SBI Chairman, Shri Challa Sreenivasulu Setty, shares his insights with ET Now, on the sidelines of Global Fintech Fest, 2026.
#SBI #TheBankerToEveryIndian #GFF2026
What Challa Setty said
Written from the verified transcript and checked against it. Every figure links to the moment it was said.
Challa Setty discussed SBI's technology and macro outlook. On AI, he said most banks, including SBI, prefer building their own small language models (SLMs) over large ones, using open-sourced LLMs with an orchestration layer for flexibility. He noted agentic AI is not yet deployed in cybersecurity, but the IBA committee he leads is working with OEMs on a framework to use AI against AI threats. Setty emphasized SBI's 'digital first, customer first, nation always' approach to lower cost to serve, and mentioned integrating 18 non-banking subsidiaries onto YONO. He said FCNR(B) inflows of $126 billion are temporary and will be absorbed in 4-5 months, with sustainable liquidity of 6-7 lakh crore. SBI raised $10.9 billion in deposits. He sees possible NIM blips industry-wide but not necessarily for SBI. Inflation and oil above $100 are worries but Q2 looks good. He expects credit growth to hold in Q3 and FY27 GDP at 7%.
Key takeaways
SBI prefers building own small language models over large ones, using open-sourced LLMs with an orchestration layer.
The $126 billion FCNR(B) liquidity is temporary and will be consumed in 4-5 months.
India's FY27 GDP growth should be at least 7%.
SBI has 18 non-banking subsidiaries, which it aims to integrate onto YONO platform.
We hear from IT applications providers as well is that there is a reluctance to use large language models. Small language models is what is preferred because it helps in privacy of data, customization much better. What has been your experience when you are talking to the regulator? Are they open to large language models or is a small language model what they are willing to look at?
No, I don't have any information whether the regulator has any preference for SLMs and LLMs, but most of the participants like us, we probably would like to have our own SLMs, but not building our own LLMs. Obviously, then you will have open-sourced LLMs and you will have an orchestration layer, your ability to use one LM in one use case and another LM in another. So we do not want to lose that flexibility. So the approach by the users or the user entities is more in terms of having your own small language models if you can afford it and use a large number of LLMs outsourced.
So far, I think none of us really had deployed agentic AI or a cutting-edge AI technology in cybersecurity. While the tools are efficient within the broader toolkit what we have for cybersecurity protection, we have been taking the best tools available, best manpower available. But fighting AI with AI is something we still are not there. IBA, where I've been leading a committee, we have given a broader framework and we have worked with a lot of OEMs on what is their experience of using AI as a cybersecurity tool. We are closely interacting with them and we are giving a broader framework of how AI can be used to fight AI threats on the cybersecurity front. But if you ask me, I think we are still not there in a big way, but we are moving in that direction.
I know one of the important things which SBI has done is the way it has moved towards technology and adopting it. And now you have an ecosystem in place with many of the companies listed already, actually. Right, I mean you have fund management which can access YONO, and the kind of customer base that is there and the way the customer base is also growing. How do you bring all that into a tech stack for SBI where you leverage it much more?
True. I think if you see, we've been following an approach called digital first. We always have been having a customer-first approach, and our slogan now is "Digital first, customer first, nation always." So for a bank of our size and a country of our size, if you don't have the digital-first approach—probably AI-first will come as we speak now—it is very difficult to serve. You may serve, but it comes with an expense. So how do you bring down the cost to serve is very core to our digital-first approach, while you use your physical network on a physical network basis. But what is important is that you continue to invest in the right technology. And you are right, I think we are also trying to bring the whole ecosystem. We have almost 18 non-banking subsidiaries, and they provide very valuable customer financial needs met through them. So how do we bring them on one single platform? YONO could be one such platform which we are building.
Sir, we have you here so I'll ask you some macro questions as well. We've seen the kind of FCNR(B) flows which have come in, $126 billion, that's brought in a lot of liquidity as well. How do you look at the liquidity scenario at this point in time?
So this is a surplus liquidity which has come in one short time. Of course, there is a window within a window everybody has to mobilize these deposits, but I believe that this liquidity gets consumed in 4 to 5 months. So this temporary liquidity RBI is trying to absorb temporarily. One good thing is that this liquidity of almost 6 to 7 lakh crore is a sustainable liquidity. It's a five-year liquidity what we have created, so I believe it's very positive for the banking industry. We've been talking about the deposit and advances gap, so that gap obviously is now partly filled. But it's not such a large liquidity that we should be worried about what kind of lending will happen and all those things. I think it gets consumed. We will be having a busy season starting from October, right?
Yeah, festive season. Yes, you have September 15th when you have advance tax things, so I don't think it need to be worried about the excess liquidity.
Setty, C. (2026, September 10). SBI Chairman, Shri Challa Sreenivasulu Setty [Interview transcript]. State Bank of India. CEOInterviews.AI. https://ceointerviews.ai/interview/1335695/
MLA
Challa Setty. "SBI Chairman, Shri Challa Sreenivasulu Setty." State Bank of India, 10 Sep. 2026. Transcript, CEOInterviews.AI, https://ceointerviews.ai/interview/1335695/.
BibTeX
@misc{setty2026_1335695,
author = {Challa Setty},
title = {SBI Chairman, Shri Challa Sreenivasulu Setty},
howpublished = {Interview transcript, State Bank of India. CEOInterviews.AI},
year = {2026},
month = {sep},
url = {https://ceointerviews.ai/interview/1335695/},
note = {Speaker-attributed transcript with timestamps}
}
