AI Infrastructure Silicon Valley - Executive Series 2024
2024-08-16
Palo Alto, United States
who it's for
ai infrastructure architects, data platform engineers, systems architects, silicon engineers, machine learning platform engineers, cloud infrastructure engineers, edge and embedded engineers
topics
ethernet fabric auto-tuning; gpu telemetry correlation; end-to-end streaming immutable log; model-to-data architectures; Apache Iceberg table format; cpu matrix extensions; kernel-level AI optimizations; at-memory computing with sram; optical compute interconnect; disaggregated HBM; CXL and NVMe controller platforms; edge inference for sensor analytics
| title | author | length |
|---|---|---|
| Raj Yavatkar, Juniper Networks | AI Infrastructure Silicon Valley - Executive Series | Raj Yavatkar | 18m |
Claims a working 800 gig switches fabric plus intent-based Apstra automation that runs auto-tuning application to adjust buffer thresholds and congestion… | ||
| Alexander Gallego, Redpanda Data | AI Infrastructure Silicon Valley - Executive Series | Alexander Gallego | 27m |
Claims that infrastructure must move from batch to realtime and that deploying models next to data — not shipping data to foundation models — is the enterprise… | ||
| Ami Badani, ARM | AI Infrastructure Silicon Valley - Executive Series | Ami Badani | 20m |
Arm argues CPUs must run more AI and shows concrete moves: matrix extensions for faster vector math, Kleidi AI kernel‑level optimizations tied into… | ||
| Bruno Geoly, Lumindt | AI Infrastructure Silicon Valley - Executive Series | Bruno Geoly | 15m |
Lumindt describes a compact, solid‑state hydrogen black box that stores hydrogen in a metal “sponge” and discharges through a fuel cell; a 30 kWh system is… | ||
| Chris Walker, Untether AI | AI Infrastructure Silicon Valley - Executive Series | Chris Walker | 18m |
Claims a reduction of the memory wall by embedding custom SRAM next to compute to avoid the 80–90% energy lost moving data, delivering at-memory computing with… | ||
| Craig Walker, Dialpad | AI Infrastructure Silicon Valley - Executive Series | Craig Walker | 27m |
Acquires TalkIQ in 2018 and embeds real-time ASR into the stack, claiming seven billion minutes of business-conversation training data and 65 full‑time AI… | ||
| Dave Lazovsky, Celestial AI | AI Infrastructure Silicon Valley - Executive Series | Dave Lazovsky | 22m |
Celestial AI describes a validated Photonic Fabric that uses silicon photonics plus TSMC 4nm control CMOS to replace copper interconnects and enable optical… | ||
| Gopi Sirineni, Axiado | AI Infrastructure Silicon Valley - Executive Series | Gopi Sirineni | 23m |
Company builds a hardware TCU/SCM single‑chip that replaces four‑to‑five chips, made at TSMC 12nm and shipped as a modular NIC/card that plugs into NVIDIA… | ||
| Hao Zhong, Scaleflux | AI Infrastructure Silicon Valley - Executive Series | Hao Zhong | 18m |
Company builds controller chips plus firmware and software that put flash and DRAM closer to CPUs, shipping designs manufactured at TSMC on multiple geometry… | ||
| Honghao Deng, butlr.io | AI Infrastructure Silicon Valley - Executive Series | Honghao Deng | 17m |
MIT spinoff builds a tiny wireless thermal sensor that measures ambient and surface temperature and runs edge models to produce high‑fidelity indoor location… | ||
| Krishna Rangasayee, SiMa.ai | AI Infrastructure Silicon Valley - Executive Series | Krishna Rangasayee | 27m |
SiMa.ai describes a purpose-built ML system-on-chip plus a software platform that aims to run multimodal AI at the embedded edge with 50+ customers and ~$270M… | ||
| Larry Yang, Phononic | AI Infrastructure Silicon Valley - Executive Series | Larry Yang | 13m |
Company builds solid‑state thermoelectric modules and reports 20 million shipped for optical transceiver laser cooling, claims compatibility with air and… | ||
| Rodrigo Liang, SambaNova Systems | AI Infrastructure Silicon Valley - Executive Series | Rodrigo Liang | 11m |
SambaNova Systems claims a native Dataflow chip architecture with TSMC manufacturing that compiles Hugging Face models without an ISA, and ships full-stack… | ||
| Sid Sheth, d-Matrix | AI Infrastructure Silicon Valley - Executive Series | Sid Sheth | 24m |
Company builds inference-focused silicon and accelerator cards that target generative AI by combining four compute engines — a licensed RISC-V data mover, an… | ||
| Charles Giancarlo, Pure Storage | AI Infrastructure Silicon Valley - Executive Series | Charles Giancarlo | 27m |
Claims that writing speed, not just read throughput, is the critical bottleneck for large model training because frequent checkpointing bottleneck pauses cost… | ||
| Vijay Nagarajan, Broadcom | AI Infrastructure Silicon Valley - Executive Series | Vijay Nagarajan | 35m |
Broadcom describes how Wi‑Fi evolution targets latency and capacity for AI and AR/VR, not just raw speed: Wi‑Fi 7 adds Multilink Optimization (MLO) for… | ||
| Shehram Jamal, NVIDIA | AI Infrastructure Silicon Valley - Executive Series | Shehram Jamal | 33m |
Defines accelerated computing as specialized hardware like GPUs and TPUs replacing one-size-fits-all CPUs, claims inference scales to dominate even as training… | ||
| nothing matches | ||