Indian artificial intelligence startup Sarvam AI has announced Token Factory, offering access to multiple models, including its own and those developed in China, through a unified inference platform. The announcement has sparked debate over India’s homegrown AI ambitions. The matter has come under scrutiny because Sarvam is among the companies selected under the government’s IndiaAI Mission to develop homegrown AI models.
What is Sarvam AI’s Token Factory
The platform handles the process of running an AI model and generating responses to users’ questions or instructions. This process is called inference. Sarvam says its system is designed to make this process faster and more cost-effective by managing how requests are sent to its GPUs and reusing parts of earlier prompts where possible. Businesses can choose a model depending on their requirements and pay based on the number of tokens processed. Tokens are small units of text that an AI model reads or generates.
How Sarvam’s Token Factory functions
Sarvam describes Token Factory as an inference stack serving its own and other open-weight models from GPU clusters in India. Its catalogue includes Sarvam 105B, DeepSeek V4.1 Flash and V4 Pro, Kimi K3.1 Pro, GLM 5.3 and Flash, and Gemma 4 31B. Developers access them through one OpenAI-compatible API by configuring the endpoint, selecting a model and sending requests.
The platform manages the work between a request and a response through three layers:
- An inference engine
- A cache
- A control plane
The engine runs the model on graphics processing units (GPUs). The cache stores reusable prompt information to avoid processing the same context repeatedly. The control plane manages requests, routes them to the appropriate resources and adjusts capacity according to demand.
The engine separates processing into prefill and decode. Prefill reads the full prompt, including earlier conversation or context. Decode generates the answer one token at a time. Sarvam says these tasks run on separate GPU pools because they have different computing requirements. This can prevent long prompts from slowing response generation for other users.
Caching is useful in long conversations and AI agent workflows, where a model may receive the same context repeatedly. Without it, the system processes those tokens again, adding cost and delay. Sarvam says its multi-tier cache uses GPU memory and shared storage to reuse prompt information. A cache-aware router can direct returning requests to servers where relevant context is available.
Why has Token Factory triggered a sovereignty debate
The central criticism is that a model’s country of origin differs from the country where it is hosted. DeepSeek, Kimi and GLM were developed by Chinese companies. Running their open-weight versions on Indian GPUs does not make them Indian, transfer their intellectual property to Sarvam or give India control over their development. This raises the question of whether India is building its own AI capabilities or mainly providing infrastructure to run models developed elsewhere.
The term covers the ability to set rules for how AI systems are developed and used, audit their behaviour, protect sensitive information and maintain access to critical computing resources. It is therefore broader than data localisation or the nationality of a model developer.
Sarvam’s role in India’s AI ecosystem has intensified the debate. It was selected under the IndiaAI Mission, which seeks to strengthen domestic AI capabilities, including indigenous foundation models, and received subsidised computing resources to train its own models. This has raised questions about whether government support for developing homegrown AI was also benefiting a commercial platform offering access to Chinese AI models.
Sarvam said its Token Factory does not use government-subsidised GPUs. Co-founder Pratyush Kumar told Moneycontrol that the platform runs on privately procured GPUs. The company said its subsidised access under the IndiaAI Mission, used to train its own models, ended in March 2026. It maintains that developing its own models remains its core mission, while offering other models generates revenue.
According to the report, Sarvam argues that model development is expensive and inference can generate revenue from computing infrastructure. It says Indian businesses should be able to use global models through a domestic provider, with local infrastructure and rupee pricing. Global cloud providers such as Amazon Web Services also offer businesses access to AI models developed by other companies through platforms such as Amazon Bedrock.
Data sovereignty is another concern. Sarvam says Token Factory runs AI models on its GPU infrastructure in India and does not send customer prompts to the overseas providers that developed those models. The company also says it offers zero data retention on request, under which prompts and responses are not stored. However, keeping data in India does not automatically address every security concern.
Businesses also need to examine how data is accessed, how long it is retained and what security measures are in place. Hosting a foreign-developed model in India also does not make the model indigenous.
What about other Indian AI platforms
Sarvam AI has faced greater criticism not only because it is hosting Chinese AI models. Other Indian companies, including E2E Networks and IndieRouter, also offer access to foreign-developed models through infrastructure located in India. E2E Networks, for instance, launched its Blaze inference platform on November 2, 2026, offering models such as DeepSeek, Qwen, GLM and Gemma through a single API. However, Sarvam’s position is different because it was selected under the government’s IndiaAI Mission to develop indigenous foundation models and has been seen as an important part of India’s effort to build its own AI technology.
The criticism has therefore centred on whether a company backed by the government to develop homegrown AI should also build a commercial business around models developed in other countries, particularly China. The Indian Express reported that the reaction was partly driven by the expectations attached to Sarvam as a symbol of India’s AI ambitions. Supporters, however, argue that hosting foreign models can be a business opportunity and allow Indian companies to serve customers using locally available computing infrastructure.
How does hosting foreign models benefit Sarvam
Chinese AI companies have developed several powerful open-weight models, including DeepSeek, Kimi and GLM. Unlike some AI services that can only be accessed through a company’s own platform, open-weight models can be downloaded and run on a company’s own servers. This makes them an option for businesses that want to use advanced AI models without relying on services such as OpenAI or Anthropic.
For Sarvam, offering these models gives it another source of business besides developing its own AI models. It can attract companies that want to use models such as DeepSeek or Kimi but do not have the GPUs or technical resources to run them themselves. Sarvam provides the computing infrastructure and access through its Token Factory platform.



