Sarvam AI faces backlash after its new Token Factory platform started offering access to Chinese models like DeepSeek and Moonshot AI. The Bengaluru-based startup previously pushed for India to build independent technology. Many now question this move, as the company got government support to develop domestic AI models instead.
Sarvam AI has been among the AI companies that promotes in-house chatbots and reduce dependence on foreign technology. But its latest business decision has introduced an interesting contradiction into the conversation around technological sovereignty.
The Bengaluru-based AI startup has launched Token Factory, an inference platform that gives developers and businesses access to several artificial intelligence models through a single application programming interface (API). Alongside its own models, however, the platform is also offering access to systems developed by Chinese AI companies, including DeepSeek, Moonshot AI and Z.AI.
That is an interesting development, particularly given Sarvam's positioning around India's need to build independent AI capabilities. The company has received government backing through the IndiaAI Mission and was among the early recipients of subsidised computing resources to develop domestic AI models.
The question, therefore, is not simply why Sarvam is offering Chinese models. It is also about the message the company has been sending about technological independence, particularly when its approach to American AI models has been more restrictive.
Sarvam has previously been associated with efforts to limit or curtail the use of American AI models in favour of developing homegrown alternatives. Its argument has broadly reflected a concern that India should not remain dependent on foreign companies for foundational technology. Yet, with Token Factory, the startup is now making Chinese alternatives available to customers.
Offering access to an overseas model does not automatically mean abandoning the ambition to develop domestic technology. An inference platform is, after all, designed to give customers flexibility, allowing them to choose models based on their requirements, performance and costs. Having multiple models under one roof could also make the platform more attractive to developers.
But the optics are difficult to ignore. If dependence on American AI companies is a concern because it threatens India's technological autonomy, why should dependence on Chinese models be viewed differently?
The timing also makes the decision more striking. India has been increasingly vocal about building sovereign AI infrastructure, developing indigenous language models and reducing its reliance on technology controlled by foreign players. Sarvam has been one of the prominent private-sector names associated with that vision.
There is also a difference between building an Indian AI model and operating a platform that distributes models developed elsewhere. Both can be commercially useful, but they serve different purposes in the larger sovereignty debate.
Ultimately, Token Factory could be a sensible business move for Sarvam AI, especially if it wants to compete in the growing enterprise AI market. Developers are unlikely to care about the nationality of a model if it delivers better results at a lower cost.
However, a company cannot champion technological sovereignty when it suits its narrative and treat foreign dependence as an ordinary business decision when it comes to its platform. Sarvam's latest move does not necessarily undermine its domestic AI ambitions, but it does invite scrutiny over how consistently it applies the principles it advocates.
