Operant AI launches infrastructure program to secure India’s 80,000-GPU surge

As India races to cement its position as a global AI powerhouse, the infrastructure supporting this ambition is growing at a staggering rate. With more than 80,000 GPUs currently in operation and data centre capacity projected to hit 8–9 GW by 2030, the country is effectively building a "National AI Factory." However, this explosive growth has created a critical security vacuum: the Inference Gap.

San Francisco-based Operant AI announced the launch of its AI Infrastructure Ecosystem Partnership Program to address this exact vulnerability. By embedding real-time AI defence directly into the inference infrastructure, Operant is providing the "inline" security layer required to govern autonomous agents and MCP-connected systems.

The security imperative of 80,000 GPUs

India’s AI infrastructure expansion is no longer just a private sector play. Under the IndiaAI Mission, the government has deployed 38,000 GPUs, with another 20,000 recently announced at the India AI Impact Summit 2026. Major players like Yotta, NxtGen, and Neysa are providing the compute, but according to the FICCI–EY Risk Survey 2026, 61% of Indian leaders now see cyber-attacks as a board-level financial risk.

The problem? Traditional security frameworks cannot keep pace with the sub-second reasoning of modern AI. When an autonomous agent accesses an ERP or CRM via the Model Context Protocol (MCP), the attack surface expands instantly to include prompt injections and tool-level manipulations.

GPU-Accelerated Defence: Blazing-fast validation

Operant AI’s solution is the AI Gatekeeper and MCP Gateway. By utilising GPU acceleration themselves, these tools provide real-time security scanning without introducing latency bottlenecks.

Key capabilities delivered at the moment of inference include:

• Prompt Injection Blocking: Preventing malicious inputs from altering model behaviour or extracting sensitive data.

• Zero-Day Protection: Monitoring for anomalous behavioural patterns in how agents interact with tools.

• Intelligent Rate Limiting: Optimising operational costs by preventing "token-draining" automated attacks.

From Apple to ARM to India: Expert-led innovation

The vision is driven by Vrajesh Bhavsar, Co-founder and CEO of Operant AI, whose pedigree includes building the Secure Enclave for Apple’s iOS and leading the ML business unit at ARM.

"We are at an inflexion point where the scale, speed, and autonomy of AI systems have outpaced the security controls designed to govern them. The path to responsible AI isn't just about building better models; it's about securing them at the moment they matter most," said Vrajesh Bhavsar, Co-founder and CEO at Operant AI. “As Indian enterprises deploy AI models and agents across financial services, healthcare, and public sector environments, the inference layer is where security must be enforced. Our GPU-accelerated AI Gatekeeper and MCP Gateway, combined with the GPU Ecosystem Program, deliver the speed and protection that India’s production AI systems demand.”

Bhavsar further emphasised the foundational nature of this shift:

“As autonomous agents become more sophisticated and models take on increasingly critical roles, securing the inference layer is no longer optional. It is the foundation on which safe, agentic systems must be built. We're transforming models and agents from vulnerable systems into trustworthy, production-ready intelligence that organisations can deploy with confidence. Our goal is to ensure that the AI momentum is not only powerful, but also secure.”

A unified stack: The Tenstorrent partnership

The program is already gaining traction with hardware innovators like Tenstorrent. By pairing Tenstorrent’s high-throughput silicon with Operant’s runtime monitoring, customers get a single, integrated stack that is both performant and verifiably secure.

“As AI moves to always‑on agents, the bar for infrastructure gets higher: it has to be performant and open by design. Partnering with Operant AI lets Tenstorrent customers pair our high‑throughput, Tensix‑based inference platforms with real‑time visibility of agents at the inference runtime layer, enabling them to scale their AI initiatives with confidence,” said Aniket Saha, VP of Product Strategy, Tenstorrent.

Conclusion

As India moves from AI experimentation to "always-on" autonomous agents, the infrastructure must be "secure by design." Operant AI isn't just offering a tool; it's offering a foundation of trust. For GPU cloud providers and enterprises alike, the ability to protect the inference layer is fast becoming the ultimate differentiator in the global AI race.

Read More:

Enterprise AI strategy 2026: Why AI, security and data will redefine technology decisions

LED manufacturing in India: LEDX funder on global supply chain shift & innovation