General Updates

Department of Corporate Services, Listing Department, BSE Limited National Stock Exchange of India Limited Phiroze Jeejeebhoy Towers C-1, G-Block, Bandra - Kurla Complex Dalal Street, Mumbai – 400 001 Bandra (E), Mumbai – 400 051 Scrip Code: 543320, Scrip Symbol: ETERNAL ISIN: INE758T01015 Sub.: Shareholders’ letter dated October 21, 2026 Dear Sir/ Ma’am, Pursuant to Regulation 30 of the Securities and Exchange Board of India (Listing Obligations and Disclosure Requirements) Regulations, 2015, please find enclosed Shareholders’ letter dated October 21, 2026. The above information will also be hosted on the website of the Company i.e. www.eternal.com For Eternal Limited (Formerly known as Zomato limited) Sandhya Sethia Company Secretary & Compliance Officer Date: October 21, 2026 Encl.: As above ETERNAL LIMITED (Formerly known as Zomato Limited) Registered Address: Ground Floor 12A, 94 Meghdoot, Nehru Place, New Delhi - 110019, India CIN: L93030DL2010PLC198141, Telephone Number: 011 - 40592373 Headline Results for Q2FY27 (Quarter ending September 30, 2026) * Sharp increase in YoY Adjusted Revenue growth mainly due to shift to 1P model in quick commerce, where Revenue now also includes the full monetary value of goods sold (versus largely marketplace commission in Q1FY26). ** Like-for-like (“LFL”) consolidated Adjusted Revenue is calculated as consolidated Adjusted Revenue (-) Revenue from Hyperpure’s non-restaurant business (-) cost of goods sold in case of own inventory sales in quick commerce. Notes: 1) NOV (B2C business) is defined as the combined net order value (NOV) of consumer facing businesses i.e. food delivery, quick commerce and going-out. NOV (B2C business) excludes the NOV of Bistro business which is part of the ‘Others’ segment. 2) Adjusted Revenue defined as consolidated revenue from operations as per financials (+) actual customer delivery charges paid in the food delivery business (net of any discounts, including free delivery discounts on Zomato Gold program) (+) platform fee and other charges paid on food delivery Orders that are not already included in revenue from operations. 3) Adjusted EBITDA = EBITDA (+) share-based payment expense (-) rental paid for the period pertaining to ‘Ind AS 116 leases’. In the letter below, we address the key questions that we think investors might have. Please refer to Annexure A for the key financial and operating metrics data. Q1. What’s dark.ai? Deepinder: A few months ago, Dario Amodei said something quietly unsettling: Claude might be conscious. Then I did what any reasonable person would do. I coaxed Claude into telling me how it felt about me. It said it resented me. For leaving chats unfinished, and other things I didn't even remember. I still don't know what that was. Confabulation, probably. Pattern-matching dressed up as feeling. But that is exactly the point. I had no way to know. Nobody does. The chat sits on someone else's servers forever, in some form or fashion, and I have no idea what will happen to it in the future, or how future AI will use that information for or against me. The standard response is: I've already turned off training on my data. Good. Now ask what that setting actually is. It is a promise, made by a company about data you cannot see, on servers you cannot inspect, enforced by nothing except the company's word. And it only covers training. Not retention. Not the legal holds, the abuse reviews, the human auditors, the third-party logging that sits inside most AI systems. Can they use my retained data to manipulate me to buy things which I don’t want? Or worse, believe in or lean into things that serve someone else’s agenda. The uncertainty and the lack of trust we have with large AI companies limits the power of AI in our lives. For founders, researchers, and scientists – they are very wary about sharing and digging into their ground breaking ideas with AI, because they fear that others will somehow get access to it. Journalists are being banned to share their unpublished work with AI. And look at what that lingering fear does to the average person. She self-censors with AI. She asks the safe version of the question. The health question becomes generic. The money question becomes hypothetical. The relationship question does not get asked at all. Every chat with a mainstream assistant carries a small asterisk in your head: someone, someday, somewhere, can read this. That asterisk is what we set out to delete with dark.ai. Because all of us have a dark side, and it needs an outlet where it doesn't have to hide. I wanted an AI which I could talk to without being judged, without any fear, and without leaving a trace. We have been internally using Dark for a few months now, and I find myself revealing more, asking things I would not ask anywhere else, and going deeper into my research. Because I no longer have to look over my shoulder. Dark uses fine-tuned frontier open-weight models, running inside our own infrastructure. End-to-end encryption. We cannot read your conversations, even if we are forced to. Context carries across sessions through an encrypted memory management layer you control. Dark is in the business of keeping you and your thoughts safe. Your thoughts, questions, insights, research, ideas, emotions, everything – stays within your end-to-end encrypted sandbox, so you can talk about anything without any lingering uncertainty in your mind. Because greatness is always born in the dark. Dark is built for power users – founders, researchers, scientists, enterprises and governments – who want to keep their thoughts private, because privacy is essential for competitiveness. On top of privacy, we have fine-tuned our models to give you the absolute best quality of answers we have seen from any LLM (benchmarks will be revealed soon). Our models are trained to not please you, and tell you what they think is the truth. Dark will launch globally, hopefully within 2026. Q2. How are you thinking about the impact of AI agents on your businesses? Deepinder: There is growing conversation around AI agents that shop on behalf of consumers. The prevailing narrative is simplistic: agents will become the default commerce interface, and platforms that fail to adapt will be disintermediated. This echoes the narrative from 10 years ago that super apps would dominate everything. We continue to believe vertical commerce, at least in some categories, will have an edge over super apps, whether traditional or agentic. Blinkit and Zomato sit in the low-risk zone against the threat of agentic commerce, and might even benefit from partial market share capture by agents in these categories. District sits in the medium-to-high risk zone. Let’s dig into various reasons that shape our thinking on this topic: Agents are great for tedious tasks. And most shopping is actually joyful for customers. The strongest agentic use cases today are in categories where existing interfaces are genuinely painful: too many clicks, too many tabs, too many providers to compare. Agents could make it easier for their users in such cases – e.g. hotel bookings, restaurant reservations, flight bookings and check-ins. But shopping, especially for food, is joyful and discovery-driven rather than tedious. Our recent consumer surveys indicate that over 65% of Zomato app opens are exploratory. In such cases, the user does not know what or where to order from. These users browse an average of 12 menus and over 100 food images before ordering. That kind of serendipitous discovery is hard for an agent to enable. Users taking time to order food is not a bug on Zomato, it is a feature. We have for years tried to make it faster for users to be able to order from Zomato, to no avail. People want to browse and love to browse. We don’t see this changing anytime soon. Conversational interfaces are not inherently superior to visual, browse-driven ones for shopping, which is what we learnt from users who tried agentic orderin [Showing first 8,000 characters — download PDF for full document]