The RAISE Summit 2025, held in Paris, showcased the latest in AI, attracting 6,500 professionals from around the globe. Key highlights included:
- Agentic AI: Autonomous systems like H Company's "SurferH" are transforming workflows and decision-making. Hackathon winners demonstrated tools for hiring, onboarding, and IT management.
- Sovereign AI: Nations are focusing on local data control, with Europe investing €200 billion in AI infrastructure. Companies like Cerebras Systems lead with cost-effective, high-performance models.
- Open-Source AI: Models like Llama 3.1 and Mistral 7B are competing with proprietary systems, offering businesses affordable, scalable solutions.
- AI Infrastructure: Innovations in compute power and storage, such as Vultr's cost-efficient infrastructure and exascale storage solutions, are driving scalability.
- Industry Applications: AI is revolutionizing sectors like cybersecurity, life sciences, and customer support, automating workflows and improving efficiency.
The event emphasized Europe's growing role in AI innovation, with a focus on collaboration, compliance, and scalable solutions.
RAISE Summit 2025 Key Statistics and AI Trends Overview
Agentic AI and Workflow Automation
RAISE Hackathon Winners and Their AI Solutions
The RAISE 2025 hackathon brought together 6,246 participants [1], all focused on crafting solutions to tackle pressing business challenges. The standout projects showcased how agentic AI goes beyond basic automation, creating systems capable of independent decision-making and action.
The first-place winner, Unmask, revolutionises hiring workflows. This platform verifies resumes, confirms candidate identities, and analyses interview data in real time. By removing manual verification bottlenecks, it enables HR teams to make quicker and more reliable hiring decisions. Onboard-Me, which secured second place, transforms static company documents into interactive onboarding experiences tailored to each employee's learning needs. Axiom Prime takes IT operations to the next level by autonomously monitoring infrastructure and diagnosing system issues, eliminating the need for constant human oversight. Meanwhile, Ringy enhances customer support with real-time transcription and insights during live calls, powered by Groq AI on Vultr's infrastructure. (Details sourced from official hackathon records [4])
These projects highlight how agentic AI is reshaping workflow automation, making it smarter and more adaptive.
How Agentic AI Changes Business Workflows
Inspired by these cutting-edge hackathon innovations, businesses are rethinking their approach to automation. Instead of just digitising workflows, agentic AI tackles fundamental problems like repetitive approvals, fragmented data, and manual task handoffs.
Experts predict that by 2028, at least 15% of workplace decisions will be automated [7]. Additionally, 93% of IT and business leaders expect AI to enable more proactive and personalised services [6]. The rise of no-code platforms, such as Make, Zapier, and n8n, empowers non-technical teams to design AI agents for specialised tasks like real-time ad testing or CRM lead analysis. This accessibility is driving the creation of hybrid teams, where AI agents act as collaborative partners rather than mere tools.
"The organisations that succeed won't necessarily be those that automate the most. They'll be the ones that integrate AI as a trustworthy and capable teammate." – Nadine Soyez, AI Consultant [7]
The ability to collaborate effectively with AI agents is expected to increase human focus on high-value tasks by 65% [7], fundamentally altering the way work is done across various industries.
sbb-itb-e314c3b
AI Infrastructure and Sovereign Computing
Enterprise Infrastructure Developments
Creating scalable AI systems isn't just about having cutting-edge processors. The real game-changer lies in the infrastructure that supports these systems, determining whether businesses can afford to deploy AI solutions at scale. At the RAISE Summit 2025, two essential elements stood out: compute power and storage capacity.
Vultr highlighted how composable infrastructure could slash costs by introducing a developer-friendly approach. Their pre-built Terraform templates help businesses quickly transition from proof-of-concept to full-scale production. Kevin Cochrane, Vultr's Chief Marketing Officer, pointed out that their AI-focused infrastructure could cut cloud expenses by as much as 90% [4]. With 32 cloud data center regions worldwide, including a major hub in Paris, Vultr is uniquely positioned to address both performance demands and the strict regulatory requirements of European enterprises [4].
Storage, often overlooked, plays a pivotal role in AI pipelines. George Kurian, CEO of NetApp, emphasized this point:
Storage is the unsung hero in AI pipelines and... data gravity is driving Kubernetes innovation [8].
