At Bloomberg New Economy 2026, Infosys CEO Salil Parekh warned that the high costs of data centres and chips, combined with emerging AI models driving down inference expenses, present significant risks to AI infrastructure investments and the sustainability of capacity usage in India’s tech sector.
- High data centre and chip costs threaten AI capacity investments.
- Cheaper AI models could reduce demand for expensive infrastructure.
- Data growth and sovereign cloud rules further constrain capacity expansion.
What happened
Infosys CEO Salil Parekh addressed the challenges faced in expanding AI infrastructure during the Bloomberg New Economy 2026 event. He underscored the significant costs involved in building data centres and procuring chips, which are currently very expensive. This raises concerns about the return on investment and how long existing capacity will remain effective amid rapid technological changes.
Parekh also highlighted the arrival of new foundation AI models developed differently from the dominant closed models, resulting in lower inference costs. This shift in AI economics could reduce the requirement for high-cost infrastructure, posing a risk to the utilization rates of the capacity being built today.
Why it matters
The costs and risks identified by Parekh have broad implications for the AI ecosystem in India, where Infosys plays a leading role. High upfront investments in data centres and chips could slow infrastructure rollout if capacity becomes underused due to cheaper AI models reducing demand for expensive inferencing power.
Moreover, the growing volume of data and stricter sovereign cloud regulations add complexity to infrastructure planning. Ensuring workforce readiness through effective training is critical for enterprises to fully adopt advanced AI technologies and transition legacy systems, maintaining competitive advantage in an evolving landscape.
What to watch next
Stakeholders should monitor how AI model innovation affects infrastructure demand, particularly if lower-cost inference models become widely adopted, potentially reducing pressure to scale expensive data centres. The pace at which enterprises integrate these models alongside legacy system upgrades will also be key.
Additionally, tracking developments in data sovereignty regulations and their impact on infrastructure investment will be important. Infosys’ ongoing efforts to train and skill employees in AI applications will influence how effectively Indian enterprises leverage new AI capabilities despite these cost and capacity challenges.