MUMBAI : Zoho Corp. is experiencing a revenue growth slowdown due to macroeconomic challenges in its key North American market. Chief executive Sridhar Vembu said the company has turned cautious regarding spending and hiring, as many of the problems are unlikely to be resolved quickly.
“We will have temporary fluctuations. But, you also have to take into account the long-term backdrop. That is why we turned very cautious over a year ago, and we remain cautious in terms of hiring and other expenses,” Vembu said.
Founded in 1996, Zoho reported $1 billion in annual operating revenue in November last year. While revenue for the company is still growing at a slower pace, the company is hiring “very cautiously”.
Vembu, however, is optimistic about long-term tech outsourcing spending, and added that company remains bullish on IT spending growth in India.
Zoho is also in talks to invest in manufacturing advanced chemistry cells—the key component of batteries.
“We are looking at this space and are in talks with some stakeholders. It’s too early to announce, but we need to build deep tech and manufacturing capabilities in our nation. There is scope here, too, with advanced cell chemistry production-linked incentive (PLI) schemes. We need good people with experience in cell R&D, and we’re talking to some right now,” he said.
However, this initiative can be run via subsidiaries and integrated later on. “For instance, we have a medical equipment subsidiary where battery technology can be integrated,” Vembu added.
Zoho is also developing “domain-specific” artificial intelligence (AI) models to cater to specific sectors. This, Vembu said, is also helping the company optimize its long-term capital expenditure.
“Very large AI models are very expensive—to both train and run. On the other hand, smaller models that are 10-20 billion parameters in size can do very well in a single domain and are much cheaper to run. Often, these models can fit into one or two GPUs—as opposed to hundreds of GPUs required for the large language models with hundreds of billions of parameters. That would be our focus area because this is where the maximum leverage is. This helps us solve domain-specific problems. We have already started building them,” he said.
Zoho is not the only one looking to cash in on demand for AI deployments across industries. On 15 September, Ziad Asghar, senior vice president for product management at Qualcomm, said that the fabless chipmaker was looking to implement smaller AI models that could run offline for use case-specific sectors. Such smaller AI models are typically tens of billions of data parameters in size—as opposed to large language models (LLMs) that are hundreds and, at times, thousands of billions of parameters in size.
Vembu said such models are also less expensive to build. “While the large AI models require $100 million-$200 million to build, smaller models consume $5 million-$10 million,” he said. Zoho will also work with government-backed AI projects, such as the local-language LLM project, “Bhashini”, to help develop and integrate open standards.
Vembu was speaking at a company conference in Bengaluru, where the company launched the collaborative remote workplace platform Cliq Rooms.
The author was in Bengaluru on invitation of Zoho.
“Exciting news! Mint is now on WhatsApp Channels 🚀 Subscribe today by clicking the link and stay updated with the latest financial insights!” Click here!
Download The Mint News App to get Daily Market Updates & Live Business News.
Updated: 09 Oct 2023, 11:47 PM IST