As AI models grow in size and training datasets expand, the ability to manage and move data efficiently becomes just as important as computing power. Companies like NetApp and DDN are stepping up with exascale storage solutions designed to eliminate data bottlenecks and ensure smooth operations.
The industry is also shifting gears, moving from "stage zero" - focused on training foundational models - to "stage one", where enterprises begin deploying open-source large language models (LLMs) for production use. Kevin Cochrane explained this evolution:
We're just moving to stage one, where enterprises will start deploying for production inference, open-source LLMs, to start adding efficiencies to all of their enterprise workflows [8].
This progression shifts the focus of infrastructure toward enabling distributed deployment, reflecting the changing needs of businesses as they adopt AI at scale.
As enterprise requirements grow more complex, factors like regional regulations and compliance are reshaping the priorities for AI infrastructure.
Sovereign AI Development Strategies
Sovereignty in AI isn't just about politics; it's about control, compliance, and fostering regional economic growth. At the RAISE Summit 2025, sovereign AI was described as:
the new firewall in a borderless tech world [3].
European nations are leading the charge, emphasizing infrastructure that keeps data within their borders and aligns with local regulations.
The European Union has committed €200 billion to AI development, with €20 billion specifically allocated to building AI gigafactories to boost regional research and innovation [5]. These investments are designed to complement advancements in infrastructure, ensuring AI deployments remain scalable and compliant. France, for instance, is accelerating the expansion of data centers powered by nuclear energy, aiming to establish itself as a carbon-neutral, sovereign AI hub [5].
Karl Havard of Nscale outlined the three key pillars of sovereign AI:
Control, Compliance, and Growth [3].
This means organizations must carefully manage where their data is stored and processed, adhere to stringent regulations like GDPR, and develop local capabilities to reduce dependence on foreign tech providers. This approach is gaining traction globally, as evidenced by the "Joint Declaration on Inclusive and Sustainable AI", signed by 58 countries - including India, China, and several EU members - during the 2025 Paris Summit [9].
The practical impact of these strategies is immense. With stricter data residency laws across Europe, sovereign cloud infrastructure is no longer optional - it’s becoming a necessity. Kevin Cochrane underscored this point:
Sovereign cloud infrastructure is becoming more relevant than ever as governments and enterprises face stricter compliance and data residency mandates [4].
Forward-thinking organizations are leveraging this opportunity to build AI systems that are not only locally compliant but also capable of scaling globally while respecting regional requirements.
Open-Source AI and Inference Optimization
Growth of Open-Source AI Models
Open-source AI has reached a level where it can now compete directly with proprietary systems. By 2025, models such as Llama 3.1 (boasting 405 billion parameters) and Falcon 180B achieved what experts describe as "frontier-level" performance, standing toe-to-toe with proprietary giants like GPT-4 and Google's PaLM 2-Large [10]. The shift is staggering: Meta's open-source models alone have been downloaded over 400 million times, with the download rate in 2025 being ten times higher than just a year earlier [10].
What’s driving this momentum is the rise of smaller, highly efficient models. For instance, Mistral 7B, with just 7.3 billion parameters, outperformed Meta's Llama 2 13B across all benchmarks while being faster and more cost-efficient [10]. This breakthrough means businesses no longer need expensive cloud setups to access powerful AI. Instead, they can run these models on consumer-grade hardware, like a single NVIDIA RTX 4090 or even a Mac with 32 GB of RAM [10].
Mark Zuckerberg highlighted this progress, stating:
This year, Llama 3 is competitive with the most advanced models and leading in some areas [10].
The impact on businesses has been clear: 51% of companies using open-source AI tools reported positive ROI, compared to 41% of those solely relying on proprietary solutions [11]. Platforms like Hugging Face, which now hosts more than 100,000 models, have made it easier than ever for developers to access and innovate [10]. Within just one week of Llama 3's release, developers created over 600 derivative models, showcasing the speed of innovation that open-source fosters [10]. This focus on efficiency and accessibility is paving the way for advancements in inference performance.
