Deepinder Goyal’s Dark.ai could be Eternal’s dark horse, but is it too early to bet?

An early-stage experiment that could eventually open a new avenue for Eternal beyond food delivery, quick commerce and going out, Deepinder Goyal’s announcement on Dark.ai has drawn attention to the privacy-focused artificial intelligence assistant as a potentially high-margin business, although its commercial viability remains untested.

Eternal announced Dark.ai after developing it for about a year. The product is being tested internally and is expected to launch in the coming months. It is aimed at power users including founders, researchers, enterprises and governments, and will use fine-tuned open source or “open weight” models running on Eternal’s own infrastructure.

The product’s central pitch is privacy. Dark.ai is expected to use end-to-end encryption and a user-controlled encrypted memory layer, meaning Eternal would not be able to read user conversations even if compelled to do so. In Goyal’s words, the company wants to remove the concern that “someone, someday, somewhere, can read this.” “That asterisk is what we set out to delete with dark.ai,” he said in a letter to shareholders.


Also Read | Eternal shares jump 2% after strong Q2 show, brokerages see up to 34% upside

For Eternal, however, Dark.ai is still an experiment rather than an earnings driver. Management has described it as a “zero to one” bet: the business could be scaled globally if it works, but shut down if it fails. Costs are currently not meaningful, while any substantial investment would largely be directed towards hardware, compute and GPUs if customer response is strong.

That creates an unusual risk-reward profile. Management’s preliminary view is that Dark.ai could become a high gross margin business, reach profitability relatively quickly and potentially fund its own expansion. But pricing, monetisation and the launch strategy are still under evaluation, according to analysts.

“Management views Dark.ai as a high-risk, high-reward opportunity,” Nuvama Research said, adding that the next 12 months would be critical in determining whether Dark.ai and Eternal’s enterprise AI initiative Nugget can become meaningful businesses.

Also Read | Dark Matter: Eternal’s chatbot will keep user conversations private even from the company — Deepinder Goyal

Analyst enthusiasm is therefore tempered by the absence of operating visibility. Emkay Global described Dark.ai as a privacy-first assistant for power users running on frontier open-weight models and Eternal’s own infrastructure. While the brokerage viewed the move positively as an entry into an adjacent technology category, it retained a neutral stance on the product because of the lack of clarity and what it called a “not-so-adjacent category.”

Jefferies was more constructive on the optionality, calling Dark.ai the key monitorable after Eternal’s second-quarter results. “Not much is known today, & the outcome could be binary, but we like the experimental DNA,” it said. The brokerage highlighted the product’s end-to-end encryption and described it as a “zero to one” bet, while noting that current costs are immaterial.

"dark.ai extends Eternal into a large but competitive category, with privacy rather than proprietary foundation-model capability as the intended USP. Near-term financial implications remain difficult to assess given limited disclosure, but management expects high GM, which will self-fund this initiative, and calls it a binary outcome. We believe user adoption, independently verifiable privacy architecture, benchmarked model quality, and inference economics will determine whether dark.ai develops into a meaningful platform opportunity," Jefferies analyst Vivek Maheshwari said.

Nomura also termed Dark.ai a potential wild card for Eternal. The brokerage said the company plans to launch the AI application for heavy users in the coming months, with modest initial capital expenditure that could increase depending on the product’s success.

The company’s existing businesses remain the immediate investment case. Eternal’s food-delivery business delivered adjusted EBITDA growth in the second quarter, while Blinkit’s order value rose 22.2% sequentially and adjusted EBITDA improved to Rs 1.84 billion. Management also raised its store targets for Blinkit, reinforcing the view that quick commerce will drive the near-term operating narrative.

Dark.ai, by contrast, has no disclosed revenue, user numbers, pricing model or profitability track record. Its economics will depend on whether users are willing to pay for privacy, how much compute the service consumes and whether Eternal can differentiate a product built on open-source models.

That leaves investors with a potentially valuable option, but not yet a visible earnings stream. Dark.ai may eventually become Eternal’s dark horse, particularly if it converts privacy concerns into a global paid AI product. For now, though, the company is asking shareholders to underwrite the experiment before the market knows whether there is a scalable business behind it.

(Disclaimer: This article has been written by Nikhil Agarwal, who is not a SEBI-registered Research Analyst or an Investment Adviser. Nikhil Agarwal and his/her ‘relative(s)’ (as defined under Section 2(77) of the Companies Act, 2013) do not hold any financial interest in the companies mentioned in this article as of the date of publication. The views/recommendations mentioned in this article, wherever applicable, are those of the respective SEBI-registered Research Analyst/brokerage and have been reproduced/reported with due attribution. They should not be construed as the views or recommendations of The Economic Times Digital or the journalist. Readers are advised to consider the original research report and make their investment decisions based on their own assessment. Brokerage disclaimers here)