Hardware Podcasts
The Circuit podcast cover art

The Circuit

Ben Bajarin and Jay Goldberg on the business and market of semiconductors: chip makers, memory, AI silicon and the supply chain behind them, with interviews from across the industry.

189 episodes · 139 hours · since 2022

Visit The Circuit

The Circuit is filed under electronics, hardware, supply chain & operations and interviews. Its most recent episode published on , so it is still actively publishing. Listeners have left 45 ratings on Apple Podcasts, averaging 4.8 out of 5.

Latest episodes

The 10 newest episodes of The Circuit, each with a summary and chapters.

  • · 1h 00m

    Unpacking Marvell's Analyst Day, OpenAI Revenue Rumors, and the Heterogeneous Rack

    Ben Bajarin and Jay Goldberg analyze Marvell's recent Analyst Day, highlighting its strategic shift from Amazon ASIC share battles to a broader data center infrastructure play involving optical interconnects, switching silicon, and new AI infrastructure management silicon. They discuss the fragility of the AI semiconductor trade, noting that physical infrastructure bottlenecks, specifically the U.S. grid's ability to energize 20-30 gigawatts, pose a greater risk to chip timelines than end-user demand. Finally, they explore a strategic theory that TSMC is granting AMD extra wafer capacity to potentially throttle Intel's cash flow as Intel Foundry becomes a credible competitive threat.

    Chips, FPGAs & semiconductorsSourcing & supply chain
    7 chapters

  • · 58m

    Has the Memory Cycle Broken? Plus: Nvidia & Broadcom’s $40B+ Financing Playbook

    Hosts Ben Bajarin and Jay Goldberg analyze whether the memory market has structurally changed or remains cyclical following Micron's earnings, debating the impact of CXMT's rapid capacity expansion and the trade-offs between HBM and legacy memory. They examine HPE's strategic pivot to networking post-Juniper acquisition, questioning its unique value proposition amid security concerns driven by agentic AI. The discussion then shifts to 'Balance Sheet as a Service,' detailing Nvidia's vendor financing disclosures and Broadcom's $42 billion loan commitment to Anthropic, analyzing why fabless designers are taking on capital risk instead of foundries like TSMC. Finally, they explore 'Tokenomics,' advocating for quantifying AI unit economics to move past hype and understand the hard-dollar value of input/output tokens and hardware optimization.

    Pricing, margins & unit economicsSourcing & supply chain
    6 chapters

  • · 50m

    Memory Pricing Pressures, Meta's Muse AI, and the Agentic CPU Crunch

    Ben Bajarin and Jay Goldberg discuss the semiconductor memory market, noting that Chinese providers YMTC and CXMT are ramping capacity and improving yields, which may offer pricing relief to OEMs squeezed by data center demand. They analyze Meta's new 'Muse' AI assistant as a potential catalyst for mainstream adoption by bridging the user skill gap through proactive orchestration, while also highlighting the massive, unforeseen surge in CPU demand required to power these continuous 'agentic' workloads. Finally, they unpack Anthropic's $11 billion equity-exchange partnership with Akamai, highlighting its implications for distributed edge inference and the critical need for CPU infrastructure.

    Chips, FPGAs & semiconductorsAI & machine learning
    4 chapters

  • · 50m

    Conversation with Ayar Labs CEO Mark Wade on Silicon Photonics, Co-Packaged Optics, and Next-Gen AI Infrastructure

    Mark Wade, CEO of Ayar Labs, and Jesse Leiter, VP of Capital Strategy, discuss the company's strategy to solve the interconnect bottleneck in AI data centers using co-packaged optics (CPO). Wade explains the technical complexity of integrating silicon photonics into TSMC's advanced CMOS nodes and the strategic importance of partnerships with NVIDIA, AMD, and Intel. Leiter details Ayar Labs' funding journey, including a recent $650 million raise and strategic investments from major tech giants, positioning the company for an eventual IPO in a market projected to reach $30-50 billion by 2030.

    Chips, FPGAs & semiconductorsScaling production
    9 chapters

  • · 30m

    EP 192: Talking Apple Silicon with Apple's Tom Boger, Kaiann Drance, and Sri Santhanam

    Tom Boger, Kaiann Drance, and Sri Santhanam from Apple discuss the transition to 2-nanometer process technology across the A20 Pro and M-series chips, highlighting the shift to a dual neural engine and unified memory architecture. They explain how vertical integration allows Apple to optimize packaging, such as moving DRAM off the SoC to improve thermal management and sustained performance for AI workloads and gaming. The panel also details the implementation of the Secure Enclave and Secure Exclave to handle computationally expensive cybersecurity tasks and protect sensitive data at the silicon level.