Improving AI Inference Performance and Cost
With open-source AI models reaching new heights, attention has shifted to optimizing inference - making AI faster and cheaper to deploy. At the RAISE Summit 2025, experts stressed that improving inference speed and cost is essential for scaling AI applications, particularly for the "Agentic Revolution." This next phase of AI demands systems capable of processing and acting on data in real time, rather than just generating text [3][12].
Hardware innovations are transforming the economics of inference. Companies like Cerebras Systems have introduced wafer-scale engines for ultra-fast processing, while Groq and FuriosaAI have developed chips tailored for large-scale data centers [3]. In July 2025, Cerebras partnered with GSK on a project named "Molecules at Wafer-Scale", leveraging this technology to accelerate drug discovery and life sciences research [3]. Meanwhile, Scaleway became the first European cloud provider to offer the NVIDIA Blackwell Ultra B300 GPUs in December 2025, delivering a 50% performance boost compared to the prior B200 model [12].
The cost savings from these advancements are tangible. Vultr’s AI-native infrastructure, for example, powered the hackathon-winning project Ringy in July 2025. This real-time customer support tool demonstrated how ultra-low latency processing can operate efficiently on globally distributed systems [4]. As Aude Durand, Deputy CEO of iliad Group and Scaleway President, summed it up:
AI is getting smarter, faster, and everywhere at once [12].
On the software side, companies like Together AI and Fireworks AI are tackling inference-specific challenges by optimizing software to bypass hardware limitations. These improvements make it possible for organizations outside of the tech giants to deploy cutting-edge models at scale [3]. The combination of open-source innovation and optimized inference is reshaping the AI landscape, making advanced deployments more accessible than ever before.
RAISE Summit 2025 | The AI Evolution: Open Source, Fast Inference, and the Agentic Revolution
AI Applications Across Industries
These examples highlight how advancements in AI are reshaping the way businesses operate across various sectors.
AI in Cybersecurity, Data Governance, and Development Tools
AI-driven systems are revolutionizing core industries by automating and enhancing critical processes. At the RAISE Summit 2025, leaders shared innovations in AI that can manage entire workflows seamlessly.
In cybersecurity, the focus has shifted to "Offensive Security", where AI is used to simulate potential attacks before they happen. Nikesh Arora, CEO of Palo Alto Networks, discussed their work on automated threat detection systems designed to proactively identify vulnerabilities. He also introduced the concept of Sovereign AI as a "new firewall" for enterprises in highly regulated environments [3].
Paul Bloch, President and Co-Founder of DDN, emphasized the importance of data governance, stating, "AI is the Pivot - Data is the Power." This underscores how AI can transform unstructured data into actionable insights. A notable example came in July 2025, when Sanofi partnered with Snowflake to develop an AI-driven framework capable of transforming vast amounts of research data into valuable tools for drug discovery, spanning from "molecule to market value" [3].
The development tools sector is also undergoing a transformation. Lovable, a startup specializing in agentic coding, achieved €70.5 million in Annual Recurring Revenue within just seven months by July 2025 [3]. CEO Anton Osika demonstrated how their technology allows AI to autonomously handle coding tasks rather than simply suggesting snippets. Similarly, NinjaTech AI and Cerebras introduced "Truly Autonomous AI Agents" in July 2025, leveraging Cerebras' rapid inference technology to execute complex, multi-step workflows without human input [3].
| Sector | Key Organisation(s) | Primary AI Application |
|---|---|---|
| Cybersecurity | Palo Alto Networks, Snyk, XBOW | Offensive security and AI-driven threat defense |
| Data Governance | Snowflake, Databricks, Snorkel AI | Structuring unstructured data for AI workflows |
| Development Tools | Turing, Lovable, Poolside AI | Autonomous coding and workflow execution |
| Customer Support | Zendesk, Parloa, Cresta | Automating 90% of service interactions |
| Life Sciences | GSK, Sanofi, Owkin | AI-powered drug discovery and pharma innovation |
These innovations are paving the way for new service delivery models that rely heavily on AI-driven automation.
Emerging Trends in Agentic AI
A major theme emerging across industries is the transition from traditional software to autonomous AI agents. Des Traynor captured this shift with the phrase, "The Death of SaaS, The Dawn of Agents" [3]. This reflects a growing trend where businesses are moving away from tools that merely assist humans to systems that can independently complete entire workflows.
