Nasscom Tells Parliament India Funds Sarvam and BharatGen but Still Buys AI on Lowest Price -- 'We Will Never Have a Level Playing Field' Against Meta and Google

MediaNama reported on October 17, 2026 that the 17th report of Parliament's Committee on External Affairs, "India's Role in Global Tech Diplomacy, Digital Governance, Outer Space Diplomacy, AI and Emerging Innovations", tabled in September, records Nasscom asking the government to review public procurement so that domestic AI foundation models get preference over global firms.

The L1 problem

Nasscom's oral submission, as quoted in the report: "We do not have an AI inclusion roadmap right now... We are investing in foundational companies like Sarvam AI and BharatGen, who are building India's AI models, but we are not talking about how they will get consumed by Indians. Government procurement for these models still has to follow the L1 [lowest bidder] tender process."

It pointed to China: "China gave real preference to its own companies for their models to be adopted; if we still compete purely on price with Meta and Google and others, who have spent hundreds of billions of dollars, we will never have a level playing field."

What the state has already funded

The models Nasscom wants bought are government-backed. In July, Electronics and IT Minister Ashwini Vaishnaw told the Lok Sabha that "Twenty indigenous sovereign model proposals have been identified for support, comprising 12 LLMs and eight SLMs", including Sarvam AI's 30-billion and 105-billion-parameter models and BharatGen's multilingual foundation models, Business Standard reported. Spending has lagged, though. MediaNama, citing a parliamentary reply, says the IndiaAI Mission released Rs 21.79 crore in 2024-25 against a revised estimate of Rs 173 crore, and Rs 379.15 crore in 2025-26 against Rs 800 crore. On the compute side, C-DAC's tender for up to 10,000 AI compute units, covered here on October 15, sets out hosting and data-localisation rules in detail. Nasscom's point is that there is no matching rule for buying the models.

What the committee said

The committee did not explicitly recommend changing the General Financial Rules or the L1 process. It did accept the diagnosis: "the gap between India's demonstrated capability and its realised leadership is not primarily a gap of resources or talent, but of follow-through, adoption pathways, strategic clarity, and the pace at which these foundations are being built."

What remains uncertain. This rests on MediaNama's reading of the committee report. We could not retrieve the report itself, and we found no government response. A preference regime would also have to answer questions the submission does not: how to define a "domestic" model, how to handle Indian models fine-tuned from foreign open weights, and how to keep a preference from protecting weaker products. MeitY's AI regulation consultation paper, promised within a month of October 8, may be the first place this surfaces.

Analysis: India has funded the supply side of sovereign AI -- GPUs, model grants and datasets -- while leaving the demand side to an L1 procurement rule designed for commodities. Price-only tenders favour whoever can subsidise inference longest, and that is not an Indian startup. A narrower fix than a blanket preference may work better: score tenders on Indic-language quality, data residency and auditability, which Indian models are built for, rather than on price alone. Builders selling into government should watch the consultation paper for how "sovereign" gets defined.