Dans Les Bonnes Choses du Numérique, on parle d’infrastructure, d’Internet, de cloud et de souveraineté… mais pas dans un studio. On en parle dans une cuisine.

Pour ce premier épisode, Quentin Adam reçoit Benjamin Bayart, pionnier français d’Internet, ancien président de FDN (French Data Network) et figure historique du logiciel libre.

Pendant que le bœuf bourguignon continue de mijoter et que les pommes de terre sont épluchées, la conversation quitte les débuts d’Internet pour s’attaquer à un autre sujet devenu central dans l’histoire du réseau : la neutralité du net. Car si Internet est souvent présenté comme un espace ouvert par nature, cette ouverture a en réalité fait l’objet de batailles politiques très concrètes, notamment en Europe.

Dans cette deuxième partie, la conversation dérive naturellement autour de :

• ce que signifie réellement la neutralité du net ;

• pourquoi Internet est avant tout une interconnexion de réseaux ;

• comment fonctionnent les relations entre fournisseurs d’accès et transitaires ;

• la naissance de La Quadrature du Net ;

• les débats politiques autour de la régulation d’Internet en Europe ;

• et comment un texte européen a fini par protéger le principe d’un Internet ouvert.

• 
  

Bref : comment l’architecture d’Internet est devenue un sujet politique majeur.

Les Bonnes Choses du Numérique est une série de conversations informelles animée par Quentin Adam autour d’un principe simple : prendre le temps de discuter des sujets qui comptent vraiment dans le numérique : Internet, cloud, open source, infrastructures, souveraineté technologique. Le tout autour d’un repas.

—

LinkedIn : https://www.linkedin.com/in/waxzce/

Bluesky : https://bsky.app/profile/waxzce.org

X : https://x.com/waxzce

TikTok : https://www.tiktok.com/@waxzce

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.

Dans Les Bonnes Choses du Numérique, on parle d’infrastructure, d’Internet, de cloud et de souveraineté… mais pas dans un studio. On en parle dans une cuisine.

Pour ce premier épisode, Quentin Adam reçoit Benjamin Bayart, pionnier français d’Internet, ancien président de FDN (French Data Network) et figure historique du logiciel libre.

Pendant qu’ils préparent un bœuf bourguignon — avec la recette de Benjamin Bayart lui-même —, la discussion remonte aux débuts d’Internet en France : les premiers modems, les nuits passées à comprendre comment fonctionnent les réseaux, et l’apparition des premiers fournisseurs d’accès.

C’est aussi l’histoire d’une autre manière de construire Internet : FDN, un fournisseur d’accès associatif, né pour défendre un réseau ouvert et maîtrisé par ses utilisateurs.

Dans cette première partie, la conversation dérive naturellement autour de :

• les premières découvertes de l’informatique dans les années 80 ;

• l’arrivée d’Internet en France et les premiers modems ;

• la naissance des fournisseurs d’accès ;

• pourquoi certains ont choisi un modèle associatif plutôt que commercial ;

• et comment ces choix techniques sont devenus… des choix politiques.

• 
  

Bref : comment Internet s’est construit, avant de devenir l’infrastructure critique qu’il est aujourd’hui.

Les Bonnes Choses du Numérique est une série de conversations informelles animée par Quentin Adam autour d’un principe simple : prendre le temps de discuter des sujets qui comptent vraiment dans le numérique : Internet, cloud, open source, infrastructures, souveraineté technologique. Le tout autour d’un repas.

—

LinkedIn : https://www.linkedin.com/in/waxzce/

Bluesky : https://bsky.app/profile/waxzce.org

X : https://x.com/waxzce

TikTok : https://www.tiktok.com/@waxzce

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.

Artificial Intelligence is no longer just a software story. It is a compute story. In this full episode of Opening Voices, Quentin Adam speaks with Steeve Morin, founder and CEO of ZML, to explore a fundamental question: who controls the compute layer of AI?

