Zoho on Wednesday marked its entry into vertical SaaS solutions focused on seven industry-specific applications at its annual event Zoholics in Chennai. Chief Scientist and Co-founder Sridhar Vembu tells Narayanan V about the rationale, the role of AI in boosting productivity and the changing economics of the technology industry. Edited Excerpts:
What prompted your entry into vertical SaaS?
If you look at our horizontal suite, we already have a very strong foundation. When you go into a particular segment like real estate, healthcare or education, they need very industry-specific solutions. Typically, they will take this and customize it. So, we wanted to work with those customers directly and offer these pre-customised solutions so that the time to deploy and time to value is much faster. This is also related to AI productivity. How are we going to tap that productivity? What is the benefit to the customer of this productivity? They get their solutions faster and it costs less because they don’t have to spend all this time and money on customising it. It’s already done for them. This is actually very much a co-creation model. We have been working with customers on this and we plan to expand this in a big way.
How do you strike a balance between AI productivity gains and overt reliance?
Four or five years ago, I said software programmers were like handloom weavers and they were hand-weaving code. Powerloom machines were coming. So, one way to adapt is to become fashion designers and the other adaptation is to build your own machine looms. We are actually doing both. The fashion looms are all the industry-specific vertical offerings that we give, and the machine looms are what will reduce technology dependence or maybe eliminate it altogether. We can raise the productivity of the workforce rapidly so a given person is able to do more. The cost to the customer stays flat even though incomes go up. That is a traditional model of prosperity, by the way: inflation-free prosperity. If productivity jumps a lot, then prices can stay [flat]. That’s how the tech industry has delivered all these things.
How has the AI gold rush and memory price increase impacted data centres?
First, the memory prices went up, thanks to artificial intelligence (AI). Even traditional data centres, which have nothing to do with AI and GPUs, saw server prices go up. So, we are countering it by making our own servers. Some of the margin that was going to vendors is being retained by us. On the AI side, GPU prices are very high. So, right now we are only renting. We are not actually buying. We are leasing that capacity.
In the long term, I think GPU prices will come down. Also, the large models will become more efficient to run. The small models will become more intelligent and more effective. So, when you look at the efficiency and these two curves, that’s where you can predict that AI will become a commodity.
Global majors have called for a slowdown in AI development. Your views?
It’s exactly like top students in a class telling other students not to study for the exam. If I, for example, want to slow down, I can slow down myself. I don’t ask the entire industry to slow down because I want to slow down. I see it a little more cynically. These companies want everybody else to slow down so that they can move faster and they can, kind of, monopolise the market.
That’s the real truth. There are too many models, prices are crashing on all dimensions, and then you cannot justify your two-year, three-year valuation based on the stock and economics that have evolved. OpenAI alone has about $350 billion or something in commitments they have made to other companies. All these accounts will come out when they go public. That’s why OpenAI also cancelled its IPO. The problem is that LLMs are becoming a commodity. You cannot meet the valuation and profit objectives in the current industry landscape, which has evolved.