1. Europe’s pivot: less regulation, more implementation
The EU’s message, voiced by Executive Vice-President Henna Virkkunen, was surprisingly pragmatic. Rather than adding new layers to the AI Act, she emphasised simplifying what already exists. A “digital simplification” package, a digital fitness check and efforts to cut duplicate cybersecurity and data reporting are all meant to make life easier, especially for startups and SMEs. Read more in the Digital Rulebook.
To reduce fragmentation, the Commission floated a “28th regime”: a voluntary, harmonised track that would let innovative firms operate across the bloc under one regulatory and tax framework. Alongside this, a Scale-Up Fund and new retail investment instruments aim to unlock household savings for startup financing.
Compute is now explicit industrial policy. The Commission highlighted AI Factories — an initiative that will expand European supercomputing capacity roughly fivefold and give researchers and startups priority access. Learn more about the initiative in the AI Factories policy page.
At RankMyAI we will reflect these shifts by giving more weight to fast, well-documented EU compliance, the ability to operate across borders and clear access to reliable compute, whether via AI Factories or equivalent cloud capacity.
2. Partnerships, not vertical empires
Many companies no longer build full compute infrastructure or frontier models themselves. Instead they partner with hardware and infrastructure providers. For example, Uber recently struck a deal with NVIDIA: Uber will leverage NVIDIA’s DRIVE AGX Hyperion 10 platform and AI infrastructure to roll out global robotaxi and autonomous-delivery fleets. See the announcement here: Uber–NVIDIA Autonomous Mobility Partnership.
Similarly, Mistral AI partnered with ASML: ASML invested €1.3 billion and now integrates Mistral’s AI models into its semiconductor-machinery operations. Read the announcement here: ASML–Mistral Strategic Partnership.
These partnerships show that the competitive edge now often comes through access to high-end compute and specialised hardware — not just owning the full stack internally.
3. Beyond GPUs: hardware as bottleneck
Another undercurrent at Web Summit was that software innovation is starting to bump into physical limits. Training and serving large models is increasingly constrained by energy, cooling and chip availability. That is why so much attention went to the next wave of hardware: more efficient GPUs, photonic chips that move computation into the optical domain, and early forays into quantum for specialised workloads.
For tool builders, the advantage will go to those who can exploit specialised hardware while hiding its complexity from end-users and offering predictable performance–cost trade-offs. RankMyAI will increasingly look at how tools use and orchestrate hardware rather than just the headline model they advertise.
Read more on this subject here: Beyond GPUs: the new compute stack powering AI’s next decade.
4. Visual intelligence steps forward
While text and tabular data still dominate many AI narratives, a growing share of real-world use cases now revolve around visual intelligence: systems that understand images, video, spatial layouts and sensor data, often combined with structured information. We see this in several more practical applications in categories like Manufacturing & Industry, Transportation & Mobility, and Logistics & Supply Chain.
The most interesting products were not “pure computer vision”, but multi-modal: they mix visual input with metadata, logs and text to make decisions. In response, RankMyAI will expand and refine our rankings around visual intelligence and multi-modal platforms, paying attention to how they handle edge cases, explain decisions and obtain or label their data.
5. Countries as platforms, not just backdrops
One striking observation from the show floor: the majority of space was taken by countries, ministries and tech hubs, not by big AI vendors. Governments from the Middle East, Latin America, Eastern Europe and elsewhere were pitching their ecosystems as the place to build and scale AI companies, promising talent pipelines, testbeds, sandboxes and funding.
For founders this creates a new decision layer: choosing the right jurisdiction becomes a strategic product decision. For policy makers it raises the question of how competitive their offer really is. This relates directly to our reports and insights like the Dutch AI report: Mapping the Dutch AI Landscape.
RankMyAI is therefore expanding its country and ecosystem profiles. We will increasingly focus on indicators such as local infrastructure, funding intensity and regulatory openness — linking tool performance to the quality of the surrounding ecosystem and helping both governments benchmark themselves and companies decide where to base operations.
6. AI for Good: from slogan to metric
Talks such as Huawei’s “AI for Good” keynote cut through the noise. Behind the trillion-dollar race there is a basic question: what does AI actually do for people and the planet? We see rising demand, especially from governments and large enterprises, for tools that can show tangible benefits in health, climate, safety or inclusion, and that are transparent about data sources, fairness efforts and energy use.
We are exploring an “AI for Good” lens across our rankings: surfacing tools with verifiable social or environmental impact, and making it easier for buyers to filter for solutions that do more than optimise ad clicks. See for example the Sustainability Analysis & Reporting Ranking.