Together, they unpack:

• Why AI makes cloud systems compute-bound again

• The real difference between training and inference

• Why inference will dominate AI workloads

• How stateful systems break 20 years of architecture patterns

• Why power, not space, now limits data centers

• Whether GPUs are a temporary solution

• The rise of TPUs, NPUs and AI-dedicated chips

• Why hardware optionality may define the next decade

As AI becomes a universal primitive across industries, control shifts from models to infrastructure.

This episode connects architecture, economics and semiconductor strategy, and explains why inference may become the industrial foundation of the AI era.

Opening Voices is also available on all streaming platforms:

• Deezer: https://www.deezer.com/show/1001774171

• Spotify: https://open.spotify.com/show/3QTe4gKhsmhWnlLZaUxNo1

• Apple Podcasts: https://podcasts.apple.com/fr/podcast/opening-voices/id1806281823

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.

What happens when AI infrastructure depends on a single compute ecosystem?

In this final part of Opening Voices, Quentin Adam concludes his discussion with Steeve Morin, founder and CEO of ZML, to explore how to bring competition back into AI compute. ZML’s approach is simple in principle, difficult in execution: make AI workloads run efficiently on any chip.

They discuss:

• why hardware abstraction is key to breaking vendor dependency

• how optionality of compute changes market dynamics

• why existing hardware can still deliver major efficiency gains

• how operating complexity locks companies into single ecosystems

• why open source can accelerate semiconductor competition

The future of AI will not be decided by models alone, but by who controls the compute layer beneath them.

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.

If AI is to power the entire economy, inference must become affordable, scalable and widely available.

In this third part of Opening Voices, Quentin Adam continues the conversation with Steeve Morin, founder and CEO of ZML, to explore what it really takes to industrialise inference. They discuss:

• why AI must move from “chatbots as products” to AI as an infrastructure primitive

• why inference will power every sector — banks, startups, industry

• how efficiency gains (sometimes 5x, 10x, even 100x+) are still possible

• why GPUs are not the only path forward

• how new chips (TPUs, NPUs and emerging players) are reopening the semiconductor market

• why power, density and optimisation now matter more than raw experimentation

This episode explains why the next wave is not about building better models, but about making inference economically viable at scale.

—

Episode Chapters: Making Inference Available

00:00 – Introduction and Context

01:38 – AI as a Primitive vs. AI as a Product

04:19 – The Economic Unit of the Token

05:15 – Scaling Compute for Inference

07:31 – A Revolution Comparable to Mobile

08:43 – Beyond GPUs

10:56 – Compiler Errors and Efficiency Waste

12:38 – Understanding Chips

15:31 – The New "Blue Ocean" of Semiconductors

20:40 – Nvidia's Strategy and Competition

21:43 – Conclusion and Next Episode

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.

What really separates training a model from running it in production? And why does this difference matter so much at scale?

In this second part of Opening Voices, Quentin Adam, CEO of Clever Cloud, continues the discussion with Steeve Morin, founder and CEO of ZML, to clarify a distinction that is often misunderstood.

Starting from concrete explanations, Steeve walks through what training and inference actually involve from an engineering and industrial perspective:

• why training is a one-off research effort, while inference is repeated endlessly

• why “more is better” applies to training, but becomes a liability in production

• why inference represents the vast majority of compute needs

• how cost, reliability and margins reshape technical decisions

• why operating systems at scale is fundamentally different from building models

This episode helps understand why the real challenge is no longer creating models, but running them reliably, efficiently and sustainably in production.

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.

Why does AI suddenly make cloud infrastructure harder to operate, scale and predict? For years, cloud systems were designed around a simple reality: most applications were limited by network and storage, not by computation.

In this first part of Opening Voices, Quentin Adam, CEO of Clever Cloud, talks with Steeve Morin, founder and CEO of ZML, about why this reality no longer holds — and what breaks when it does.

