Investing.com -- Analysts turned more constructive on several AI infrastructure providers this week, arguing that the next phase of artificial intelligence spending will be driven increasingly by inference workloads and enterprise adoption of open-source models.

Truist upgraded CoreWeave to Buy from Hold, saying growing adoption of open models and sovereign AI should broaden demand beyond large hyperscaler customers. The brokerage said the company remains the leading neocloud provider, with about 1 gigawatt of active power online, a large revenue backlog and significant contracted capacity. Truist also sees potential upside to profitability if GPUs remain economically useful beyond their current six-year depreciation schedules.

The analyst calls reflect a growing belief on Wall Street that the next phase of the AI infrastructure cycle will be driven less by model training and more by inference, as enterprises increasingly deploy open-source models and AI agents in production.

Firms such as CoreWeave, DigitalOcean and Nebius are seeking to capitalize on that shift by expanding beyond GPU rentals into full-stack platforms that combine computing capacity with software, data and inference services, positioning themselves as alternatives to traditional hyperscale cloud providers.

The firm acknowledged risks from Meta's planned cloud initiatives and potential policy constraints on data center development, but believes demand for AI compute remains strong enough that any capacity not renewed by large customers could be absorbed by other enterprises. Truist cut its price target slightly to $126 from $131 while maintaining revenue forecasts.

Separately, Baird initiated coverage of DigitalOcean with an Outperform rating and a $165 price target, describing the company as a cloud platform evolving from its traditional developer-focused roots into a full-stack AI infrastructure provider. The brokerage highlighted DigitalOcean's focus on digital-native enterprises, its growing inference and agentic AI capabilities, and a more flexible capacity model that avoids large take-or-pay commitments.

Baird said DigitalOcean's existing developer community, extensive product portfolio and AI-native cloud platform position it to benefit as customers increasingly seek integrated solutions spanning infrastructure, data management, inference and AI agents. The firm expects accelerating AI-related recurring revenue growth and views the company as well placed to capture demand from smaller enterprises underserved by hyperscalers.

Baird also initiated coverage of Nebius with an Outperform rating and a $250 price target, arguing the company is among the strongest beneficiaries of the shift from AI training to inference. The firm pointed to Nebius' full-stack cloud platform, expanding software capabilities and aggressive acquisition strategy aimed at strengthening its inference offerings.

According to Baird, Nebius has differentiated itself by owning much of its infrastructure, designing hardware in-house and building software products such as its Token Factory inference platform. The brokerage expects the company to deliver some of the fastest growth in the sector as it expands capacity, diversifies its customer base and capitalizes on rising enterprise demand for AI infrastructure and inference services.

Across the sector, analysts said growing enterprise adoption of AI, increasing use of open-source models and rising inference workloads are creating opportunities for specialized cloud providers that offer both computing capacity and software services, supporting a favorable outlook for AI infrastructure spending over the coming years.