Blog

SaaS unicorn Zoho Corporationis planning to roll out its foundational AI models by the end of the year, said the firm’s newly appointed CEO Shailesh Kumar Davey.

This announcement comes at a time when DeepSeek’s challenge to OpenAI has renewed questions about India’s ability to build its indigenous AI model.

In his first media interaction since taking over, Davey stated that Zoho is developing two AI models— based on 7 billion to 13 billion parameters each, as it pushes ahead with its larger AI ambitions.

Davey said building an LLM (large language model) requires dealing with experimental methods and constraints; DeepSeek, OpenAI, Gemini, and Anthropic are all tackling similar challenges.

“We have been training a 3-13 billion parameter model and have been working on it over the last eight months… Whatever DeepSeek did was experimental. There was a similar set of experiments conducted by OpenAI, Gemini, or Anthropic… and DeepSeek had done a certain other set of experiments, because the constraints in which they were working were very different,” said Davey, at a press conference in Mumbai.

“From an intellectual challenge perspective, I would consider all of them to be almost equal. The space in which all of these are working are just almost similar. Just that the constraints that some would prioritise is what would differ,” he added. 

Beyond the general-purpose model, the SaaS firm is also building an Indic language model designed specifically for Indian users. As part of this effort, it is working with IITs, AI4Bharat (a research lab at IIT Madras), and data collection firms such as Karya to create datasets, including those for voice-based models.

AI4Bharat is working on developing open-source datasets, tools, models and applications for Indian languages.

While Zoho expects to roll out its indigenous LLM by the end of the year, the company has not disclosed a formal timeline for the Indic LLM.  

As of 2024, Zoho had invested over $20 million in NVIDIA for AI infrastructure, along with forging partnerships with AMD and Intel for training their AI models efficiently.  

“We want to ensure that we know what technology is there today and how we can bring value to the customer… as well as when it is evolving, be the leader there. The whole AI space would have a newer set of training and inference, and when it comes, we want to be in the forefront of it. So we are investing a lot now, understanding and having the engineers work around the muck,” Davey said. 

Chennai-based Zoho has been valued at Rs 1,03,760 crore, marking a 58% jump from its previous valuation of Rs 65,700 crore, according to the 2024 Burgundy Private Hurun India 500 report.

Meanwhile, Arun Kumar, Regional Director of ManageEngine, a division of Zoho Corp, revealed that India is set to become the second-largest market for the enterprise IT management arm by the end of 2025, after the United States. 

“India was not even in the top 10, but today we are ranked number three. By the end of this year or next year, we’ll move to number two. We cut across close to 7,500 enterprises in India, across different industry verticals—BFSI, public sector, healthcare, education, IT, ITES—with major markets revenue coming in from Mumbai, NCR, Bangalore, Hyderabad, and Tamil Nadu,” stated Kumar. 

Eighty to ninety percent of ManageEngine’s revenue is generated through channel partners and system integrators; the Indian market accounts for roughly 10% of its overall global revenue, he added. 

Manage Engine, which claims to serve nine of every 10 Fortune 100 organisations, has 18 data centres, 20 offices, and over 200 channel partners worldwide. Eyewear retailer Lenskart and WAISL, a provider of digital airport consulting services, are among its customers in India.