Semi Doped

Wayne Nelms, CTO of Ornn and former quantitative trader, is building a financial exchange to trade GPU compute and memory as standardized commodities. He discusses pricing dynamics, hedging with futures, memory trading, depreciation and residual-value products. Short, forward-looking conversation about how financial markets could reshape GPU utilization, data-center finance, and hardware lifecycle risk.

Val Bercovici, Chief AI Officer at Weka and former NetApp CTO, explains AI-native storage and context memory challenges. They explore memory tiering from HBM to NVMe. Discussions cover latency, high-bandwidth flash, dedicated context memory networks, Axon pooling local NVMe into memory, augmented memory grids, token warehouses, and networking innovations for AI infrastructure.

They dig into the rise of agentic AI and why 2026 could be a turning point. The conversation covers Claude Code, OpenClaw, and practical tool integrations like Google Drive. Security tradeoffs of giving agents data and safer local deployment options get attention. They also explore rising CPU demand, shifting workloads off GPUs, and infrastructure choices for running agents.

Saurabh Dighe, Corporate VP leading Microsoft’s custom AI accelerator efforts, walks through Maia 200 and why it was built for inference. He discusses the design tradeoffs that target price-performance, the scale-up architecture, on-die NICs and large SRAM plus HBM, and how Maia will be delivered across Microsoft’s fleet and partners.

They dig into where the KV cache lives as AI demands week‑long, massive context runs. They debate whether SRAM-heavy accelerators like Cerebras can avoid offloading to HBM or external memory. They explore heterogeneous compute strategies and whether pre‑GPT chips will converge with GPUs. They spotlight next‑gen transformer-first accelerators to watch in the race to solve long‑context workloads.

Omer Keilaf, CEO of Innoviz, shares insights on his company, a leader in the LIDAR market. He discusses how Innoviz secured partnerships with major players like BMW and Volkswagen within a year. The conversation dives into the drive for Level 3 and 4 autonomy, how the automotive-first strategy sets them apart, and the urgent need for LIDAR resilience. Omer also highlights innovative designs for behind-windshield solutions, along with untapped opportunities in non-automotive sectors. His unique perspective paints a vibrant picture of the future of LIDAR technology.

Explore the fascinating world of LiDAR technology and its pivotal role in autonomous vehicles. Dive into the differences between Time of Flight and FMCW systems, and learn why wavelength choices impact performance and safety. Hear about how LiDAR has evolved, the competition in the market, and the significance of Chinese suppliers amid supply chain risks. Plus, find out how LiDAR performs better in extreme conditions compared to traditional cameras. The discussion reveals trade-offs between 905 nm and 1550 nm wavelengths, shedding light on this vital tech.

The hosts dive into NVIDIA's CES 2026 keynote, highlighting the revolutionary DGX Spark, which brings server-class architecture to desktops. They explore the implications of on-prem AI and the historical evolution from mainframes to minicomputers. Discussions on the Vera Rubin platform showcase its integrated architecture and the rising importance of CPUs for agentic workloads. The Rubin GPU is designed for efficiency, with cutting-edge HBM4 technology. They also analyze market reactions to NVIDIA's announcements and competitive dynamics with AMD.

Austin and Vik explore groundbreaking insights from the IEDM conference. They discuss innovations like penta-level NAND flash memory, which could disrupt the SSD market, and the power efficiency enhancements of GaN-on-silicon chiplets. The future of transistors shines bright with expectations of further advancements. The challenges in scaling silicon photonics and the vital role of complementary FETs in creating denser logic layouts are also highlighted. True to the nature of such gatherings, networking and collaboration opportunities take center stage.

The discussion dives into Nvidia's unconventional acquisition of Groq and the confusion it sparked. Expectations around GPUs and HBM are challenged as the hosts explore Groq's ultra-low latency architectures and the realities of SRAM versus HBM. They highlight unique use cases for LPUs, from ad personalization to real-time translation in robotics. Insights into Nvidia's strategy reveal an expanding focus on workload-specific optimizations, affirming that while GPUs aren't obsolete, LPUs serve a distinct purpose in inference technology.