The Circuit

In this episode of The Circuit, Ben Bajarin and Jay Goldberg explore the evolving dynamics of the semiconductor memory market, consumer AI agents, and critical compute infrastructure. The hosts begin by analyzing the aggressive capacity expansions and yield improvements of Chinese memory providers YMTC and CXMT, discussing how their growing output might provide pricing relief for consumer electronics OEMs currently squeezed by massive data center demand. The conversation then pivots to Meta's new "Muse" AI assistant, which the hosts view as a potential catalyst for mainstream AI adoption that could bridge the user skill gap by acting as a proactive personal and enterprise orchestrator. Finally, they unpack Anthropic's $11 billion equity-exchange partnership with Akamai, highlighting its implications for distributed edge inference and the massive, unforeseen surge in CPU demand required to power these continuous "agentic" workloads.

In this episode of The Circuit, hosts Ben Bajarin n and Jay Goldberg speak with Mark Wade, CEO of IR Labs, and Jesse Leiter, VP of Capital Strategy and Investor Relations. The conversation focuses on the necessity of solving the interconnect bottleneck in large-scale AI data centers by transitioning to optical connectivity. Mark explains the immense complexity of integrating optical engines into advanced CMOS nodes and highlights how the release of ChatGPT accelerated the market's demand for these large-scale computing solutions. Additionally, the discussion explores IR Labs' strategic reliance on leading-edge foundries like TSMC, their ecosystem partnerships with tech giants like NVIDIA, AMD, and Intel, and their calculated funding strategy aimed at an eventual IPO in a market projected to reach $30 to $50 billion by 2030

In this episode of The Circuit, host Ben Bajarin sits down with Apple's product marketing and engineering leaders at Apple Park to unpack their latest silicon and hardware announcements. The discussion explores the benefits of Apple’s vertically integrated approach, diving into the transition to 2-nanometer process technology and its impact on performance, power efficiency, and advanced thermal management across the A20 Pro and M-series chips. The panel also highlights the architectural decisions driving robust on-device AI capabilities—such as the dual neural engine and unified memory—and underscores Apple's deep commitment to hardware-level security and privacy through innovations like the Secure Enclave and Secure Exclave.

In this episode, Ben and Jay discuss the launch of their firm's new AI knowledge agent, Atlas, before diving into Broadcom's latest earnings and the growing complexity of modeling the custom ASIC market. They debate the semiconductor supply chain bottleneck, with Ben arguing that constraints will peak in 2027 before massive new capacity arrives in 2028, while Jay cautions that this incoming oversupply could shift pricing power and lower margins. Finally, they analyze Nvidia's acquisition of Hugging Face and explore the future of AI models, agreeing that while open and closed models will coexist, closed "frontier" models remain essential for enterprise cybersecurity and defending against rogue actors.

Ben Bajarin and Jay Goldberg break down NVIDIA’s earnings, examining how supply constraints shape growth and rising component costs pressure margins. They discuss NVIDIA’s financing commitments to neoclouds and the risks of using its balance sheet to support the ecosystem. Marvell’s XPU-attach strategy offers another route into custom silicon, with questions around when its Google opportunity will convert into revenue.
From Hot Chips, Ben shares why competing inference architectures suggest the industry is still working out how to serve AI at scale. The conversation explores the software and system-design challenges facing accelerator startups, along with custom HBM base dies and their implications for memory costs and supplier flexibility. Waymo’s workload-specific silicon raises another question: how much of inference will converge on a common architecture?