Latest Interviews
Showing 76–90 of 304 transcripts.
Clear all filters- RAISE Summit42 min
Sovereign AI: Piece by Piece | NetApp, Radiant, Hydra Host, Coatue & More | RAISE Summit 2026
George Kurian, Mahdi Yahya, Aaron Ginn, Max Cook, Benjamin Germin, Thomas Koehrer, Thomas Kova, Maddy, Ben
Industry leaders convened to define sovereign AI as a spectrum of national control over critical infrastructure, data residency, and model provenance rather than a binary state. While experts debated the feasibility of achieving true autonomy through hyperscalers, they agreed that geopolitical constraints and physical hardware limitations are driving significant investments in localized data centers and domestic chip ecosystems. The consensus projects a bifurcated future where the US and China maintain full-stack dominance, compelling other nations to navigate a complex landscape of duopoly dependency or costly independent infrastructure development.
- RAISE Summit41 min
From Research to Reality: Why Open Source Is the Engine of AI | RAISE Summit 2026
Ion Stoica, Thomas Wolf, Dillon Rolnick, Eiso Kant, Robin Rombach, Mark Porter, Dylan Rolnick
Founders from Poolside, Databricks, Neuse Research, and Hugging Face, moderated by Mark Porter, convened to debate preventing AI monopoly through open-weight models rather than closed ecosystems. The panelists argued that open source remains the only viable path for critical sectors like robotics and scientific discovery while highlighting risks of geopolitical bias and regulatory lockout under current centralized control. Ultimately, the group emphasized a strategic shift toward commoditized infrastructure and real-world evaluation to avoid security fears and drive the next wave of physical and scientific breakthroughs.
- RAISE Summit17 min
Fireside Chat with Sid Sheth, Founder & CEO of d-Matrix | RAISE Summit 2026
D-Matrix, founded in 2019, has transitioned its SRAM-based Corsair accelerator into mass production while partnering with NVIDIA to deploy heterogeneous compute solutions for the inference-focused agentic era. Moving beyond current SRAM limitations, the company is developing a 3D stacked In-Memory Compute architecture scheduled for a 2027 tape-out, which aims to simultaneously optimize latency and throughput for high-interactivity workloads. This strategic trajectory envisions a market where specialized inference accelerators coexist with HBM-based GPUs, handling distinct layers of the AI pipeline from heavy pre-fill computations to memory-intensive decoding.
- RAISE Summit38 min
The Chip Revolution: Beyond GPUs to Tomorrow's Processors | RAISE Summit 2026
Robert Wachen, Sara Achour, Walter Goodwin, Antoine Moyroud, Rob
Industry experts from Unconventional.ai, Etched, and Fractile outline a pivotal shift where inference spending will soon surpass training, driven by the emergence of autonomous agents and a transition from GPU-centric compute to system-level challenges in memory bandwidth and thermal management. While Unconventional.ai bets on physics-based dynamics to achieve 1,000x energy efficiency, Etched pursues vertical integration for trillion-parameter models and Fractile optimizes for time as the primary bottleneck. These startups aim to capitalize on hardware design expansion and software automation to overcome current utilization limits that leave data centers operating at only 30-40% matrix multiply efficiency.
- RAISE Summit18 min
Fireside Chat with Philip Johnston, Starcloud & Chetan Puttagunta, Benchmark | RAISE Summit 2026
Philip Johnston, Chetan Puttagunta
StarCloud is a Seattle-based company that has successfully launched StarCloud 1, proving that terrestrial-grade GPUs can perform high-power training and inference in space while overcoming thermal and radiation challenges. The firm plans to rapidly scale its operations through partnerships with SpaceX and other launch providers, targeting a gigawatt-scale constellation of 88,000 satellites by late 2028 to provide near-instant data processing for government, telecommunications, and lunar exploration clients. With a founding team split between aerospace and cloud engineering giants, StarCloud aims to bypass terrestrial energy permitting delays by establishing a fully vertical infrastructure capable of delivering up to 10 terawatts of compute capacity.
- RAISE Summit38 min
AI Has a Memory Problem | Gelsinger, SK hynix & More | RAISE Summit 2026
Gelsinger, Hoshik Kim, Vijay Shilpiekandula, Andrew Homan, Louis Delahaye
The global semiconductor industry is currently navigating a critical bottleneck where exponential AI demand for larger context windows outpaces the linear supply growth of High Bandwidth Memory, leaving the market in a 26% deficit. While a duopoly of SK Hynix and Samsung dominates existing capacity, architects like Cerebras and D-Matrix are pioneering stacked memory solutions expected to boost bandwidth by 50 times, though these innovations will not reach volume production until the late 2020s. Consequently, memory has evolved from a volatile commodity into a strategic infrastructure asset, characterized by long-term contracts and deep supply chain vulnerabilities that threaten global stability amid geopolitical tensions and complex manufacturing constraints.
- RAISE Summit39 min
Welcome to the New Frontier of Agentic AI | Wiz, Merge, You.com & More | RAISE Summit 2026
Yinon Costica, Gil Feig, Richard Rizk, Saahil Jain, Ed Huang, Jordan Nanos
Organizations are shifting from single-model reliance to multi-model routing architectures driven by cost sensitivity, governance compliance, and task-specific performance benchmarks. Experts emphasize that successful agentic deployment now depends on engineering harness quality, stable data graphs, and end-to-end token-level encryption rather than raw model intelligence. As the industry transitions into an engineering challenge, future differentiation will stem from strategic tool integrations and robust governance layers that enable safe, long-horizon autonomous workflows.
