Groww Builds In-House AI Stack to Reduce Compute Costs
Investment platform Groww is developing its own internal AI stack to lower compute expenses. This move comes as the company increases its AI integration and faces rising operational costs.
Indian investment platform Groww, operated by parent Billionbrains Garage Ventures, is building its proprietary in-house AI stack. The primary goal is to reduce significant compute costs that have escalated with increased AI adoption.
Co-founder and CEO Lalit Keshre stated that developing an in-house stack will lower per-token costs compared to relying on third-party commercial models. "If we had relied on third-party commercial models, our cost would have been much higher," Keshre said during the company's Q2 earnings call.
Groww has been integrating AI across its services, including trading, bonds, and wealth management. The company recently unveiled GR-1, an AI investing assistant designed to analyze markets, track news sentiment, and provide personalized portfolio insights. The tool does not execute trades automatically and requires user approval.
The company's cost to serve increased by 24% to INR 206 crore in Q2, from INR 166 crore in the previous quarter. Conversely, cost to grow declined by 29%. The rise in service costs was attributed to AI-related expenses, higher compute spend for AI workloads, and an increased number of trading days.
Groww has also integrated the wealth management offerings of Fisdom, which it recently acquired. This integration expands Groww's service portfolio and incorporates Fisdom's user base into its active customer count. The company reported a 5% increase in active users, reaching 18.2 million in Q2.