Ask in Claude
Ask in ChatGPT
- Podcasts
- Technology
- Semi Doped
22 episodes
Semi Doped
by Vikram Sekar and Austin Lyons · English
The business and technology of semiconductors. Alpha for engineers and investors alike.
Latest episodes
All episodes
Jensen Huang Resigns
NVIDIA CEO Jensen Huang’s resignation and what it means for corporate governance, AI-hardware leadership, and the near-term “AI trade.” Guests: Austin Lyons (ChipStrat; newsletter for institutional…
12 Nov 2026 · 32m · 10 chaptersIntel's Earnings Disaster: Bad Luck, Bad Planning, or Both?
Intel’s stock plunge after earnings, despite “good” quarterly results, driven by weak guidance.
9 Nov 2026 · 57m · 22 chapters3D Stacked DRAM for Inference: d-Matrix's Sudeep Bhoja
The episode explains why AI inference is memory-bound (weights + KV cache must be read per token) and how D-Matrix builds SRAM-based in-memory inference (Corsair) to cut data-movement power.
5 Nov 2026 · 1h 16m · 33 chaptersWhy Marvell Wants Lumentum: Vertical Integration in AI Optics With Cignal AI
Marvell Technology in “advanced talks” to acquire Lumentum, a data-center optics/laser supplier, potentially announced next week.
30 Oct 2026 · 40m · 13 chaptersHow Intel Took EMIB From Science Fiction to High Volume: Intel Foundry's Ravi Mahajan
Intel Foundry’s EMIB (Embedded Multi-Die Interconnect Bridge) history and how Intel moved it from “science fiction” to high-volume manufacturing, plus comparisons to silicon interposers/COAs,…
28 Oct 2026 · 46m · 15 chaptersRF Over Plastic Is a Third Option Between Copper and Optics: Point2's Sean Park
Point2 Technology’s “RF over plastic” interconnect as a third option to copper and optics for scale-up data center networking, targeting 5–10m reach between racks using an active RF plastic waveguide…
23 Oct 2026 · 1h 6m · 26 chaptersMicroVCSELs vs MicroLEDs, a MicroExplainer
Semi-Doped “MicroVCSELs vs MicroLEDs, a MicroExplainer” compares micro-emitters for AI data-center interconnects, focusing on Credo’s shift from micro-LED expectations to micro-vixels (VCSEL-based)…
20 Oct 2026 · 39m · 13 chaptersAI Inference Is a Data Movement Problem: Microsoft's Saurabh Dighe on Maia 300
Microsoft’s Maia 300 AI inference system, framed as “AI inference is a data movement problem.” Saurabh Dighe argues Microsoft builds custom silicon-in systems to maximize token throughput per…
15 Oct 2026 · 49m · 14 chaptersSynopsys’s Amazon Deal and GPT-Synopsys: Ravi Subramanian
Synopsys investor-day announcements: a multi-year $1B+ Amazon deal for application-optimized custom semiconductor interface IP, and “GPT Synopsys,” a specialized OpenAI frontier model delivered as a…
8 Oct 2026 · 42m · 13 chaptersEvery Muse User Gets 2 CPUs and 8GB in Meta's Cloud. Your Phone Already Has That
Edge/phone AI vs cloud inference. The hosts argue that app companies (especially Meta) can subsidize “agentic AI” locally because they already control consumer distribution and can reduce cloud costs…
1 Oct 2026 · 51m · 21 chaptersFlash Hit Its Physical Limit. ReRAM Is the Replacement: WeeBit Nano's Ilan Sever
Non-volatile memory for embedded SoCs—why embedded flash (eFlash) fails to scale below ~22–28nm, and how WeeBit Nano’s OxRAM resistive RAM can replace it.
29 Sep 2026 · 57m · 23 chaptersQualcomm's Durga Malladi: HBC vs HBM, Dragonfly AI 250, and Winning on TCO
Qualcomm’s Durga Malladi explains “high bandwidth compute” (HBC), based on near-memory computing, and how it targets the “memory wall” in AI inference (especially decode).
24 Sep 2026 · 21m · 9 chaptersPacing AI Means More Compute, Not Less
The episode discusses Dario Amodei’s essay “We Must Pace the Frontier,” arguing frontier AI labs should slow capability development and invest more in safety—without pausing training entirely—because…
18 Sep 2026 · 37m · 17 chaptersTSMC Will Buy High NA After All: ASML's New 6x12-Inch Mask
ASML’s new high-NA EUV mask standard and why coordinated adoption by TSMC, Samsung, and Intel makes the high-NA ramp (and ASML’s 2030 financial model) more likely.
14 Sep 2026 · 50m · 23 chaptersOpenAI’s Jalapeño! Feeling Hot Hot Hot!
OpenAI’s “Jalapeno” inference chip announcement at Hot Chips, focusing on end-to-end user latency (time to last token) and energy per request, plus system architecture (HBM locality, KV-cache…
27 Aug 2026 · 55m · 20 chaptersGrok Bots and How CPUs are used in Agentic AI
Release of GrokBot, an “agentic AI platform” that runs agents in a dedicated cloud VM, and how that shifts compute demand toward CPUs (host-node CPUs vs “agentic CPUs”) alongside GPUs.
24 Aug 2026 · 45m · 23 chaptersTensordyne's R K Anand: HPE Juniper Fabric, Logarithmic Math, MoE Inference, Air Cooling, 3nm
TensorDine’s approach to high-interactivity LLM inference (especially 2T+ parameter Mixture-of-Experts) using logarithmic math, a new accelerator architecture, and a rack-scale “scale-up fabric”…
18 Aug 2026 · 56m · 19 chapters
NEWS TAKE: China's Optical Ban, Volta's $10B Anthropic Deal, AMD's Earnings
News roundup on (1) a proposed Trump administration ban on new U.S.
11 Aug 2026 · 35m · 15 chapters
GlobalFoundries Thomas Barber: CPO, Silicon Photonics, 300mm, SiGe, OCI, NRZ
GlobalFoundries’ role in silicon photonics and CPO/OCI for AI “scale-up” networks, including why 300mm manufacturing, near-pluggable-to-CPO packaging, and NRZ-based OCI matter for power, density, and…
7 Aug 2026 · 40m · 15 chapters
How Retimers Built an $80B Company: The Story of Astera Labs
How Astera Labs grew from a retimer startup into an $80B public company by solving copper PCB signal-loss problems inside AI server trays.
1 Aug 2026 · 48m · 18 chapters
News Take: Hyperscaler CDS, SK Hynix Earnings, China's DUV
Market news roundup on (1) hyperscaler credit risk via rising credit default swap premiums tied to AI GPU CapEx and “circular financing,” (2) SK Hynix earnings shock despite strong results, and (3)…
29 Jul 2026 · 18m · 7 chapters
Datacenter Interconnects: Copper vs. Optics, Nvidia's 78-Layer PCB, Co-Packaged Optics (CPO)
Data center interconnects for AI/LLM training and inference—how to connect tens to hundreds of thousands of GPUs using scale-up (within rack/near), scale-out (rack-to-rack), and scale-across…
25 Jul 2026 · 49m · 16 chaptersWe use cookies.They keep VO running smoothly and help us improve it. Privacy policy