Vipul Ved Prakash, Together AI | RAISE Summit 2025 скачать в хорошем качестве

Vipul Ved Prakash, Together AI | RAISE Summit 2025

In this RAISE Summit interview, Vipul Ved Prakash, co-founder and CEO of Together AI, sits down with theCUBE’s John Furrier to unpack how an “AI acceleration cloud” is redefining supercomputing access. Prakash explains why generative AI demands both raw horsepower and hardware-software co-optimization, sharing that Together AI now fields its first Grace Blackwell 200 cluster delivering 1.4 petaFLOP per rack and is adding tens of thousands of NVIDIA GPUs each quarter. He details how software tuning cut DeepSeek model inference from USD 8 to 55 cents per million tokens and forecasts another 3-4× drop with new silicon. The discussion dives into open source momentum, with more than 200 ready-to-use models on Together’s platform, and explores emerging “neo-cloud” architectures tailored for AI-native workloads. Prakash highlights sovereign AI as a defining trend, noting two gigawatts of European capacity in the pipeline across multiple decentralized data centers. Key takeaways include the shift from proof-of-concept to production at AI-native startups, the operational demands of high-density GPU fleets, and the race for regulation-ready infrastructure that blends observability, fault tolerance and data compliance from day one. Find more SiliconANGLE news and analysis https://siliconangle.com/ Follow theCUBE’s wall-to-wall event coverage https://siliconangle.com/events/ Learn about the latest theCUBE events https://www.thecube.net/ 00:00 - Intro 00:07 - Title: "Unveiling Together AI: Foundations, Innovations, and Vision 04:07 - Driving Innovation: Harnessing Open Source and Market Intelligence in AI Acceleration 09:22 - Sovereign AI and European Expansion 11:44 - AI Regulation and Infrastructure Ready: The NVIDIA Influence 16:04 - Future of AI and Infrastructure 18:10 - Conclusion #theCUBE #RAISESummit #TogetherAI #GenerativeAI #AIInfrastructure #Supercomputing #SovereignAI