Google, Intel Forge AI Infrastructure

Google is betting billions on custom chips and cloud muscle to power the next wave of AI, striking multi-year deals with Intel and Broadcom to lock in hardware and scale out data centers. The stakes are high: Google’s partnership with Intel expands beyond traditional CPUs to custom Infrastructure Processing Units—hardware designed to handle the networking and storage bottlenecks that throttle AI workloads. With Anthropic reserving “multiple gigawatts” of Google Cloud capacity for its AI models and enterprise customers like Coinbase and Shopify coming aboard, the AI arms race is shifting from raw GPU power to smarter, more efficient systems. But Google’s gamble ties its fate to Intel’s execution and Broadcom’s manufacturing, raising the risk that supply chain hiccups or missed performance targets could slam its bottom line.

Here’s the catch: all this expansion hinges on the physical realities of data center power and global geopolitics. Iran’s recent threat to treat U.S.-based cloud data centers as military targets throws a wrench into how regions are built, insured, and secured—potentially driving up costs and forcing Google to redesign facilities to keep defense and civilian workloads apart. Meanwhile, scale commitments like Onix’s $500M Google Cloud deal promise faster AI deployment for big businesses, but actual value won’t be clear until those contracts renew and the real-world consumption comes in. All eyes are on whether Google’s new chip infrastructure makes it from slide decks to live data centers by year’s end.

Based on reporting from Reuters, TechCrunch, and Silicon Canals.

Powered by Apisod.com