Piotr Krzepczak

Europe’s Technology Independence Strategy Achieving Technological Independence is the "Grand Challenge" for Europe. To decouple from the US "Hyperscalers" (Google, Microsoft, AWS) and hardware gatekeepers (Nvidia), Europe is shifting from a purely regulatory "watchdog" to an active "builder" using a mix of sovereign infrastructure and open-source pragmatism. Here is the blueprint for how Europe is navigating this path as of 2026. Breaking the Nvidia Monopoly (Hardware) Hardware is Europe’s biggest vulnerability. Currently, nearly all AI training happens on Nvidia GPUs. Europe’s strategy focuses on: The European Chips Act: Aiming to double the EU's share of global semiconductor production to 20% by 2030. This includes funding for "Open-Source Hardware" based on RISC-V architecture, which allows European firms to design chips without licensing US IP (like ARM or x86). EuroHPC & AI Factories: Instead of relying on Google’s TPUs or Nvidia’s proprietary clusters, the EU has launched 19 "AI Factories". These are supercomputing hubs (like LUMI in Finland or Leonardo in Italy) that provide startups and researchers with massive sovereign compute power. SiPearl: A key European venture developing "Rhea" microprocessors—high-performance, low-power chips specifically designed for European supercomputers to run AI workloads without US-made CPUs. Replacing the "Alphabet" Stack (Software & Cloud) To move away from Google (Alphabet) and the US cloud dominance (83% of the European market), the focus is on Interoperability and Data Spaces. Sovereign Cloud (EuroStack): A coalition of European providers (IONOS, Nextcloud, OVHcloud) is building "EuroStack" a suite of open-source cloud tools that ensure data never leaves European jurisdiction. Common European Data Spaces: Rather than letting Google silo all the data, the EU is building sector-specific data pools (Health, Manufacturing, Energy) where European companies can share data securely to train models that are "culturally and legally" European. Leveraging "Open Weights" (Mistral, DeepSeek, and Beyond) Europe has realized it doesn't need to build a "Google-sized" model from scratch if it leverages the global open-source ecosystem. The "Mistral" Model: France’s Mistral AI has pioneered the "Open Weight" approach. By releasing models that anyone can run on their own servers, they remove the need to send private data to US-based APIs like GPT-4 or Gemini. The DeepSeek Opportunity: The rise of Chinese models like DeepSeek-R1 changed the game in early 2025. Because DeepSeek is highly efficient and open-source, European firms are using it as a "foundation" to build their own specialized tools. Why this works: It breaks the US monopoly. If Europe uses Chinese open-source code on European-owned hardware, it achieves a "tactical independence" from Silicon Valley. #europe #ai #usa #china #google #alphabet #microsost #aws #nvidia #nanoinvest #freedom 

5

1 commentaar

White-collar jobs will change completely in the next 5 years. Not all of them. Not the leadership ones. But the operational, repetitive, process-heavy ones. Those are being replaced by AI agents. Faster than most people expect. At bont, we've already built our own pipeline to run 90% on automated, AI-interpreted processes. Human interaction only kicks in when AI can't judge. For us it’s no longer a future scenario. We're living it right now. The question for every business leader is no longer: "will AI replace jobs?" It's: "which parts of my team's work should have been automated two years ago?" If the society doesn't have a good answer for the people affected, this will create real pressure on employment markets. That's something I think about a lot. What’s your opinion on this?

17

26 commentaren

Most episodes of our Data-powered Innovation Jam we record. Others we live. Guess to which category our newest episode belongs. It comes fresh from a tech-soaked tour of San Francisco and Silicon Valley by my co-host Robert (Dr Bob) Engels and our 'Tech Guy' buddy Alex Bulat- van den Wildenberg. From TechCrunch to GitHub Universe and whatever Google is calling their AI centre this week: the duo delivered a set of ‘live’ interviews with some crucial movers & shakers in the San Francisco realm. And honestly, these interviews felt more like street-corner improv than a podcast schedule (a 'jam', anyone?). And what do you know … I finally managed to sneak in a mention of my heroes from Metallica as well (took me only dozens of episodes; ah, the small victories.) But the Bay Area has so much more to offer musically: funk, jazz, psychedelia, garage rock - all that glorious, weird SFO energy. Somehow, it all blends very well with the conversations Robert and Alex had with founders, researchers and innovators, powering data and AI to solve problems that actually matter. So tune in on your favourite podcast platform and pump up the volume - because in San Francisco, tech always comes with a backbeat. Weiwei Feng Arne Rossmann Sudeshna S. Sanna Östberg Kevin M. Campbell

33

2 commentaren

🤖 Agentic AI. 🧩 Multi-agent orchestration. ⚡ Autonomous execution. These aren’t buzzwords anymore — they’re shipping in production. A month ago, we wrote about the shift from features to experiences. About AI moving from answering questions → taking action. That future arrived faster than even we expected. 🚀 THE AGENT LAYER IS HERE 🔹 Claude Code — Terminal-native AI with full repository context. Not autocomplete. Not a copilot. An autonomous coding agent executing multi-step workflows. 🔹 n8n Agent Systems — RAG-powered agents with tool-calling across 400+ integrations. They reason, retrieve context, make decisions, and complete end-to-end workflows. 🔹 Moltbot / Clawdbot — Chat-to-compute interfaces where natural language triggers real system commands. Conversational AI that doesn’t just respond — it executes. 🔹 Google Antigravity — Multi-agent IDE orchestration. You’re no longer writing code line by line. You’re managing a fleet of specialized agents in parallel. ⚠️ And the Google Drive deletion incident? A reminder that AI safety isn’t optional — it is the product. 🔍 THE PATTERN IS CLEAR We’ve moved from: ❌ Prompt → Response ✅ Intent → Autonomous Execution AI has left the inference layer. It’s now in the action layer, with: 🧠 Long-term memory 🔌 Tool use via real APIs 👁️ Multi-modal understanding 🗺️ Planning before execution 🧾 Persistent state 🕰️ BY 2030, THE GAME ISN’T “WHO HAS THE BEST MODEL” It’s who builds the best orchestration layer: 🔐 Permission frameworks 🧑‍✋Human-in-the-loop triggers 📜 Audit & rollback systems 🛡️ Safety guardrails 🎨 Experience design The intelligence is commoditized. Trust is the moat. 🛠️ WHAT WE’RE BUILDING (IN PRODUCTION) We believe unconstrained agency doesn’t scale. We’re building: 🔸 Agentic workflow systems with real RAG & tool-use 🔸 Context-aware dev copilots 🔸 Embedded AI presence inside communication experiences All with: ✅ Explicit permissions ✅ Transparent audit trails ✅ Human checkpoints ✅ Rollback & recovery Because autonomy without control isn’t innovation — it’s liability. 🌐 The infrastructure layer for autonomous AI is forming right now. If you’re building here, you’re not just shipping features — you’re defining how humans and autonomous systems co-exist. #AgenticAI #AIAgents #LLMOps #AIInfrastructure #AIOrchestration #AutonomousAI #RAG #BuildingInPublic #FutureOfAI

14

1 commentaar