Predictions are a tough sport to play. If you get them all right, you played it safe by recapping existing trends or gave wishy-washy statements that are hard to verify. If you got them all wrong, you likely went for the headlines with little accountability on those claims. Every year, we try to find that balance to help guide enterprises in the year ahead. One of my favorite pastimes, however, is looking backward. The goal isn’t self-validation — it’s accountability. We ask ourselves the same questions every year: Were we right? Were we early? Did we miss important signals? Or did the market move in a direction we simply didn’t anticipate?

We just published our latest report card. And as always, we don’t grade ourselves generously. Some might call me a “tough grader.” Read the full report, Scoring Our 2025 Cloud Predictions, for more.

Now let’s talk about 2025. Although a global pandemic didn’t strike, it was a year of extreme volatility as AI acceleration collided with the cloud market amid a time of chip shortages, new (and sometimes vanishing) tariffs, increasing sovereignty concerns, and rapidly approaching sustainability deadlines. Overall, we scored better than previous years. We were spot on or partially right for a few and faced external challenges for other predictions.

Our biggest hit: Integrated RAG services become the hottest new cloud services.

We predicted that hyperscalers would rapidly integrate vector databases, embeddings, and vector search into first-party retrieval-augmented generation (RAG) services, leaning on both homegrown and open-source frameworks. This prediction was spot on. Amazon Web Services (AWS), Microsoft, and Google set the tone globally, making RAG a core platform capability in Bedrock Knowledge Bases, Azure AI Search, and Vertex AI RAG Engine. Chinese cloud providers followed in kind by embedding RAG deeply into native knowledge management offerings. Even as industry attention shifted to agents and multicloud container platforms, RAG never faded. Instead, it became the foundational backbone of context engineering for agentic workflows.

Our biggest miss: Hyperscalers miss CO2 emissions goals — again — but with improvements YOY.

We predicted that hyperscaler emissions would rise from 2019 levels but improve year over year in 2024, signaling a rebound from AI‑driven increases. That only happened for Microsoft; AWS and Google emissions went up. There are some early signs for a rebound (e.g., Google reported decreases in scope 1 and scope 2 emissions), but AI is still hitting sustainability goals significantly.

If you want the full breakdown of how each 2025 prediction scored and why, read the full report. And if you’ve followed this series before, you know what comes next: We’ll do this again next year with our 2027 predictions.

If you want to dive into this report or have any questions on FinOps, cloud cost management tooling (watch for the incoming Forrester Wave™ in Q4 2026!), cloud platform teams, cloud skills, cloud governance, or industry clouds, you can find me at inquiry@forrester.com.