13 Feb 2026
58m

Memory Mayhem & AI Capex Madness

Podcast cover

Semi Doped

The podcast explores the escalating costs of memory in semiconductors and its wide-ranging effects on the tech industry. It highlights how the increasing demand for high-bandwidth memory (HBM) in AI accelerators is squeezing the supply and driving up prices of DRAM for consumer devices like gaming GPUs. This shift is incentivizing memory manufacturers and companies like NVIDIA to prioritize AI chips over consumer products due to better profit margins. The conversation touches on potential solutions like context memory and optics, but emphasizes HBM's unbeatable performance for AI applications. The hosts also analyze the capital expenditure plans of major hyperscalers like Google, Meta, and Amazon, discussing how their investments in AI infrastructure are reshaping the semiconductor landscape.

Outlines

Part 1: Memory Crisis, Market Shifts

Part 2: AI Architecture, Inference Costs

Part 3: Hyperscaler CapEx, Google Strategy

Part 4: Microsoft, Meta, Amazon Analysis

Part 5: Future Outlook, Industry Impact

Sign in to continue reading, translating and more.

Open full episode in Podwise