By 2028, it’s expected that 68% of service and support interactions with tech providers will be handled by Agentic AI [6]. Gartner projects that by 2029, 80% of routine customer service requests will be managed by these systems, potentially cutting costs by around 30% [6]. Parloa, which became a European AI unicorn by 2025, exemplifies this shift. Their agentic AI solutions go beyond basic chatbots, enabling fully autonomous agents to resolve customer issues effectively [3].
In sectors like cybersecurity and data governance, 93% of IT and business leaders anticipate that Agentic AI will deliver more personalized and proactive services across the board [6]. However, 89% of customers still prefer a mix of AI and human support, emphasizing the need for a balanced approach. Businesses are likely to start by automating low-risk tasks, such as internal support or routine customer queries, before expanding to more critical operations where human judgment remains crucial [6].
Key Takeaways from RAISE Summit 2025
RAISE Summit 2025 highlighted Europe’s strong position in the AI sector, gathering leading professionals at the Carrousel du Louvre in Paris for meaningful B2B interactions. The event also celebrated advancements in AI by awarding a €5 million prize pool. As co-founders Michael Amar, Henri Delahaye, and Hadrien de Cournon explained:
For our second edition, RAISE Summit has again shown that it is the premier event for B2B professionals seeking to disrupt, build, and connect in the AI landscape... to find the right partners to deliver cutting-edge solutions. [1]
The summit wasn’t just about networking. It illustrated how partnerships across industries are reshaping AI, featuring insights from Sanofi alongside tech giants like Google Cloud, AWS, Nvidia, and Mistral AI.
Hackathon Projects and Networking Opportunities
The "RAISE Your Hack" event brought the spirit of innovation to life, becoming the largest AI hackathon globally with 6,246 participants [1]. Projects showcased at the hackathon demonstrated practical solutions for real-world challenges. For instance, some focused on automating hiring verification processes, while others introduced AI-powered surgical guidance systems that reduced both operating times and radiation exposure. These projects underscored how AI can enhance efficiency across various sectors.
Global AI Sovereignty and Cross-Industry Collaboration
One of the summit’s key themes was "Algorithmic Independence", emphasizing the importance of Sovereign AI in securing national infrastructure. Discussions revolved around achieving three main objectives: control, compliance, and growth [3]. Europe’s focus on developing sovereign GPU capacities and localized cloud infrastructure reflects its dedication to maintaining data residency standards while staying ahead in AI advancements.
The 4F Compass (Foundation, Frontier, Friction, Future) provided a framework for understanding AI’s transformative potential across industries. This approach encouraged participants to explore innovative applications while ensuring responsible implementation [2]. Mike Mattacola, GM International at Coreweave, summed up the event’s energy:
Amazing people, great vibe, and it is something Europe needs badly. We need everyone to get together in Europe more often and focus on how we can accelerate AI. [2]
FAQs
What is agentic AI, in plain terms?
Agentic AI represents a new level of artificial intelligence, where systems operate independently, make decisions on their own, and manage tasks without needing continuous human oversight. Unlike older AI models that simply respond to commands, agentic AI can take the initiative, analyze data, and autonomously handle intricate workflows. This development allows businesses to simplify operations, automate tasks, and improve decision-making processes, leading to a significant change in how AI is used across industries and reshapes workplace dynamics.
How can my organisation adopt sovereign AI in Europe without breaking compliance rules?
To embrace sovereign AI in Europe while adhering to regulations, it's essential to rely on AI systems developed using local infrastructure, data, and expertise. This approach ensures greater control and aligns with EU standards, such as the AI Act, by emphasizing transparency, safety, and ethical practices. Focus on creating AI solutions that are secure, adaptable, and designed to address local requirements. Partnering with trusted regional providers can help strike a balance between advancing technology and staying compliant with legal and ethical guidelines.
Which open-source models are most practical for production, and what hardware do they need?
In 2025, open-source models suitable for production will likely be those backed by active community involvement and offering scalable deployment capabilities. To handle the demands of these models, high-performance GPUs like the NVIDIA A100 or dedicated inference servers are often necessary. Additionally, optimized inference engines play a crucial role in ensuring efficiency for large-scale operations. Examples of solutions tailored to meet these needs include vLLM and Ollama.
