🔆 Lumiera's Paris observations

Was this email forwarded to you? Sign up here

🗞️ Issue 78 // ⏱️ Read Time: 5 min

Hello 👋

We’re spending this week in Paris, where we are attending RAISE Summit and hosting an Executive Dinner together with Windsurf, a leading AI code assistant company. Below we are sharing some of our insights based on the conference.

For context: This conference stands out in a couple of ways in contrast to other events we’ve attended, in that there has been a heavy focus on the infrastructural side of AI, and it clearly has a strong commercial purpose, with few other interests being represented. Most participants are business representatives with a clear economic interest in AI infrastructure.

We’re highlighting three things today: The infrastructure discussions, the human element of technological innovation, and some exclusive details from our event.

Our Three Observations from this Week

1. The infrastructure discussions

What does it take to fuel AI? Data centres (physical spaces), the processing units/GPU’s (the hardware chips that these models run on, stored in the data centers) and the foundation models that infrastructure serves.

Based on the conversations we’ve taken part of, it seems like the majority of AI infrastructure leaders are making a strong effort to win the race against each other, focusing on market domination, rather than considering the impact on society and the experience of end users. There is a degree of separation here, between providing the backbone of AI to model companies and the final use case, that allows for this narrow scope of what it means to “win.”

The traditional innovation cycle typically flows from identified need → solution development → infrastructure scaling. In AI, we're seeing this reversed: massive real estate investments in data centers are driving the push for AI applications, rather than genuine user needs driving infrastructure development.

This creates several dynamics:

• Lobby-driven demand: Infrastructure stakeholders with significant capital investments are pushing for expanded AI use cases to justify their capacity

• Misaligned incentives: Success metrics focused on GPU utilization rather than meaningful problem-solving

• Resource allocation distortions: Enormous capital flowing toward infrastructure before sustainable business models are proven

Are we building AI infrastructure to solve real problems, or are we creating problems that justify our infrastructure? How do we ensure that the physical constraints of data centers don't determine the direction of AI development? And perhaps most importantly: Who benefits when infrastructure capacity drives adoption rather than genuine need?

2. Cracking the code on AI adoption

Has anyone cracked the code on AI adoption at scale or the human side of AI innovation? In cases when the topic was brought up, it was clear that getting people bought into the AI journey is more important for success than the technology itself.

Reality Check: Present vs. Future

In the panel "Code, Copy, Conquer: How AI Agents Redefine Enterprise," journalist Karen Kwok from Reuters was doing a great job encouraging a conversation around what is real today versus what is future potential. It's natural for business leaders to promote the latter, but it's important for the public to know what the current limitations of AI are.

There seemed to be a focus on short-term thinking that celebrates immediate gains in velocity and efficiency. What was missing was a plan for what comes next: where humans fit in if/when AI can do it all. This gap between current capabilities and future promises creates unrealistic expectations that can undermine successful adoption.

The People Problem

Few conversations recognized the importance of people and change management in AI implementation. If the people are not on board, the technology is not going to make a difference. The transformation process is not just about training: It's about trust, understanding, and meaningful involvement.

One notable exception was Sanofi's approach, which started with quality: establishing governed and high-quality data infrastructure before adding AI capabilities. This foundation-first strategy was built in anticipation of the EU AI Act, demonstrating how regulatory compliance can actually drive better implementation practices.

So, the question remains: How will we actually manage the transition into the AI era. If we will be working differently, and if some jobs will be replaced, how will we handle the upskilling and transition? Most discussions focused on the technology's capabilities rather than the human impact and adaptation strategies.

3. Executive Dinner on the Topic of Responsible AI

We’ve had many questions about the dinners and events we host in different cities around Europe. This time we created an experience with the CTO of Windsurf in Paris, and made curated invitation list of mostly technical leaders. Out of respect for the privacy for the profiles invited, this dinner is confidential, meaning we won’t share many details. What we can share is that we had a great time, the food was delicious and one of the topics discussed was if AI impacts our cognitive abilities (following the study ‘Your Brain on ChatGPT’ that we wrote about a couple of weeks ago).

Are you a company leader or an investor in the tech/AI space that would like to host a similar event with us? Feel free to get in touch by sending an email to [email protected].

Until next time.
On behalf of Team Lumiera

Emma - Business Strategist
Allegra - Data Specialist

Lumiera has gathered the brightest people from the technology and policy sectors to give you top-quality advice so you can navigate the new AI Era.

Follow the carefully curated Lumiera podcast playlist to stay informed and challenged on all things AI.