Drawing from real production experience, Steeve explains:

• why some workloads push hardware to its limits instead of waiting on I/O

• how sequential computation changes everything for scaling and scheduling

• why memory becomes the real bottleneck long before CPU usage

• why moving workloads around is no longer trivial

• and why many cloud design patterns simply stop working

This episode is about understanding why systems that worked for years start failing under new constraints, and what that means for engineers and organisations operating modern platforms.

Opening Voices is also available on all streaming platforms:

• Deezer: https://www.deezer.com/show/1001774171

• Spotify: https://open.spotify.com/show/3QTe4gKhsmhWnlLZaUxNo1

• Apple Podcasts: https://podcasts.apple.com/fr/podcast/opening-voices/id1806281823

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.

How do you build Europe's first sovereign high-performance processor from scratch? What does it take to challenge Intel and AMD with just 200 engineers and €150 million?

In this comprehensive episode of Opening Voices, I sit down with Philippe Notton (CEO) and Craig Prunty (VP) from SiPearl to explore Europe's most ambitious semiconductor project:

🔹 How 200 engineers across 5 R&D sites design a 60 billion gate processor

🔹 Why ARM is becoming the new battlefield for server processors

🔹 The €20 million data center built to emulate Europe's most complex chip

🔹 How hyperscalers like Amazon and Google are driving the custom silicon revolution

🔹 Why AI inference on CPUs can be 70x more energy efficient than GPUs

🔹 The chiplet revolution and Europe's opportunity in post-Moore's Law era

We dive deep into the technical challenges, market dynamics, and geopolitical implications of building sovereign computing infrastructure in Europe.

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.

Why are CPUs making a comeback in the age of AI? And how can Europe build a sovereign compute infrastructure in a world dominated by American and Asian tech giants?

In this final episode of Opening Voices with the SiPearl team, I sit down with Philippe Notton (CEO) and Craig Prunty (VP) to explore the future of processor design and its impact on AI, sovereignty, and Europe’s industrial strength.

We discuss:

🔹 Why CPUs can outperform GPUs in inference workloads

🔹 How SiPearl’s chip design drastically reduces energy consumption

🔹 The growing importance of chiplet architectures for modular compute

🔹 Why Europe needs its own electronic design cloud infrastructure

🔹 How to reclaim industrial skills and build a sovereign value chain

🔹 The difference between SiPearl and ASML in the semiconductor ecosystem

🔹 Why sovereignty isn't about isolation—but about control and resilience

What role should Europe play in the next wave of AI infrastructure? Join the conversation in the comments.

Subscribe for more deep dives into Europe’s tech future.

🎧 Opening Voices is also available on all streaming platforms:

— Deezer: https://www.deezer.com/show/1001774171

— Spotify: https://open.spotify.com/show/3QTe4gKhsmhWnlLZaUxNo1

— Apple Podcasts: https://podcasts.apple.com/fr/podcast/opening-voices/id1806281823

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.

Why are tech giants like Apple, Amazon, and Google building their own processors—and what does it mean for Europe?

In this new episode of Opening Voices, I sit down with Craig Prunty, VP at SiPearl, to explore how processor design is becoming a strategic lever for performance, sovereignty, and efficiency.

We discuss:

🔹 Why ARM is taking over the server market

🔹 The real reasons hyperscalers are going boutique in chip design

🔹 What makes European players like SiPearl competitive—despite limited funding

🔹 The opportunities for Europe to scale its processor ambitions with partners like TSMC

Join the conversation—can Europe reclaim control over its compute infrastructure?

Subscribe for more deep dives into the future of European tech.

🎧 Opening Voices is also available on all streaming platforms:

— Deezer: https://www.deezer.com/show/1001774171

— Spotify: https://open.spotify.com/show/3QTe4gKhsmhWnlLZaUxNo1

— Apple Podcasts: https://podcasts.apple.com/fr/podcast/opening-voices/id1806281823

Hébergé par Ausha. Visitez ausha.co/politique-de-confidentialite pour plus d'informations.