SoftBank, OpenAI & Amazon | Manthan Gupta (AI Engineer) | Himanshu Dubey (Upsurge Labs) | Episode 40

Today’s episode covers the capital + compute shift — and what it means for builders shipping in AI. 🗞️ News Highlights SoftBank “back in action” from 2026 — deal pace re-accelerates after the reset; ~$5.8B raised via Nvidia share sale; focus on AI infra + late-stage, with fewer, bigger, higher-conviction checks. “We want to continue investing in India.” — Sumer Juneja OpenAI turns 10 — founded Dec 2015; now a ~$20B annualized run-rate enterprise AI giant with ~$1.4T+ commitments and deep Azure alignment. Amazon adds $35B more to India by 2030 — AI benefits for 15M businesses and 4M students; exports targeted at $80B; supports 3.8M jobs. Plus Satya Nadella’s India AI tour: agentic-AI partnerships with Infosys/TCS/Wipro/Cognizant and more Azure zones + in-country capacity. 🧠 Guest Segments 1/ Manthan Gupta: A sharp technical teardown — can you infer LLM architecture via conversation, why OpenAI skipped RAG for memory, what that implies for “better RAG” startups, and where this approach breaks (Claude/Gemini next). 2/ Himanshu Dubey (Upsurge Labs / Ground Zero): How they build and kill products at speed (Bhindi, Helium, OpenBio, BlueBrain) — first 10× wedges, real user use-cases, and what Ground Zero is becoming. Plus Instadapp, DeFi’s next decade, and AI × on-chain. 🎧 Watch, subscribe, and join our WhatsApp community: https://chat.whatsapp.com/BK69EsNWP8BJIaoy3k7EfX?mode=ac_t Follow us: X → https://x.com/OfflineOnAir Instagram → https://www.instagram.com/theofflinenetwork LinkedIn → https://www.linkedin.com/company/offlineonair Website → https://www.theoffline.network