India now has a data platform to train its own AI model, starts working on its own GPU

New Delhi: India unveiled its library of non-individual data sets called AI Kosh to help create the latest steps (LLMS) by the Center to help increase the development of artificial intelligence (AI) in the country.

AI Kosh will also be used to approve startups to reach 18,693 GPU of India AI Mission, or graphic processing units, which will be introduced at a subsidized fee, on Thursday, Electronics and Information Technology said.

At the same time, Union IT Minister Ashwini Vaishnav said, “comprehensive stakeholder counseling” has begun to manufacture “India’s own GPU” – a process that he said that “three to four years” would take. “The construction of the GPU is a very, very difficult process, but we will reach there. As we progress, we will share more information in the appointed time. ,

The GPU is considered to be the most important part of the construction of the AI ​​model – Elgorithm who helps generate speech, text or video like humans. Since the arrival of the US tech firm Openai’s chat in October 2022, GPus has become one of the world’s most demanded commodities-which has been flattering with an increase in evaluation from NVIDIA Corp’s $ 300 billion to $ 3.6 trillion.

Also read Hiranandani’s AI Infra firm Yotta is planning an American listing. But at what cost?

However, India currently does not patent for the manufacture of its own JPU, which depends on the country on US Tech firm Nvidia, Advanced Micro devices (AMD), Amazon Ink, Qualcomm Inc., Intel Corp and others for these prized chips.

“The idea is to achieve our own advanced chip patent and make them in more sophisticated semiconductor building plants, which will come to India in the coming years,” a senior government official said about India’s move to create a circular supply chain for AI. “Till then, we have a lot of GPUs to arrive to make the expected basic AI models.”

Mint On 5 February, India said that India was aiming to develop its AI chip in about five years.

Till then, Vaishnav said that Matty has received applications to develop foundational models from 67 startups – which will be provided access to dataset of mety via AI dictionary platform.

“To start, we will choose three to five fundamental AI startups, and will provide GPU infrastructure adjacent to them,” the minister said. “The platform to offer AI compute is now live, and India’s GPU cost will be the lowest in the world.”

Also read NVIDIA’s Blackwell GPU can reach Indian coasts in early October

Vaishnav said that 22 applicants have applied to create a basic AI model, while the rest can manufacture models for a wide range of cases.

GPU per hour per hour – Training owners in India entering the field of AI model will be a significant obstruction in India 67 or $ 0.77. Global data centers sell GPU access to enterprises at an average cost of $ 2–2.5 per hour per hour.

“Compute platforms also include public datasets introduced by various ministries, which will allow startups up to non-individual data in different languages. These datasets will cover areas such as agriculture, logistics and more. Even Bhashini will provide its Indian language data for startups to create AI models for India’s specific needs, Vaishnav said.

Bangalore-based AI Startup Corpoor.AI, who sells Indic-language AI application based on its secondary, open-based language model Bharatgpt, is one of the applicants in Meety AI missions to make its own JPU.

Also read Chips, Indic LLM, Flying Taxis: Inside India Trests with Tech Independence

The company’s founder AI model will consume consumer and “a general purpose model covering all beats, which will be extended on the basis of demand for sector-wise expertise,” said Ankush Sabarwal, Chief Executive Officer of Korover. Mint,

The compute infrastructure provided on Thursday includes GPUs from four companies- NVDia, AMD, Intel and Amazon web services, 10 shortlisted data centers and technology infrastructure vendors are being supplied to the mety. In the long run, the GPU mandate of the IT Ministry will be an important revenue stream for these vendors, such as CEO of data services Sunil Gupta told Mint On 10 February.

Vaishnav said that cabinet-appointed fund of over 10,000 crores will be used to “offer 40% subsidy” for the cost of accessing GPU by startups. However, the minister did not disclose whether the Center would hold a stake in the developed AI model or they would be completely private owned.

Read more Mint Interpreter: India-centric LLM’s mercurial rhise