Krutrim Targets First Silicon for Bodhi AI Accelerator in 2026

Krutrim's Chip Ambition: India's First AI Silicon Program at Scale

Krutrim, the artificial intelligence venture of Ola founder Bhavish Aggarwal, is targeting first silicon for its Bodhi AI chip in 2026, making it the most ambitious indigenous AI hardware program initiated by an Indian technology company to date. The Krutrim chip roadmap covers three distinct product lines: Bodhi, an AI training accelerator for large language model and multimodal AI workloads; Sarv, a general compute processor for cloud and enterprise infrastructure; and Ojas, an edge inference processor for on-device AI applications in mobile devices, embedded systems and IoT endpoints.

Technology Partners: ARM and Untether AI

Krutrim has entered strategic partnerships with ARM for chip architecture and instruction set licensing, and with Canadian AI hardware company Untether AI for processing architecture support. ARM's involvement ensures Krutrim's chips benefit from the world's most widely deployed processor architecture ecosystem, giving Bodhi, Sarv and Ojas compatibility with the vast software and development tool ecosystem built around ARM's instruction set. Untether AI, known for its at-memory compute architecture designed to eliminate data movement bottlenecks in AI inference, contributes architectural expertise relevant to Krutrim's AI accelerator design goals.

Manufacturing: Global Foundry Partnership

Krutrim plans to manufacture its chips at a global tier-I or tier-II foundry, with TSMC and Samsung cited as the most likely candidates. The company has not disclosed the target technology node for Bodhi 1, the first product in the family, but industry observers expect the initial chip to be built at a mature node suitable for a first silicon tape-out, with subsequent generations migrating to more advanced nodes as Krutrim's foundry relationships and chip design maturity develop.

Krutrim Cloud: The Business Context

Krutrim's chip program exists within the context of its cloud and AI services business, which has introduced over 50 services on the Krutrim Cloud platform, including Bhashik, a multimodal language hub supporting translation across text, voice and video, and a Customer Experience AI product offering multimodal enterprise AI agents. Running these services on externally sourced GPU infrastructure imposes significant compute costs, and Krutrim's chip program is partly motivated by the long-term economic case for operating AI cloud services on proprietary silicon rather than renting capacity from Nvidia or AMD.

Data Center Infrastructure Plans

Krutrim is planning to scale its data center capacity to 1 GW by 2028, a massive infrastructure commitment that would require thousands of AI accelerators and general compute processors. If Bodhi and Sarv reach commercial production and yield rates that make them cost-competitive with imported alternatives, the Krutrim data center buildout becomes a large captive market for domestically designed chips, a deployment scale that very few chip startups in any country have access to at this stage of development.

Why This Matters for India's Chip Ecosystem

Krutrim's chip program demonstrates that India's private sector is investing in chip design IP at the product level, not just providing services to foreign chip companies. Combined with Netrasemi's A2000 edge AI chip heading to mass production and a growing cohort of DLI-scheme-backed startups, Krutrim's Bodhi program signals a qualitative shift in the ambition level of Indian semiconductor design, from contract engineering to product IP ownership and potentially to designing chips that compete directly in the global AI hardware market.