Red Hat Summit 2025: AI on the foundations of the hybrid cloud

“With each new technological wave, the old one does not disappear.”

This is how Ashesh Badani , Senior Vice President of Red Hat, opened the Red Hat Summit Connect 2025. Facing an audience of CIOs and IT managers, the American publisher unveiled its strategy to support the transition to AI without sacrificing existing investments.

On the agenda: a presentation of Red Hat AI, its enterprise AI platform built on the open-source foundations of vLLM and InstructLab; a global announcement of 100% European technical support for government sectors; and a look back at strategic partnerships with Intel and AWS. The message is clear: AI doesn’t replace the hybrid cloud, it complements it. The results speak for themselves, with ROIs reaching up to 668% for customers who migrated to OpenShift, from Ford to Emirates NBD. While 60% of humanity has already adopted generative AI, Red Hat is banking on continuity, interoperability, and digital sovereignty to capitalise on this momentum.

Devoteam and Open Source

At Devoteam, we integrate open-source technologies into the heart of our architectures and have contributed to the projects that have shaped the ecosystem for over twenty years. Indeed, Open Source has been a major source of innovation for DevOps and Cloud technologies, where we have developed our expertise through mastery of DevOps practices, automation of CI/CD pipelines, and adoption of Cloud-native platforms based on open standards.

Today, this approach naturally extends to emerging fields. Open-source frameworks are becoming a strategic foundation for quantum technologies and AI engineering, from model training to secure and scalable deployment. Our positioning: to maintain and develop our expertise in Open-Source technologies, ensuring innovation and interoperability.

Red Hat Study: Industry Trends and Constraints

Rémy Mandon, Country Manager for France, explains in the introduction that Red Hat Europe surveys 900 IT decision-makers every year. This year, three strategic priorities are shaping CIOs’ roadmaps:

• AI for 75%
• Hybrid and multi-cloud for 73%
• Virtualisation for 72%

Virtualisation remains the number one topic. But the emergence of containerisation has sparked debate about its future, as containerisation offers flexibility and operational advantages while questioning the long-term viability of traditional architectures.

After an intensive phase of cloud migration, the focus is now on modernising legacy applications. Companies are seeking to future-proof their infrastructures while integrating new technologies, particularly AI, which inspires as much hope as uncertainty. This dual transition necessitates a thorough review of infrastructure choices and investments.

Faced with these priorities, IT decision-makers are confronted with major constraints that hinder their transformations. Data security and sovereignty are at the forefront of concerns, reinforced by the emergence of new certifications in France and Europe, which are supplemented by the European Union’s proposed sovereignty scores. The skills challenge is also being felt. Teams are calling for a simplification of tools to facilitate their adoption in an increasingly complex technological environment.

Some quick findings from this study:

Security issues and skills

• Customers are demanding stronger guarantees with certifications such as SecNumCloud in France, EUCS and C5 in Europe, complemented by new sovereignty scores
• Companies are calling for a simplification of tools in the face of increasing technological complexity, in order to facilitate the development of internal skills.
• Uncertainty surrounding AI and virtualisation is prompting a rethinking of infrastructure choices.

Obstacles to AI adoption

• The prohibitive entry cost, concerns about confidentiality and data sovereignty, particularly in a European context
• Organisational silos between business and IT complicate integration on ageing infrastructures.
• The risk of projects being rushed by enthusiastic professionals when the infrastructure is not ready

Strategic priority areas

• Moving beyond Proof of Concepts to generate concrete value through operational deployments
• Promoting collaboration between business and IT for the successful adoption of AI
• Reduce technical debt by modernising infrastructure to ensure its sustainability

Open source as a catalyst

Faced with security and sovereignty challenges, open source is emerging as an essential solution, according to Rémy Mandon. 93% of the 900 decision-makers surveyed consider it absolutely critical, as the transparency of the source code directly addresses the requirements of trust and compliance.

Red Hat capitalises on this momentum by transforming innovation from 100,000 communities and one million developers into proven solutions, subjected to rigorous non-regression, functional and security testing before any production deployment.

Furthermore, to address the skills challenge, Red Hat offers a pragmatic approach with a unified platform. This platform leverages technologies and expertise already mastered by the teams, thus reducing the learning curve and enabling a rapid transition from proof-of-concept to tangible value creation.

45 years of open source to support the cloud and AI transition

Ashesh Badani opened his speech with a historical perspective. “ We are living in an era of unprecedented change. “ In these moments, it’s essential to look back,” he said. He then recalled the succession of technological waves driven by dominant companies: IBM and mainframes, Microsoft and Intel with PCs, Cisco in networking, and Google and Apple in the internet and mobile sectors.

But one constant remains, according to him: “ With each new wave, the old one doesn’t disappear. ” Mainframes and Windows servers are still here. They must be maintained alongside innovations, a reality that Red Hat is committed to supporting.

