Citi analyst calls infrastructure software the top bet in the AI trade era

Fatima Bulani, co-head of US software equities research at Citi, on July 8, 2026, recommended in an interview with Bloomberg that investors bet on infrastructure software — as the most resilient segment amid AI-trading and a correction in the broader tech sector.

Three

Tiers of the AI Market: Why Infrastructure Is Not the Same as Models

The modern AI market can be conveniently divided into three tiers. The first tier comprises developers of frontier models: OpenAI, Anthropic, Google DeepMind, xAI. They compete for technological leadership, and it is their stocks that are subject to the greatest volatility: each new benchmark shifts the balance of power. The second tier is AI applications: companies that build products on top of other models. The third tier is infrastructure software.

The third tier includes cloud platform providers and specialized database vendors, workload orchestration tools and MLOps, security systems, data management and compliance tools, monitoring, observability and logging platforms, as well as tools for deploying and integrating models into corporate IT systems.

Infrastructure companies serve all participants in the AI market simultaneously: their clients include OpenAI, Anthropic, Meta, and thousands of corporate customers. Recurring subscription revenue and long-term contracts provide predictable cash flows — regardless of which model proves to be dominant.

In the investment community, this logic is called the "pickaxe and shovel strategy": in a gold rush, not only gold seekers become rich, but also those who sell equipment. In the AI race, this role is played by database providers, DevOps tool vendors, and cloud infrastructure suppliers.

Why Tech Sector Correction Is an Entry Point

After several quarters of rapid growth, the broad tech sector is experiencing a correction in mid-2026. The reasons are typical for overheated markets: inflated monetization expectations, concerns about the payback timeline for large AI capital expenditures, and instability in trade conditions.

According to Bulani, for infrastructure software this is more of an opportunity to buy rather than a signal to exit. Demand for AI computing is not decreasing — it continues to grow with corporate AI deployment. The more companies that adopt AI agents and build their own AI pipelines, the greater the demand for tools that enable their operation.

Infrastructure software historically demonstrates lower volatility compared to frontier model developers: broad diversification of the customer base and the contractual nature of revenue reduce dependence on competitive dynamics in the model race.

Corporate clients rarely abandon basic infrastructure even during cost-cutting periods: the switching cost is too high and the dependence on uninterrupted system operation is too great. This creates stable "sticky" demand even amid market turbulence.

What This Means

Citi analyst's position reflects a broader consensus among institutional investors: in the AI economy, it is not only those who create the smartest model who win, but also those who ensure its operation. Against the backdrop of uncertainty about winners in the AI model race, infrastructure software offers predictable and diversified exposure to AI trading — without betting on any specific winner.

*Meta is recognized as an extremist organization and is banned in the Russian Federation.