- RAISE Summit39 min
Pilot to Production: The Valley of Death and How to Cross It | RAISE Summit 2026
Barak Kaufman, Jonathan Corbin, Euro Beinat, Ivan Burazin, Ariel Shulman, Venkat Pullela, Yvonne, Juro
A panel of industry leaders from companies including Keysight, Bright Data, and Prozus discussed the strategic pivot from AI proof-of-concept pilots to secure, scalable enterprise production. The group emphasized that overcoming legacy system bottlenecks and establishing robust data guardrails are essential for transitioning agents into critical workflows, citing Prozus's success with over 60,000 internally built agents. Furthermore, the discussion highlighted a shift in adoption strategies toward model-agnostic architectures and revenue-focused ROI metrics rather than mere cost reduction.
- RAISE Summit20 min
Beyond the Helpdesk: AI Turning IT Into a Strategic Function | Serval | RAISE Summit 2026
Serval is an AI-native platform that completely rebuilds IT Service Management infrastructure to replace legacy systems like ServiceNow by integrating dual-agent architecture for full automation of enterprise workflows. Co-founded by Jake, the company has driven an 80% automation rate for major clients such as SeatGeek and Ramp, enabling thousands of IT staff to shift from support roles to building cross-functional solutions in marketing, legal, and HR. By treating AI as a core foundation rather than an add-on layer, Serval allows enterprises to eliminate support queues and rapidly reinvent their operational roadmaps every six months.
- RAISE Summit17 min
Keynote by Sunghyun Park, CEO & Co-Founder of Rebellions | RAISE Summit 2026
Seoul-based Rebellion, led by CEO Sunghyun Park, is establishing itself as a mass-production alternative to NVIDIA by delivering air-cooled AI server racks that achieve three times lower capital expenditure and superior efficiency through a custom hybrid SRAM and HBM3e architecture. Backed by strategic investors including SK Hynix and Samsung Foundry, the company's full-stack platform currently processes 14 million daily API calls for mission-critical applications at SK Telecom while prioritizing native open-source software compatibility. By optimizing for tokens per second per watt per dollar, Rebellion aims to diversify the global supply chain with a heterogeneous compute solution that scales via modular chiplets for both inference and emerging AGI workloads.
- RAISE Summit13 min
Keynote by Maor Shlomo, Founder & CEO of Base44 | RAISE Summit 2026
Founded 18 months ago as a one-person "vibe coding" platform, Base44 recently secured a Wix acquisition despite generating over $150 million in annual recurring revenue without external investment. The company achieved 50x growth by restructuring around "Teams of Ones," where individual contributors leverage AI agents to handle end-to-end operations without traditional management layers. This model prioritizes compounding ambitions by reinvesting 30% of time into custom tooling, enabling a single entity to execute hundreds of daily deployments and expansive product initiatives that outpace legacy organizational structures.
- RAISE Summit21 min
Building Global out of Europe in the AI Era | Jan Oberhauser, n8n | RAISE Summit 2026
N8N positions itself as an open-source AI workflow orchestration layer that connects diverse models and applications, a strategy validated by adoption from major enterprises like Microsoft, SAP, and the German government. By prioritizing model agnosticism and a "fair code" license, the platform enables organizations to maintain data sovereignty and switch providers to optimize costs and mitigate supply chain risks. Supported by a global network of partners and community ambassadors, N8N continues to expand its enterprise footprint by evolving from a tool for technical builders into an accessible infrastructure for non-technical users to deploy robust, ROI-driven AI solutions.
- RAISE Summit21 min
The Foundation Model Revolution for Structured Data | Frank Hutter, Prior Labs | RAISE Summit 2026
Founded 18 months ago by Frank Hutter, Souraj Gambier, and Noah, Prior Labs has developed TabPFN, a tabular foundation model that processes structured enterprise data in seconds without retraining. The company achieved state-of-the-art performance on the TabArena leaderboard and delivered measurable results for major clients like Hitachi and Credit Agricole before signing a definitive acquisition agreement with SAP. Under this deal, SAP will invest over €1 billion to scale Prior Labs as an independent European sovereign AI hub while preserving its open-source models and research leadership.
- RAISE Summit23 min
Scaling Global AI Infrastructure | Kevin Cochrane, Vultr | RAISE Summit 2026
Vulture's representative, a recent ultra-running record holder, announced the "Vulture AI Factory," a unified infrastructure stack launching with SUSE to support sovereign cloud operations and decouple GPU inference from CPU orchestration. The speaker outlined a new 30-year "super cycle" of compute innovation while revealing open-source breakthroughs like the Voltron Retriever model and new integrations with NVIDIA, AMD, Dell, and NetApp. Highlighting enterprise adoption through a Marriott International use case and recent hackathon victories in clinical trials and telco operations, the event emphasized a strategic shift toward decentralized, agentic AI architectures.
- RAISE Summit19 min
Pat Gelsinger & Elad Raz: Where Chips Go After the GPU Era | RAISE Summit 2026
Next Silicon has pivoted from high-performance computing to artificial intelligence, deploying its Maverick accelerator which integrates custom RISC-V CPUs with data flow architecture to unify serial and parallel workloads in national laboratory environments. The company's four-generation development strategy emphasizes adaptable, software-driven designs that support evolving non-Transformer models while maintaining compatibility with legacy Fortran and modern CUDA ecosystems. Looking forward, Next Silicon anticipates a market shift toward 10,000x efficiency gains and 3D physical modeling, requiring unified memory spaces to handle continuously learning systems rather than the fragmented training and inference phases of current AI.