Lessons from the Cloud Era

The cloud era has delivered several fundamental lessons. Location has become less important with the emergence of sovereign clouds in Europe and Asia, addressing governance and compliance challenges. Consistency in operations and automation has become crucial, as has the ability to deploy at scale with built-in security. Innovation has exploded in hyperscaler services and open source, as evidenced by the CNCF, which hosts hundreds of projects. Furthermore, collaboration between teams has intensified under the banners of DevOps, DevSecOps, and platform engineering.

Ashesh Badani emphasised Red Hat’s consistent philosophy over the past 45 years: “ To bring the power of open source, the strength of collaboration and community work, to deliver enterprise products that can be deployed at scale. ” This approach has been applied across eras, from Linux to the hybrid cloud with Kubernetes, and now AI. The Red Hat platform enables the execution of containers, virtual machines, and cloud-native or legacy applications on the same infrastructure, regardless of its deployment: physical, virtual, private, public, sovereign, or edge.

Case study: the migration of the National Bank of Dubai

The case of Emirates NBD illustrates this strategy. The Dubai-based bank migrated its traditional virtualisation to OpenShift, unifying containers and virtual machines. After several months, 37,000 virtual machines are now running on this infrastructure.

Ford and General Motors are following this path, achieving impressive ROIs: between 540% and 668% according to third-party analyst studies. This success also relies on collaboration with hyperscalers.

“ Whether you choose AWS, Azure, Google Cloud, IBM Cloud, or Oracle, Red Hat OpenShift and Ansible are available directly from their consoles,” concluded Ashesh Badani, adding that “ the ecosystem is at the heart of our open-source economy. ”

Applying to AI the strategies that made hybrid cloud successful

Ashesh Badani continued his presentation by addressing AI, beginning with a reminder: “ We tend to overestimate the value of a technology in the short term and invariably underestimate it in the long term .” But generative AI is breaking all records. According to recent studies, 60% of the world’s population has already used generative AI. “ Never before have 8 billion people agreed so quickly on adopting a technology .” An unprecedented surge in technological history.

The Red Hat executive was keen to apply the lessons learned from the past. “ The principles we have successfully applied to Linux, containers, Kubernetes, private and public cloud, and automation, we will now apply to AI .” Red Hat is conducting approximately 1,000 AI engagements worldwide this year, falling into three categories. Risk management remains the most traditional use case, widely used by financial institutions and e-commerce businesses for fraud detection. But with generative AI, two other areas are emerging rapidly: revenue generation and cost optimisation. Conversational assistants are proliferating, while code generation tools and process optimisation are producing measurable results.

Beyond AI POCs

However, widespread adoption carries significant risks. Badani noted that 60% of CEOs and CIOs consider AI a top priority, while 58% of CIOs plan to allocate at least 5% of their IT budget to it. This pressure from both the top and bottom is accompanied by a worrying observation: “A significant number of AI projects fail or are abandoned after the proof-of-concept phase ,” with an MIT study revealing even higher failure rates.

To shed light on the future of AI, Badani invoked Jevons’ paradox. This 19th-century British economist observed that improvements in the energy efficiency of steam engines did not reduce coal consumption; they caused it to skyrocket. “ We anticipate exactly the same phenomenon with AI. The more efficient we make the use of underlying resources, the more use cases will proliferate,” he explained.

Red Hat’s strategy remains true to its principles: “ Any cloud, any AI accelerator, any model. ” This multi-environment abstraction fully applies to AI. vLLM and InstructLab form the foundation of this new era because “ inference efficiency is becoming the central challenge. ”

An agent framework based on the MCP

Above this layer, Red Hat is developing an agent framework utilising the Model Context Protocol (MCP) for seamless integration with existing systems and new AI APIs. Coexistence is the guiding principle. CPUs and GPUs side by side, Linux and vLLM together, Kubernetes and InstructLab in parallel, traditional applications and new AI agents on the same infrastructure. “ We build on your existing hybrid cloud investments, without disruption. ”

Red Hat AI, the recently announced platform, consolidates all these technologies. It integrates optimisation techniques like slicing, quantisation, and scheduling, offers GPU-as-a-Service and Model-as-a-Service, and includes guardrails, bias detection, and hallucination protection. “ Red Hat AI Stack is built on the open hybrid cloud stack you already know,” Badani concluded. “ All the investments you’ve made, all the technological skills you’ve developed, remain relevant. ”

Intel and Red Hat

Jean-Laurent Philippe, representing Intel France, took advantage of the Red Hat Summit to highlight the profound transformation the American semiconductor giant is undergoing under the leadership of its new CEO, who arrived in March. The company is pursuing a dual strategic objective: to catch up in the foundry sector with the launch of the Intel Core Ultra 3 processor, manufactured using 18A technology (1.8 nanometers), while simultaneously making a major push into AI.

“ By 2028, approximately 80% of AI infrastructure will be dedicated to inference and only 20% to training .” This explains why Intel is focusing its efforts on this phase, where its Xeon processors excel, particularly thanks to their doubled energy efficiency in the latest generation. He emphasised Intel’s DNA of interoperability. From the invention of x86 to the democratisation of Wi-Fi, this finds a natural continuity in open source.

