January 4, 2024
Gen AI startups received US$ 566 million in 2022 and US$ 144 million in 2023
Investments cover a range of Gen AI offerings, with text content creation, chatbots and virtual assistants, and image and video generation
The past year saw tech giants and startups developing large language models trained in Indic languages
Flipkart and Ozonetel are exploring small language models tailored to domain-specific use cases and trained on smaller datasets
According to Nasscom, Indian Gen AI startups have received a total investment of US$ 700 million in the past three years. India has a thriving ecosystem with over 100 GenAI startups.
Sangeeta Gupta, the Senior Vice President and Chief Strategy Officer at Nasscom, emphasised the importance of recent funding from emerging Gen AI ventures like Sarvam AI, a Hindi large language model (LLM) maker, and Adobe’s acquisition of Bengaluru-based Rephrase.ai. She highlighted these as positive validations of the growing AI landscape in India.
The year 2022 peaked with US$ 566 million invested in Gen AI startups. This was followed by US$ 144 million in 2023, according to data. These investments cover a range of Gen AI offerings, with text content creation (18%), chatbots and virtual assistants (18%), and image and video generation (16%) comprising significant portions of the funding.
Nasscom estimates that AI startups received private investments totalling US$ 8 billion between 2013 and 2022. Notably, US$ 3.24 billion was invested in 2022 alone across 1900 AI startups, marking it the most significant funding period in AI. The past year saw remarkable breakthroughs in Gen AI innovation, with tech giants and startups developing large language models trained on Indic languages.
Several noteworthy initiatives include the launch of BharatGPT by Conversational AI startup Corover.ai, a 7 billion-parameter model trained in 14 Indian languages for text, voice, and video interactions. Sarvam AI introduced OpenHathi-Hi-0.1, the first open-source Hindi language model built on Meta’s LlaMa 2-7B model. Ola unveiled Krutrim LLM, which is capable of comprehending 22 Indian languages and generating responses in 10 languages. Tech Mahindra is actively engaged in ‘Project Indus,’ developing an LLM trained in Hindi and 37 Indic dialects.
Additionally, companies such as Flipkart and SaaS provider Ozonetel are exploring small language models tailored to domain-specific use cases and trained on smaller datasets.
Source: Economic Times