Why AI success depends on a resilient, cloud-first digital core

By Saju Sankarankutty, SVP, Service Offering Head, Cloud, Infosys

Worldwide AI spending will reach $3.49 trillion in 2027, with AI infrastructure being the biggest spend bucket; this is as per a recent Gartner research Investment in AI infrastructure is climbing sharply, but production maturity has largely not kept pace. Probable causes include unclear business gains, inconsistent human-machine collaboration, and weak data foundations.

While capital investment in computational infrastructure is accelerating significantly, production maturity has not advanced at a proportional rate. Contributing factors could range from ambiguous financial returns and erratic human-machine interaction protocols to deficient data architectures. In a nutshell, investment in AI spending is rising faster than actual returns.

A robust digital core is essential for the operational foundation that AI needs for speed, scale, and autonomy. An integrated combination of cloud platforms, networks, data pipelines, and security controls, forms the foundation on which an organization can move experiments to production, and reliably deliver services to millions of users every day.

Start with reliable data for business

A recent research by Forrester on data quality platforms indicates that buyers want platforms that govern all kinds of input data formats. Platforms must be traceable and validated continuously. Together, these capabilities are a practical starting point for AI at scale.

Under the RBI’s Digital Lending Directions, 2025, borrower data must be collected with explicit consent accompanied by an audit trail, stored on servers in India, and repatriated within 24 hours if processed beyond sovereign boundaries. This indicates that a lender running credit decisions through a model has to show which fields fed a decision, on what consent, and where the data sat when it was made.

Strong governance and compliance are table stakes in regulated industries.

Prioritize cloud-first as an operating model

Production AI is typically unpredictable and compute-intensive, with training cycles, inference bursts, and agentic workflows creating uneven demand patterns. Hybrid architectures that enable data movement efficiency, consistency, and recoverability are building blocks of a solid cloud-first foundation. They offer the elasticity, managed AI services, and global reach needed to scale AI workloads.

Teams often find it difficult to manage the complexity of cloud footprint expansion, placing strain on observability and response. The solution is operations that monitor, triage, and respond automatically inside set governance boundaries. Escalation to people is by exception.

Define resilience as a key metric

As per a survey by the Uptime Institute, 57 percent of respondents indicated that the cost of their most recent major outage was more than $100,000. For the second year running, 20 percent of respondents said that the costs were more than $1 million.

A research by Infosys Knowledge Institute frames this resilience as three connected disciplines covering the mitigation of operational disruption via dynamic scalability for elasticity and high availability, the containment of impact through automated failover mechanisms to maintain fault tolerance, and the execution of rapid service restoration procedures following critical incidents for disaster recovery.

Combining AI across customer service, fraud detection, supply chain, and other core operational areas will bring system availability, revenue protection, and continual real-time decision-making. The reliability of systems goes beyond the engineering of the technical components and establishes an essential legal and fiduciary obligation to customers and shareholders.

Assurance of operational continuity and rapid recovery driven by automated failover protocols and regional redundancy delivers continuous availability. Utilization of diagnostic tools identifies and remediates infrastructure anomalies prior to end-user impact through advanced observability.

Strengthen security

AI deployments often expand attack surfaces. Models pull data from disparate sources, connect to more systems, and increasingly act on their own, making secure, well-governed networks essential. Zero-Trust architectures, encryption in transit and at rest, and identity-based access controls protect data as it moves across distributed environments.

In a nutshell

Organizations must treat the digital core as a strategic asset by being cloud-first for scale, resilient for continuity, secure and data-ready for trust. Those that build AI on brittle infrastructure will keep generating pilots that don’t reach production, or production systems that fail when they matter the most. Successful AI means that organizations’ infrastructure can carry what they are trying to build.

AI ambition is abundant and increasingly well-funded. The key requirement is a digital core that is strong enough to carry it – reliably, securely, and at scale.

AIAI SpendingCloud
Comments (0)
Add Comment