Zoho co-founder Sridhar Vembu, who took on the role of Chief Scientist last year, often finds himself ruminating about AI a lot. And his posts on X show that—whether reflecting on how mathematics could increasingly become an area dominated by machines or sounding the alarm on AI voice scams.
The transition from CEO to Chief Scientist allows him to spend substantially more time on technology—researching, writing, or reviewing code. Over the past 18 months, he has filed around 25–30 patents, and roughly 70% of his effort now goes toward technology.
“I used to never get this much time. Now I’m very much dedicating myself to this,” Vembu tells AIM in an exclusive interview.
Vembu says that if he were starting Zoho in 2026, the company would still be built around SaaS, but with intelligence embedded into the software.
“The SaaS part is very necessary, but we’re combining it with the intelligence,” he says. “AI is the new UI, and we call this ‘ubiquitous intelligence’.”
Zoho is also investing in its own AI models. Vembu expects the company to launch a 100 billion parameter model this year, while also developing smaller models for specific industries.
Vembu says Zoho is also developing specialised AI models for sectors such as healthcare, education and real estate, alongside its general-purpose models. They could be cheaper to operate and could be bundled with Zoho’s existing software offerings.
Zoho also recently launched Zia Chat, its conversational AI interface for businesses, and its vertical offerings are central to the company’s reinvention.
Trillion Dollar AI Investment Bubble
Vembu thinks the flood of funding washing over the AI ecosystem is testing the structural limits of venture capital investment.
“The massive AI investment is a bubble, but AI technology is real,” he says.
However, he argues that India should continue investing aggressively in foundational AI research and scientific talent, while remaining cautious about the current investment cycle.
Vembu expects infrastructure costs to eventually fall as GPU and memory prices decline and domestic semiconductor manufacturing expands. He cites technologies such as UPI, telecom networks, and Digi Yatra as examples of how technology can become widespread without requiring trillion-dollar companies.
Leasing GPUs to Manage AI Costs
For now, Zoho is managing AI infrastructure costs by leasing GPU capacity while also purchasing some hardware.
Memory costs remain a bigger challenge. Since Zoho hosts its own servers, the company has little choice but to absorb higher memory prices while working to make its software more memory efficient.
“Long term, it’s going to pay off, but short term, it is still a challenge for us,” he says.
Vembu also sees local AI compute as a way to reduce inference costs. He says future versions of Zia Chat could combine Zoho data from the cloud with smaller models running locally on users’ computers.
“That can give you the same interface, same intelligence you are used to, but zero token cost, because it’s running locally,” he quips.
Vertical SaaS
Zoho is moving beyond its traditional SaaS roots, investing in proprietary models, vertical SaaS applications, chip design, and edge computing.
However, Vembu says the rise of AI coding tools does not necessarily eliminate the market for specialised software. Companies may be able to build their own applications, but many will still pay vendors for support, security, compliance, upgrades, and reliability.
“If the price is right, people will buy rather than build,” he remarks.
This is also shaping Zoho’s vertical SaaS strategy. The company is working on specialised offerings for areas such as healthcare and insurance, with Vembu arguing that domain expertise will become increasingly valuable as generic software becomes easier to create.
Sovereign Push
Around 10% of Zoho’s revenue currently comes from India, with the remaining revenue spread across the US, Europe, the Middle East, and Africa.
He says India is growing rapidly, with Zoho continuing to compete on a global scale rather than limiting its ambitions to India.
That’s why Zoho is investing in chip design and edge CPU devices, while India is also developing semiconductor fabrication capacity. However, he does not expect India to achieve technological self-reliance anytime soon.
“It cannot happen overnight. It is a 20-year journey,” he says. “But I do believe in the next 10 years, we will cross a very reasonable threshold.”
AI Will Change Software Jobs
Vembu believes the traditional software engineering job market will change as AI automates more coding work. His advice to students is to focus on fundamentals and develop expertise in a specific industry.
“Make sure you learn the foundations, fundamentals. Don't outsource all your assignments and everything to AI,” he suggests.
For software engineers, he recommends building expertise in domains such as finance, insurance, or healthcare.
“Generic software may not get paid, but specialised expertise will get paid,” Vembu says.
He expects new jobs to emerge in adjacent sectors, including semiconductors, medical equipment, and vertical software.
Will Zoho go for IPO?
Vembu also rules out the logic of taking Zoho public if doing so would constrain its ability to make long-term investments.
“If we were a public company, we couldn’t be investing in semiconductors, medical tech, robotics, etc,” he says.
He also links Zoho’s decision to remain private with its approach to employees during the AI transition.
“Public companies all laid off [employees]. We never laid off anyone,” Vembu boasts. “Those things are only possible because we ignore the quarter-to-quarter pressures.”
That independence is central to Zoho’s current strategy as it moves beyond traditional SaaS into AI, vertical software, and deeper technology bets.
Your reaction