Zoho’s FY25 profit margins dropped to 27.8% from 34.8% in FY23 as AI infrastructure costs rose. Co-founder Sridhar Vembu warned that current AI spending levels are becoming economically unsustainable, forcing the company to pivot toward more efficient, deterministic AI solutions to protect its bottom line.
Zoho is re-evaluating its artificial intelligence strategy as rising infrastructure costs begin to pressure the company's financial performance. Sridhar Vembu, co-founder and chief scientist, highlighted the increasing strain on corporate balance sheets caused by heavy investments in large language models and massive data center requirements. This shift reflects a broader industry challenge where the ability to monetize AI investments remains uncertain for many software businesses.
Financial Pressure on Margins
The impact of these rising costs is visible in Zoho’s recent financial results. The company’s consolidated net profit margin fell to 27.8% in FY25, compared to 34.8% in FY23. Total expenses increased to ₹6,234 crore in FY25, up from ₹4,899 crore in the previous year. This tightening of margins, driven largely by the high price of computing hardware and memory consumption, has led the firm to slow down hiring and focus on internal efficiency to protect its profitability.
Strategic Shift to Deterministic AI
To manage these operational costs, Zoho is shifting its focus from general-purpose large language models toward what the company calls deterministic AI. By prioritizing tools like Zia Chat, which embeds guardrails to control business data, the company aims to reduce the unpredictable and expensive nature of token-heavy AI. This strategy is designed to provide enterprise-grade accuracy while lowering the compute resources required for each task, helping the company avoid the high operational overhead associated with standard generative AI models.
Future Outlook and Hardware Costs
Looking ahead, Zoho is exploring the development of sovereign data centers to gain more control over its infrastructure. However, the company remains cautious about the financial commitment, indicating that such investments depend on the normalization of chip and hardware costs. The company is wary of relying on borrowed capital for these projects while the industry is still determining the true return on investment for generative AI. For now, the key monitorable will be whether this efficiency-focused strategy can stabilize margins or if the rising cost of computing hardware eventually necessitates price increases for customers.