This philosophy has fueled a 25-year partnership with Red Hat, enabling the rapid deployment of optimised AI solutions. Beyond raw performance, Intel prioritises security with technologies like Intel TDX (Trust Domain Extensions). TDX isolates entire virtual machines in secure enclaves. This approach has been adopted by BBVA, Spain’s second-largest bank, as part of its Red Hat OpenShift Virtualisation infrastructure.

Digital Sovereignty

Next up was Rémy Mandon, who placed digital sovereignty at the heart of his speech. This is a central concern for European CIOs. Drawing on statistics from Red Hat’s annual survey, he reiterated that sustainability and digital sovereignty are the two main drivers behind choosing a sovereign cloud.

For Red Hat, this sovereignty rests on five fundamental pillars:

• The transparency and trust inherent in open-source software. Public code offers complete traceability, reinforced by the upcoming European Cyber Resilience Act, which will impose strict cybersecurity criteria on all open-source solution providers from 2027.
• The openness and hybridity of the cloud enable the portability of workloads between providers, thereby reducing dependence.
• Infrastructure reliability: Faced with hundreds of thousands of open source communities constantly producing patches and new features, Red Hat ensures sorting, testing and offers stable roadmaps ranging from 10 to 12 years for Red Hat Enterprise Linux.
• The ecosystem of certified local partners is essential for the concrete implementation of the solutions.
• Finally, AI itself must be sovereign. This requires the integration of open standards such as vLLM, InstructLab, and the MCP protocol to maximise interoperability.

Rémy Mandon emphasised the total absence of intellectual property in Red Hat code, which is systematically shared with communities, and the Open Source Assurance that protects customers against any legal risk.

Last but not least, an exclusive global announcement. Red Hat unveiled the launch this coming January of “Confirmed Sovereign Support.” This is a 100% European technical support service delivered by staff based in the EU and available to all 27 member states. It specifically targets sovereign sectors such as defence, healthcare, banking, and government ministries.

AWS and Red Hat

Antoine Binétruy and Fabien Schwarz (AWS France) detailed the AWS-Red Hat strategic alliance, which began in 2008. Red Hat Enterprise Linux joined AWS in 2008, just two years after the cloud was created.

Today, the partnership is structured around three major axes:

• ROSA (Red Hat OpenShift Service on AWS)
• OpenShift Virtualisation allows migrated virtual machines and containers to coexist on AWS infrastructure.
• OpenShift AI for artificial intelligence workloads.

A recently signed Strategic Collaboration Agreement (SCA) is mobilising significant resources to accelerate co-innovation and joint go-to-market strategies.

Customer use cases illustrate this depth of integration:

• Emile Frey France has migrated autosphere.fr to ROSA
• SNCF, RATP and Air Liquide are modernising their IBM Maximo software on OpenShift, deployed in the AWS cloud
• Safran has reduced its developers’ deployment time by a factor of ten, thanks to ROSA and Red Hat Dev Spaces, which has gone from two months to a few days.

Access to Red Hat solutions via AWS is available through two channels: directly from the AWS console for deployment in just a few clicks, or via the AWS Marketplace. This offers a dual operational and financial advantage, as purchases count toward AWS customer commitments, allowing them to secure their rebates while accessing the Red Hat portfolio. With the planned December 2025 launch of the AWS European Sovereign Cloud, operated exclusively by EU citizens and completely separate from the existing AWS network, Red Hat is a key partner in this strategic deployment for the European market.

From the Cloud era to the AI era

To conclude the morning, Pauline Truong , AI Solution Architect at Red Hat, presented the major evolution of the enterprise AI offering. Four fundamental pillars structure the offering: inference, personalisation, MLOps, and agentic.

Red Hat AI

Red Hat AI positions itself as a complete solution, from development to production, by fully leveraging hardware accelerators. At the heart of this offering is the vLLM inference server, to which Red Hat is a major contributor. It provides the ideal starting point for serving language models while optimising performance and cost.

OpenShift AI

The vendor also offers a flexible toolkit for customising LLMs. This allows for model enrichment via RAG or fine-tuning with the InstructLab methodology, all supported by a vast catalogue of validated and optimised models from Mistral, Llama, and OpenAI, available as open source on Hugging Face and integrated into OpenShift AI. This platform is natively applicable to OpenShift to facilitate the scaling of AI workloads in hybrid or multi-cloud environments. It facilitates collaboration between IT, Ops, and Data Science teams. As needs evolve, the Red Hat AI portfolio offers industrialisation capabilities:

• Regularly updated development environments
• Pipelines for automating workflows
• Agents to orchestrate complex tasks
• Real-time observability
• Safeguards for safety and compliance
• Performance evaluation tools.

GPUs and Model-as-a-Service

Red Hat is also introducing two new paradigms: GPU-as-a-Service and Model-as-a-Service, as well as the distributed vLLM project announced in May, which will be available next week in OpenShift. The ecosystem is enriched by support for the Model Context Protocol (MCP) to connect models to external tools and data sources, the implementation of LlamaStack offering a single entry point to a wide range of AI capabilities, and data processing tools like Dockling for parsing various document formats and InstructLab for generating synthetic data, thus closing the complete data-to-inference cycle.