Zoho Launches In-House Server ‘Nathu La’ To Lower AI Inference Costs
Zoho unveiled Nathu La, an in-house server it says can cut AI inference costs and power use while reducing foreign dependencies.
Intelligence analysis by GPT-5.4 Mini
Zoho says its five-year, India-built server platform is designed to make AI inference cheaper and more efficient across its own stack. The company frames it as both a cost play and a move toward greater technology sovereignty.
Zoho built its own computer box for running AI jobs, like making its own kitchen instead of renting one. It says the box uses less electricity and costs less to run, so the company can do more AI work without paying as much to others.
Analysis
What Zoho launched
Zoho has introduced an in-house server platform called Nathu La, built to reduce AI inference costs across its application and SaaS infrastructure. The company says the platform was developed over five years by its internal R&D teams and manufactured with Indian electronic manufacturing partners.
What Zoho claims it can do
According to Zoho, Nathu La offers 20-30% lower total cost of ownership and up to an 18% reduction in power consumption. The hardware is based on Intel Xeon 6 processors and includes in-house engineered motherboards, DC-SC (Data Centre Secure Control) modules, and network cards. Zoho also says it has filed five patents related to thermal management and modular server architecture.
The company’s pitch is not only about cost. It also says owning more of the stack reduces exposure to foreign licensing dependencies and external security audits. Zoho said the servers will be used across its software and infrastructure for AI inference, virtualisation, high-performance computing, and storage, with the added goal of improving data governance for global customers.
How far along it is
Zoho says it has already deployed a few hundred Nathu La servers, with around 1,000 in production and pre-production, and a target of 2,000 by year-end, according to Moneycontrol. CEO Shailesh Davey said the company is compounding benefits by using contextual, right-sized models on its own platform and its own servers.
Why the timing matters
The launch comes as enterprise AI usage pushes up spending on inference, which Zoho and its cofounder Sridhar Vembu have linked to India’s technology import bill. Vembu has compared the issue to an “oil import bill” for the AI era, warning that dependence on foreign AI compute infrastructure could become a vulnerability for countries like India.
Key points
- Zoho says Nathu La is its in-house server platform for lowering AI inference costs.
- The company claims 20-30% lower total cost of ownership and up to 18% lower power consumption.
- Zoho says the platform was built over five years and manufactured with Indian partners.
- The company says it has already deployed a few hundred servers and wants 2,000 by year-end.
- Zoho frames the launch as a step toward greater technology sovereignty and less dependence on foreign infrastructure.
If Zoho’s claims hold up, Nathu La could lower the company’s AI running costs and cut electricity use at the same time. It could also reduce reliance on outside software licenses and security checks, which Zoho says improves control over its own systems.
The reported savings are company claims, so the real benefit may be smaller once the hardware is deployed at scale. Even with in-house servers, Zoho still depends on outside chips and manufacturing inputs, which means full independence is limited.
