Artificial intelligence has shifted from pilot projects to enterprise-wide deployment, but organizations still cannot effectively measure whether their AI spending delivers business value, according to InfoQ's 2026 DevOps and Cloud Trends Report published this week. The report, compiled by InfoQ editors and industry practitioners, found that token consumption has become a significant operational cost while reliability problems across major cloud providers have forced architects to refocus on resilience fundamentals. The analysis positions AI infrastructure, governance automation, and digital sovereignty as the dominant strategic priorities reshaping cloud architecture.
The report places enterprise AI platforms and AI gateways in the Innovators category, describing a transition from individual teams connecting their own models to centralized governance systems with approved model catalogs and provisioned workspaces. Platform engineering has advanced to Early Majority status, with teams now standardizing AI capabilities rather than debating infrastructure provisioning methods. AI agents for cloud engineering moved from Innovators to Early Adopters as hyperscalers shipped agent registries, DevOps agents, and sandboxed execution environments, though enterprise adoption remains constrained by governance and compliance requirements. The Model Context Protocol, introduced by Anthropic in late 2024, has reached Early Majority adoption as the default integration standard for AI tooling, with centralized authentication plugins addressing earlier security concerns. Sovereign cloud strategies entered the Innovators category as European organizations evaluate keeping data and models within regional boundaries, though practitioners acknowledged that full sovereignty would require rebuilding nearly every application and system built on American platforms.
The report identifies token economics as an unsolved challenge, with existing FinOps tools unable to connect AI expenditure to outcomes. According to panelist Matt Saunders, "The FinOps tools can tell you that Matt Saunders spent X, Y, Z dollars on tokens in Opus, in Fable, in Sonnet, et cetera. But none of them that I know of can actually relate that back to outcomes." The FinOps Foundation has established a separate Tokenomics Foundation to address what panelist Shweta Vohra characterized as "agents' chaos" exceeding the complexity of earlier microservices proliferation. On reliability, panelist Renato Losio stated the most surprising development had nothing to do with AI: "What really surprised me is how poor the reliability of the major cloud services has been the last year." The report notes one AWS region remained offline for six months, while a US Virginia outage in October disrupted large portions of the internet for multiple hours.
The report attributes the measurement gap to AI costs functioning as proxies for unpredictable variables rather than deterministic line items like storage classes or data transfer. Organizations face uncertainty about whether current model pricing reflects subsidized lock-in strategies or genuine long-term costs. The reliability problems stem from new AI workloads straining capacity at the same time organizations prioritized adopting novel services over multi-region design and operational readiness. In regulated environments, enterprise AI agent adoption encounters a specific bottleneck: individual teams deploy agents without shared foundations, duplicating work and colliding with security frameworks including DORA requirements. The sovereign cloud challenge reflects architectural constraints where European infrastructure providers can supply basic computing and storage but lack equivalent platform and software-as-a-service offerings to replace systems like Salesforce. One panelist working with regulated European clients reported that approximately half are gradually migrating workloads back to on-premises infrastructure to maintain regional data control.
The report recommends organizations resist "agent washing" where products add autonomous features without delivering value, particularly in regulated sectors where human oversight remains mandatory. Platform teams should focus on becoming AI-native enablers that standardize capabilities and prevent shadow platform proliferation rather than creating bottlenecks. The report warns that several rapidly rebranded cloud AI services will likely disappear within twelve months as developers prioritize outcomes over specific model implementations. Organizations positioned for durable results will pair AI-enabled execution with platform governance and measurement systems that convert investment into value rather than complexity and untracked costs, according to the analysis. The core tension enterprises must resolve is between the momentum behind AI adoption and the absence of tools proving whether that adoption justifies its expense. The reliability failures suggest that operational fundamentals erode quickly when leadership attention shifts entirely toward new capabilities, creating compounding risk as AI workloads add unpredictable demand to already strained infrastructure.