Hardware Podcasts

Episode 105: Sanity Check on Custom ASICS?

The Circuit · 17 Feb 2025 · 41m · Episode 106

What this episode covers

Ben Bajarin and Jay Goldberg discuss the debate between custom ASICs and GPUs in AI compute, arguing that GPUs retain significant value due to software flexibility and installed base. They analyze the uneven maturity of hyperscaler silicon efforts, noting Google's TPU success versus the struggles of Amazon's Trainium and Meta's MTIA. The conversation then shifts to analog semiconductors, highlighting their critical role in humanoid robotics and data center power management, with specific mentions of Infineon's inventory buildup linked to NVIDIA's Blackwell.

Most relevant to

large enterprises

Topics

Mentioned

Companies
NVIDIA, AMD, Google, Amazon, Meta, Microsoft, Intel, Texas Instruments, Infineon, NXP, Analog Devices, Deep Seek
Products
TPU, Trainium, Inferentia, MTIA, Maia, Blackwell, CUDA, PyTorch

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

  • EE Times Current · 49m

    Can the Nvidia Monopoly on AI Chips Be Broken?

    Sally Ward-Foxton of EE Times discusses the AI chip market with hosts Dr. Sunny Bains and Dr. Giulia D’Angelo, analyzing the dominance of Nvidia and the potential for neuromorphic and specialized inference chips to break its monopoly. The conversation covers the spectrum of AI hardware from data center GPUs to ultra-low-power edge devices, highlighting the critical role of software stacks and latency in determining adoption. Professor Ralph Etienne-Cummings joins to critique the industry's focus on power versus learning capabilities and to discuss the future of neuromorphic computing, including memristors and photonic integration.

    Chips, FPGAs & semiconductorsAI & machine learning
    9 chapters

  • Oxide and Friends · 1h 54m

    Heterogeneous Computing with Raja Koduri

    Raja Koduri, former chief architect at Intel and AMD, joins Bryan Cantrill and Adam Leventhal to explain the differences between CPUs, GPUs, FPGAs, and ASICs through historical and analogical lenses. Koduri traces the evolution from 3dfx and Direct3D to CUDA, arguing that NVIDIA’s success stems from a tight coupling of programming and execution models, which creates a sticky ecosystem. He uses transportation analogies to categorize these technologies—cars for CPUs, buses for GPUs, and construction vehicles for FPGAs—and discusses how advanced packaging and memory bandwidth are becoming the primary constraints in modern AI hardware.

    Chips, FPGAs & semiconductorsMicrocontrollers & dev platforms
    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

  • Built 2 Scale · 37m

    AI NEWS | Episode 31

    Hosts Matt Perrott and Scott Wilcox analyze Google's Gemini 3 release, highlighting its integration with the Google ecosystem and the strategic shift to proprietary TPU chips to reduce reliance on Nvidia. They discuss the diminishing returns of scaling LLMs, citing Ilya Sutskever's view that the 'age of scaling is over,' and explore the necessity for application-layer companies like Lovable to niche down. The conversation concludes with a debate on the importance of Silicon Valley for AI talent versus the viability of distributed global teams and non-frontier AI strategies.

    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.