Hardware Podcasts

EP 138: Talking GPU Depreciation, OCP Takeaways

The Circuit · 20 Oct 2025 · 51m · Episode 140

What this episode covers

Ben Bajarin and Jay Goldberg discuss the debate over GPU depreciation cycles, arguing that while NVIDIA's Blackwell offers significant performance gains, physical infrastructure limits and economic viability suggest a 4-5 year lifecycle rather than the feared 2-year cycle. Goldberg shares takeaways from the Open Compute Project (OCP) forum, highlighting the rise of 'neoclouds' as a key growth driver for hardware vendors and the strategic pressure on hyperscalers like Amazon and Microsoft to maintain market share. The hosts also analyze the supply chain constraints in power equipment, noting severe backlogs for transformers and gas turbines, and touch on TSMC's cautious approach to capital expenditure despite record demand.

Most relevant to

large enterprises

Topics

Mentioned

Companies
NVIDIA, Open Compute Project, Cisco, Rista, Google, Oracle, XAI, Aligned Data Centers, BlackRock, Microsoft, Honeywell, TSMC, ASML
Products
Blackwell, Hopper, A100, Vera Rubin, VMware, Lepton, AWS, Azure

Related episodes from other shows

Other shows on the same ground, closest first.

  • Built 2 Scale · 1h 08m

    Google TPU vs Nvidia GPUs: Token Per Watt Dominance, or the Era of Scaling Is Over?

    Founders Matt Perrott and Scott Wilcox discuss Google's Gemini 3.0 training on TPUs, which bypasses Nvidia's hardware to achieve a structural cost advantage, and the resulting threat to 'vibe coding' startups like Lovable. They analyze the 'age of scaling is over' consensus among researchers like Ilya Sutskever, arguing that future breakthroughs require real-world spatial intelligence rather than just more compute. The hosts also debate office culture, remote work dynamics, and the three critical roles founders must hire early: product, brand, and sales.

    Chips, FPGAs & semiconductorsAI & machine learning
    6 chapters

  • Catalyst · 34m

    Who benefits from the AI power bottleneck?

    Shanu Mathew, senior vice president and portfolio manager at Lazard, analyzes the conflicting incentives of various actors in the AI energy supply chain regarding the 'power bottleneck' narrative. He explains how hyperscalers like Microsoft and Amazon, equipment makers like Quanta and Vertiv, and land developers benefit from amplifying scarcity to drive demand or secure competitive advantages. Conversely, he identifies independent power producers (IPPs) like Constellation Energy and LNG providers as having incentives to downplay constraints to maintain high commodity prices or favorable arbitrage margins.

    Policy, trade & tariffsEnergy & climate tech
    6 chapters

  • The Amp Hour · 59m

    Back on the road in '26

    Chris Gammell of Contextual Electronics and Dave Jones of EEVblog discuss their upcoming travel to Embedded World and various hardware meetups, highlighting the challenges of organizing community events. They critique a TechCrunch article on solid-state transformers, arguing that traditional AC transformers remain superior for grid infrastructure due to efficiency and reliability. The conversation then shifts to the fragile global supply chain for AI hardware, tracing dependencies from ChatGPT back to ultra-pure quartz from Spruce Pine, North Carolina, and discussing the marketing tactics of Donut Lab regarding their battery technology. Finally, they review modular computing options like the Framework laptop and Flow Z13, while analyzing the latest developments in KiCad 10 and the Paduk microcontroller used by Rode.

    PCB design & layoutEnergy & climate tech
    8 chapters

  • EE Times Current · 29m

    How Much is Enough for Hyperscale Companies?

    Brian Santo hosts a discussion on how hyperscale companies like Google, Facebook, Amazon, and Microsoft are reshaping the semiconductor industry by demanding extreme power efficiency and designing custom AI accelerators. Experts Rick Merritt, Gary Hilson, and Dylan McGrath explain that these giants are pushing beyond standard roadmaps, collaborating directly with equipment vendors like Applied Materials, and leveraging their vast capital to drive innovation in memory and process technology. The episode concludes with Wally Rhines' insight that these companies' ultimate goal is owning the data and systems rather than just selling silicon, raising concerns about their growing societal power.

    Chips, FPGAs & semiconductorsSourcing & supply chain
    6 chapters

  • Oxide and Friends · 2h 01m

    Predictions 2025

    Bryan Cantrill and Adam Leventhal of Oxide Computer Company host a panel with Simon Willison, Steve Klabnik, Steve Tuck, and Mike Cafarella to review past predictions and forecast trends for 2025, 2028, and 2031. The group discusses the failure of agentic AI, the rise of AI efficiency and open weights models like DeepSeek, and the potential for a chip crisis due to geopolitical tensions and Intel's struggles. They also predict shifts in hardware markets, including NVIDIA's Blackwell struggles, AMD's software acquisitions, and the resurgence of manufacturing as a VC focus.

    AI & machine learningChips, FPGAs & semiconductors
    8 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by The Circuit. The episode itself belongs to its publisher.