Перейти к содержимому

EXCLUSIVE: Ziroh Labs CEO Says You No Longer Need GPUs to Run AI and He Has the Numbers

AIM Network

0:00 / 0:00

EXCLUSIVE: Ziroh Labs CEO Says You No Longer Need GPUs to Run AI and He Has the Numbers

14 498 просмотров · 2 дн. назад
AIM Network
384 тыс. подписчиков
14 498 просмотров · 2 дн. назад
While global technology enterprises race to secure multi-billion-dollar GPU clusters, Ziroh Labs is challenging the assumption that AI inference requires dedicated hardware accelerators. Through its Kompact AI platform, the company enables production AI models to run on standard CPU architectures, claiming up to three times the performance of existing CPU-based solutions at 70% to 80% lower operational costs. Founded by Hrishikesh Dewan, whose early career involved building an Assamese-language Linux operating system kernel at IIT Guwahati, Ziroh Labs leverages cryptographic optimization and decentralized computing principles to make AI deployment accessible across resource-constrained environments. In this episode of Front Page, Sudhi Sachdev sits down with Hrishikesh Dewan, Co-Founder & CEO of Ziroh Labs. They explore the shift from centralized GPU infrastructure to decentralized CPU execution, domain-specific small language model (SLM) architectures, the launch of Kompact AI Box for healthcare and education, collaborating with cryptographer Whitfield Diffie, partnering with IIT Guwahati to launch free LLM foundations coursework, and fostering a research culture that embraces experimentation and failure. Can CPU-based inference platforms democratize enterprise AI deployment across the Global South? ✅ Cypher 2026: Join 5,000+ AI leaders. Book now - https://cypher.analyticsindiamag.com/ ✅ Chess League: Test your skills - https://chess.analyticsindiamag.com/ CHAPTERS: 0:00 Intro: Kompact AI runs models on CPUs 1:40 Assamese OS kernels & optional GPUs 4:13 Decentralized computing & natural architecture 7:10 Large vs. small models for domain tasks 12:48 Preserving quality & CPU throughput 15:15 Coexisting with GPUs & hardware bounds 21:28 Private AI in healthcare & education 24:18 Kompact AI Box: Local software stacks 29:43 Running local SLMs on $200 smartphones 35:45 Cryptography origins & Whitfield Diffie 41:25 Centralized mainframes to edge AI 46:40 IIT Guwahati partnership & free LLM course 52:10 Shifting research from service to IP 58:45 Final thoughts: Learning from failure #KompactAI #ZirohLabs #CPUInference #EdgeAI #FrontPage #AIMNetwork ------------------------------------------------------------------------------------------ 📌 Note: The video thumbnail is for representational purposes and was created with the assistance of AI. 📩 For inquiries or questions, contact us at: info@aimmediahouse.com ------------------------------------------------------------------------------------------ 📌 ABOUT FRONT PAGE Front Page is a daily live news briefing covering AI, enterprise tech, semiconductors, and India’s role in the global technology economy. Built for people who make decisions, not just consume headlines. ------------------------------------------------------------------------------------------ 👋 ABOUT AIM NETWORK AIM Network is India's leading voice on the intersection of AI, Technology, and Business. We bring you sharp analysis, breaking news, and original perspectives on everything from Generative AI breakthroughs to geopolitical market shifts. 🔔 Subscribe for daily insights:    / @aimmediahouse   📺 WATCH OUR POPULAR SERIES: Front Page (Daily Tech News):    • Front Page   Tech Talks (Podcast):    • Tech Talk   -Simulated Reality (Podcast) -    • Simulated Reality Podcast   📰 STAY UPDATED: Website: https://analyticsindiamag.com/ LinkedIn:   / analytics-india-magazine   Twitter/X: https://x.com/Analyticsindiam Instagram:   / aim.media.house   ------------------------------------------------------------------------------------------