    Microcontrollers & dev platformsAI & machine learning
    8 chapters

  • · 51m

    Ep 191: Agents Gone Wild: Broadcom’s ASICs and the AI Frontier

    Ben Bajarin and Jay Goldberg discuss Broadcom's recent earnings, highlighting the shift in custom ASIC customers from Google to Anthropic and OpenAI, and the increasing complexity of modeling mixed-chip programs involving competitors like MediaTek and Marvell. They debate the semiconductor supply chain, with Bajarin predicting peak constraints in 2027 and capacity relief in 2028, while Goldberg warns that this oversupply could shift pricing power and compress margins. The hosts also analyze Nvidia's $13 billion acquisition of Hugging Face, arguing it secures Nvidia's position in AI software without locking in open models, and conclude that both open and closed frontier models will coexist, with frontier models remaining essential for enterprise cybersecurity and national security.

    Chips, FPGAs & semiconductorsSourcing & supply chain
    7 chapters

  • · 53m

    EP 190: NVIDIA and Marvell Earnings, Hot Chips Hot Takes

    Ben Bajarin and Jay Goldberg analyze NVIDIA’s earnings, highlighting a 70% revenue growth guide driven by supply constraints and TSMC capacity, while noting margin pressure from rising memory and component costs. They discuss Marvell’s strategic focus on XPU-attached silicon and its $1.2 billion Google deal, contrasting its 60% growth with NVIDIA’s aggressive outlook. The hosts also cover the Hot Chips conference, describing an 'era of ferment' with competing inference architectures, custom HBM base dies, and workload-specific silicon from companies like Waymo and AMD.

    Chips, FPGAs & semiconductorsPricing, margins & unit economics
    9 chapters

  • · 38m

    EP 189: Analyzing ADI Earnings, Custom Silicon Shifts, and Semiconductor Debt

    Hosts Ben Bajarin and Jay Goldberg analyze Analog Devices' (ADI) recent earnings, highlighting a doubling of data center revenue driven by 800-volt architectures and high gross margins, while noting their conservative focus on Serviceable Addressable Market (SAM). They discuss Marvell's strategic win with Google for custom silicon periphery and optical interconnects, clarifying that this diversifies Marvell's customer base beyond Amazon without significantly impacting Broadcom's core TPU business. Finally, the hosts examine the emerging trend of 'balance sheet competition,' where chipmakers like NVIDIA and Broadcom use Special Purpose Vehicles (SPVs) and debt to fund data center capacity, a risky departure from the industry's historical low-debt model.

    Analog & power electronicsChips, FPGAs & semiconductors
    8 chapters

  • · 56m

    EP 188: The Wild West of Optical, Cisco's AI Hack, and Is Compute an Investable Asset

    Hosts Ben Bajarin and Jay Goldberg analyze the optical networking sector, contrasting Lumentum's success with Coherent's struggles in transitioning to 6-inch indium phosphate wafers, while debating the adoption timelines for Near-Packaged Optics (NPO) versus Co-Packaged Optics (CPO). They discuss Cisco's resurgence driven by its Silicon One chips and a strategic partnership with Anthropic to use the Mythos model for security audits, and examine Applied Materials' strong earnings backed by Long-Term Agreements (LTAs), attributing guidance hesitation to Intel's pending capital plans. Finally, the hosts critique NVIDIA's $500 billion data center financing initiative and Jensen Huang's claim that compute is a non-obsolescent asset class, arguing that rapid hardware obsolescence makes such an investment model risky.

    Chips, FPGAs & semiconductorsSourcing & supply chain
    5 chapters

  • · 40m

    EP 187: Inside the Future of Memory Summit & AMD’s "Boring" Bullishness

    Ben Bajarin and Jay Goldberg analyze AMD's recent earnings, noting that while the execution was solid and data center CPU demand doubled year-over-year, the market reaction was muted due to conservative guidance and high capital expenditure expectations. Ben reports from the Future of Memory Summit, highlighting the maturation of the CXL (Compute Express Link) ecosystem, driven by hyperscalers' need to reuse legacy DDR4 memory for inference workloads. The hosts also discuss the durability of long-term agreements (LTAs) in the memory supercycle, citing Sandisk's strong margins, and identify SiTime as a key beneficiary of the rack-scale AI infrastructure boom.

    Chips, FPGAs & semiconductorsSourcing & supply chain
    6 chapters

If you like The Circuit

Popular

The Amp Hour

Electronics design, manufacturing, and entrepreneurial discussion with a focus on hardware development.

hardwareelectronics
Visit Website

EMS@C-LEVEL

Interviews with executives of electronics manufacturing services companies on their businesses, their customers and the state of the contract electronics manufacturing industry.

manufacturingsupply chain
Visit Website

Catalyst

Latitude Media's weekly interview show on the energy transition, covering the technologies, costs and investment behind clean power, the grid, fuels and decarbonization.

climate energysupply chain
Visit Website
Popular

Embedded.fm

Discussions with experts about embedded systems, electronics, and hardware development.

electronicsembedded
Visit Website