India weekly AI roundup flags cost crunch, DeepSeek price war

This Indian weekly roundup stitches together what many feel as separate threads into a coherent signal: the economics of frontier AI are diverging. On one side are labs like Anthropic, raising $65 billion at near‑trillion valuations and keeping headline API prices high while quietly leaning on hyperscaler credits and loss‑leading enterprise deals. On the other side are aggressive challengers like DeepSeek, which just made a 75% promotional price cut permanent, effectively setting a new floor for high‑end model pricing. ([financialexpress.com](https://www.financialexpress.com/life/technology-ai-affordability-wakeup-call-anthropics-65bn-mega-round-and-indias-first-12nm-ai-chip-weekly-ai-roundup-4255352/))

Layered on top are IBM and Red Hat’s $5 billion Project Lightwell, which aims to use AI to secure open source at scale, and Netrasemi’s 12 nm edge AI chip, signaling India’s ambition to play in the hardware stack, not just services. ([financialexpress.com](https://www.financialexpress.com/life/technology-ai-affordability-wakeup-call-anthropics-65bn-mega-round-and-indias-first-12nm-ai-chip-weekly-ai-roundup-4255352/)) The article’s core takeaway—that enterprises are running into budget walls as agentic workloads chew through tokens—should matter for anyone betting on AGI timelines. If only the very largest players can afford frontier inference at scale, we may see a bifurcation: elite labs racing ahead on super‑clusters while a broader ecosystem standardizes on cheaper, slightly‑behind open or regional models. That could speed up top‑end capability development even as it forces more pragmatic, cost‑sensitive deployment elsewhere.