menu
In 2025 Q1, GenAI was defined by hype and rapid experimentation. In Q2, the focus shifted to strategic positioning as both startups and incumbents raced to stake out territory. Q3 brought a more grounded reality. Compute and AI infrastructure emerged as the foundation on which everything else depends, while enterprise adoption shifted from chat interfaces to embedded systems. Agentic AI is now starting to show what it can do to optimize workflows and business processes.
According to a recent survey from Cloudera, 70% of enterprise IT leaders now say they have achieved significant success with AI initiatives. However, Integrating AI is not without security concerns: Half of respondents say data leakage during model training was a concern related to AI security, with 48% saying unauthorized data access, and 43% saying unsecure third-party AI tools.
As GenAI moves into business workflows, demand for reliable and controllable infrastructure is exploding. The new wave of startups is building around the biggest bottlenecks in the stack. Some are working on compute, creating alternatives to scarce GPU capacity. Others are tackling data pipelines, helping enterprises connect their ERP, CRM and internal systems. A third group is focused on orchestration, embedding agents directly into workflows. These shifts are redefining where value in the AI stack will accrue, with consequences for everyone from chipmakers to SaaS founders.
One of the biggest signals in Q3 came from the EU’s push to build large-scale AI training hubs known as AI Gigafactories. The Commission confirmed 76 proposals across 16 countries after its call for interest closed in June, and in September signaled plans to consolidate these into a handful of hubs backed by a €20 billion envelope for regional compute infrastructure.
Meanwhile, US capacity ramped up again. OpenAI and NVIDIA signed a letter of intent to deploy at least 10 GW of NVIDIA systems, with NVIDIA intending to invest up to $100 billion as capacity comes online. Days later, OpenAI expanded its contract with CoreWeave, a US-based cloud provider specialising in GPU infrastructure for AI workloads, by another $6.5 billion. An investment that brings OpenAI’s total CoreWeave commitments in 2025 to around $22.4 billion.
The cloud companies (AWS, Azure, GCP, Oracle Cloud, etc.) are taking center stage as important AI infrastructure providers offering compute, storage, and increasingly, AI-specific accelerators.
Another theme also stood out in Q3. Agentic AI, systems that can reason, plan and execute tasks with minimal input, is moving from prototype to production.
The strongest signal came from startups targeting enterprise systems. Most recently, Factory raised a $50 million Series B to launch Droids, a platform that lets companies create and run autonomous agents for internal workflows. Another example is AppZen that just secured a $180 million Series D to expand its AI finance platform, which automates auditing and compliance directly inside ERP systems.
These systems go deep into enterprise workflows. By integrating directly with ERP, CRM and internal APIs they create complexity, compliance requirements and high switching costs. That’s what makes them defensible, and it aligns with Cloudera’s finding that 83% of enterprise leaders see investment in AI agents as critical to maintaining a competitive edge.
For B2B founders, the real opportunity lies in building intelligence into the systems where business actually happens.
In September, enterprise knowledge platform Sana signed a definitive agreement to be acquired by Workday for about $1.1 billion. One of the largest Nordic AI outcomes to date.
Sana’s success was not about novelty. It was about embedding AI where knowledge actually lives, across documentation, conversations and people, and doing so in a way that fit seamlessly into existing workflows. For Nordic founders it’s proof that billion-dollar outcomes are being built quietly in the systems that matter most.
Another Nordic signal came from Denmark where AI-native finance platform Light just raised $30 million in Series A funding. Built with AI at the core rather than retrofitted, Light is rethinking enterprise finance systems for companies that scale faster than traditional tools can handle.
The clearest macro signal in Q3 came from capital markets. According to Crunchbase, $103.5 billion, or 93%, of all scaleup tech investment in Silicon Valley this year has gone into AI. An unprecedented level of concentration that shows AI is no longer a vertical. It’s the market.
The consequences ripple across the ecosystem. Startups outside AI are finding capital scarcer while AI companies are raising mega-rounds on the back of infrastructure partnerships and talent wars. For founders, this means valuations are increasingly benchmarked against AI-native peers and investor expectations are shifting toward clear AI depth rather than surface features.
US players dominate foundational models, cloud infrastructure, and the biggest scale rounds. Faced with brute force, here’s no magic bullet for how Europe can compete (or at least differentiate) versus US incumbents. However, it seems like there is still space to win by embedding AI where workflows, compliance and integration matter most. That’s where defensibility lives and where Europe’s strengths can count.
Read also
Q2 Recap: Navigating the Shifting Terrain of GenAI
GenAI Landscape (Q1 2025): Europe's Position in the AI Race – Catching up or Falling Behind?
Generative AI Landscape Q4 2024: AI Investments Reach Historic High
