Public clouds add RAG, lose out on CWS; invest in new regions In 2025

A newly dynamic AI-focused public cloud platform vendor landscape is providing customers with more options, and in 2025, their choices will become clear. As sovereignty, cost, and data ownership concerns rise, private clouds will thrive — but not among the usual suspects. VMware revenue will continue to drop as the customers  accelerate cloud migrations in search of relief from price hikes. 

Hyperscalers will adapt by planting sovereign-focused cloud regions in Europe and beyond, while maneuvering for dominance in emerging markets. But, hyperscaler data center (DC) buildouts and AI focus will come at a cost. They must provide GPU firepower for a growing number of GenAI services, forcing postponement of sustainability goals. Furthermore, hyperscaler AI pre-occupation will strengthen the hand of independent multi-cloud-capable security vendors.

Integrated RAG services become the hottest new cloud services. AI has dominated business and tech conversations over the past two years. Cloud is no exception. Last year cloud leaders attempted differentiation through AI infrastructure and foundation models (FMs). 

In 2025, the focus will shift to retrieval-augmented generation (RAG) services with every major hyperscaler launching  solutions. FMs sit at the core of GenAI, but they struggle with hallucination, low accuracy, irrelevance, and coherent nonsense in business contexts. RAG combines the strengths of information retrieval and generative FMs to address these challenges. 

Amazon Web Services (AWS), Azure, Google Cloud, and Alibaba  Cloud will integrate vector database, embedding, and vector search in RAG services leveraging homegrown or open-source RAG frameworks like Lang Chain and Llama Index.

Private cloud gains momentum with VMware alternatives. On-premises computing (by any name) is on the rise again as companies solve sovereignty, cost, and data ownership / security challenges. Most large enterprises already take a hybrid cloud approach, and they will further invest in private cloud for workloads requiring data storage and processing on premises for security, privacy, and regulatory compliance (e.g., pretraining and fine-tuning of FMs, RAG integration, AI agent automation). 

However, newcomers and private cloud expanders won’t likely increase business with dominant private cloud player VMware given recent bundling and pricing changes. In 2025, Forrester predicts that most major public cloud providers will increase investments in private cloud. Hyperconverged offerings like Nutanix and open-source projects like OpenStack will see increased interest.

OpenStack’s success will require an improved tech vendor ecosystem. But, keep in mind that private cloud growth will occur in parallel to public cloud’s, not at its expense, and many VMware shops will renew at a limited scale.

Platform specialists squeeze out cloud infrastructure platforms’ security capabilities. Cisco, Fortinet (with its recent Lacework acquisition), Palo Alto Networks, and Wiz will continue to significantly invest in coherent cloud workload security (CWS) solutions with cloud security posture management, cloud infrastructure entitlement management, infrastructure-as-code scanning, as well as container security functionality sets, centralized and GenAI-supported policy management, detection, response, and reporting. 

With Google’s failed $23 billion acquisition of Wiz, platform specialists will pose stiffer competition to AWS, Azure, and Google’s native CWS capabilities. By the end of 2025, 60% or fewer cloud customers will prefer the platform’s native CWS capabilities, while the remaining 40% will use a platform specialist CWS vendor.

Hyperscalers compete over African cloud regions, while Oracle expands in Middle East. Over the past five years, major cloud players have doubled down on investments in Europe and APAC, and pushed further into the Middle East. In 2025, the focus will shift to the next battleground: Africa. US and Chinese providers already have footholds in South Africa. Microsoft and G24 have pledged $1 billion to launch new green East African cloud regions. 

Huawei opened a Cairo-based DCearly in 2024. Huawei and Alibaba announced hundreds of millions of dollars for two new regions each. Over 2025, expect more announcements from the major players to strengthen coverage across South Africa, Nigeria, and Kenya. All are partnering locally to appeal to sovereignty concerns, but Oracle will play the strongest card as part of its Middle East expansion. We expect Oracle to win at least five major cloud accounts across the Gulf region through its dedicated-region OCI and telco-delivered Alloy OEM program offerings.

Hyperscalers miss CO2 emissions goals — again — but, with improvements YOY. Power demand from AI and GenAI workloads, and new DC buildouts, made Google and Microsoft miss their CO2 emissions goals. Google emissions were up by 48% since 2019, and Microsoft’s rising 29.1% since 2020. The International Energy Agency believes that global DCs’ electricity could more than double by 2026 (up from 460 TWh in 2022) largely due to AI workloads, and the amount of data feeding into the models. 

To slow this trend, cloud players incorporate renewable energy, custom silicon designs, and direct-to-chip liquid cooling while exploring new areas like microfluidics at the chip level. In 2025, hyperscalers will announce more of such initiatives, but they won’t make a dent in the near term.

Forrester predicts that Microsoft and Google emissions (from its 2019 baseline) will rise to 20%, but with marked improvement from 2024. With liquid cooling efforts and few DC expansion plans, we expect emissions to decrease by 10% YOY.

-- Forrester Research, USA & Australia.