Mumbai, Nov 12 Over half of Indian companies expect AI workloads to grow by more than 50 per cent in three to five years, with 51 per cent building new data‑centre capacity within 12 months, a report said on Wednesday.
The report by networking gear major Cisco noted that 91 per cent of Indian organisations are deploying autonomous AI agents, adding that only 37 per cent can properly secure them.
Globally, 13 per cent of organisations that have deployed artificial intelligence (AI) are outperforming peers by making fundamental infrastructure choices that create compounding advantages, the report added.
Calling such organisations "pacesetters", Cisco said that 97 per cent of global pacesetters have deployed AI at the scale and speed needed to unlock use cases and achieve ROI.
"They build network-first foundations, prioritize power infrastructure, optimise continuously, and embed security from day one," the report said.
"As much as 45 per cent of Indian companies without this level of architectural foresight risk accumulating ‘AI Infrastructure Debt'," said Simon Miceli, Managing Director, Cloud and AI infrastructure, Asia Pacific, Japan and Greater China, Cisco.
This technical debt can compound into operational risk, security exposure, compliance challenges, and competitive disadvantage, the report noted.
Cisco identified four architectural choices that distinguished AL leaders as they prioritise continuous optimisation after deployment, treat the network as a foundation, estimate power early, and build security as infrastructure from day one.
While most respondents focus on compute, pacesetters prioritise network infrastructure, with 81 per cent of global Pacesetters rating their network as 'optimal' for AI workloads, the report noted.
“Pacesetters” are winning not because they spend more, but because they made architectural bets early—before workloads demanded them, before bottlenecks emerged, before security became urgent, the report said.
Disclaimer: This post has been auto-published from an agency feed without any modifications to the text and has not been reviewed by an editor