Indian startups shift to lower-cost Chinese LLMs

Indian companies are increasingly using Chinese open-weight LLMs from DeepSeek, Alibaba and Moonshot AI to reduce AI spending, according to Nikkei reporting cited in the piece. Puneet Kumar, CEO at Mirae Asset Venture Investments India, said several consumer technology startups he has met since mid-2025 are using such models because publicly accessible parameters can be downloaded, modified and locally hosted, lowering costs by an “order of magnitude.”

Pricing is a key driver. Microsoft’s Foundry platform in southern India offers DeepSeek models with charges between 19 cents and $1.74 per million input tokens, while output prices range from 51 cents to $5.40; Moonshot’s Kimi input costs go up to 95 cents and output costs up to $4. OpenAI’s GPT 5.5 series is listed at $5 to $12 per million input tokens, with output costs between $30 and $54.

Adoption is expanding as US model providers court India. OpenAI and Anthropic have opened offices or expanded offerings, and Anthropic has signed Air India, Cognizant and Razorpay, while OpenAI counts Tata Consultancy Services as a client. Companies including Tesla, Amazon, Uber and Walmart are also capping AI usage, and Open Router data showed Chinese LLM usage more than doubling to 25 trillion tokens in the final week of June from the end of May, 78% more than U.S. models.

The shift raises concerns about India’s AI sovereignty. India earmarked about 104 billion rupees (about $1.1 billion at current exchange rates) for the technology over a five-year period in 2024, but reported spending lagged allocations. Analysts cited geopolitical risk, warning that dependence on American or Chinese models could leave Indian companies exposed if access is suddenly restricted.