiming something worth examining carefully. Its GLM-5.3-Flash inference system has reportedly served sixty-two trillion tokens using one hundred thousand domestically made processors, including chips from Huawei, Hygon, and Moore Threads.
The Bureau of Industry and Security has opened an investigation into Apex Logistics, a freight forwarder alleged to have filed false export classification codes to move Nvidia B300 servers from Taiwan through Hong Kong into China. What makes this significant isn't just the alleged violation.
Intel's foundry group reported a two point one billion dollar operating loss on five point eight billion dollars in Q2 revenue. External customers contributed only two hundred ninety-three million of that revenue.
Three things are worth tracking closely from here. First, whether Phase 2 tariff rates and carve-out categories get formally published, because until they do, hyperscaler capex planning remains structurally exposed.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.