ETH Zurich and TUM Achieve Milestone with First 7nm AI Chip Tape-Out in Europe

Europe's Academic Leap into Advanced AI Hardware

In a landmark achievement for European higher education, researchers at ETH Zurich and the Technical University of Munich (TUM) have successfully completed tape-outs of the continent's first 7-nanometer (7nm) AI chips. Tape-out refers to the final step in chip design where the layout is sent to a foundry for manufacturing, marking the transition from digital blueprint to physical silicon. This milestone underscores the growing prowess of European universities in semiconductor technology, particularly for artificial intelligence (AI) applications, amid global pushes for technological sovereignty.

These developments are not isolated feats but part of a broader renaissance in European academia's role in hardware innovation. With AI permeating every sector from healthcare to autonomous systems, the ability to design efficient, secure chips at advanced nodes like 7nm positions these institutions as leaders in bridging the gap between research and real-world deployment.

TUM's Neuromorphic AI Chip: Pioneering Local Processing

At the forefront is TUM's neuromorphic AI chip, unveiled on February 4, 2026, as the European Union's first university-designed processor using modern 7nm technology. Neuromorphic computing mimics the human brain's neural structures, enabling ultra-efficient pattern recognition and data processing directly on the device—eliminating the need for power-hungry cloud connections.

Led by Prof. Hussam Amrouch, Chair of AI Processor Design, the chip adheres to standards set by Taiwan Semiconductor Manufacturing Company (TSMC), the world's leading foundry. Built on the open-source RISC-V instruction set architecture (ISA)—a flexible alternative to proprietary designs like ARM—it allows customization for specialized tasks. Imagine analyzing electrocardiograms (ECGs) for arrhythmias or electroencephalograms (EEGs) for neurological insights without transmitting sensitive data externally.

"You can buy a Ferrari, but that doesn't necessarily make you faster in the city. An e-bike is more efficient here," Amrouch analogized, highlighting how specialized chips outperform general-purpose giants in targeted scenarios. The design emphasizes energy efficiency, privacy via local processing, and security against hardware Trojans—critical for sectors like defense and automotive.

This project stems from TUM's Munich Advanced Center for High-Tech AI Chips (MACHT-AI), funded by Bavarian ministries with over €10 million. Over five years, it will train 300+ students and researchers in AI chip design, fostering a pipeline of talent for Europe's semiconductor ecosystem.

ETH Zurich's Picobello: A Massive Multi-Core Testbed

Complementing TUM's work, ETH Zurich's Picobello project represents the first large-scale 7nm tape-out entirely by a European academic team, led by the Parallel Ultra-Low Power (PULP) platform under Prof. Luca Benini. This behemoth integrates 144 RISC-V cores across 16 clusters, tensor coprocessors for AI acceleration, and over 10 MB of on-chip memory, clocked at 1 GHz.

Picobello serves as a test vehicle for inference acceleration—running pre-trained AI models—in edge and cloud environments, prioritizing extreme energy efficiency for agentic AI systems that act autonomously. Fabricated via EUROPRACTICE and Synopsys tools, it contributes to the EuroHPC Joint Undertaking's EUPILOT initiative for indigenous high-performance computing (HPC) accelerators.

Involving 20 researchers, including two from TUM, Picobello demonstrates scalable open-source design flows. "AI is now driving the evolution of computing chips and brings enormous opportunity," Benini noted, stressing Europe's momentum in closing the innovation gap.

The Synergistic Collaboration Fueling Progress

ETH Zurich and TUM's efforts intersect through shared tools and personnel, amplified by Synopsys' Academic & Research Alliances program. Both leveraged Fusion Compiler for physical design, full Synopsys flows, and ASIP Designer for RISC-V customization. TUM's ultra-low-power multi-core accelerator (2 mm² die) was taped out in just four months—a testament to streamlined workflows.

This partnership exemplifies how university-industry ties accelerate academia's entry into commercial-grade silicon. Frank K. Gürkaynak from ETH emphasized team coordination for billion-transistor complexities, while Amrouch praised Synopsys support: "We don’t have the luxury of time to spend another five or six years catching up."

Explore research jobs in semiconductor design at leading European universities to join such groundbreaking teams.

Technical Deep Dive: Mastering 7nm for AI

7nm refers to the transistor gate length in nanometers, enabling denser integration—billions of transistors per chip—for superior performance-per-watt. Step-by-step, designing at this node involves:

• Architecture definition: RISC-V cores with AI extensions for vector/tensor ops.
• RTL coding: Hardware description language for logic.
• Synthesis and place-and-route: Using EDA tools like Synopsys to map to silicon.
• Verification: Simulating against TSMC process design kits (PDKs).
• Tape-out: GDSII file submission to foundry.

Neuromorphic elements in TUM's chip emulate spiking neural networks for sparse, event-driven computation, slashing power by up to 90% versus traditional von Neumann architectures.

Training Tomorrow's Chip Designers

Beyond silicon, these projects prioritize human capital. MACHT-AI's first workshop launches in March 2026, blending engineering and computer science curricula with hands-on tape-outs. ETH's PULP educates via open-source repositories, democratizing access.

Europe's semiconductor skills shortage—projected to need 100,000+ engineers—makes this vital. Bavaria's ecosystem, including GlobalFoundries and Infineon, benefits from university-fed talent. Prof. Amrouch envisions graduates wielding industry flows with commercial IPs.

Check faculty positions or career advice for advancing in AI hardware academia.

European Chips Act: The Policy Backbone

These breakthroughs align with the €43 billion European Chips Act (2023), aiming for 20% global market share by 2030. Initiatives like Chips Joint Undertaking fund pilot lines (e.g., imec's NanoIC) and design platforms. ESMC Dresden (16/12nm initially, scaling to 7nm+) will produce TUM designs from 2028.

• Benefits: Supply chain resilience post-COVID/Ukraine war.
• Risks: Lagging Asia's sub-3nm nodes.
• Solutions: University fabs, cross-border consortia.

TUM President Thomas Hofmann hailed it as sustainable tech sovereignty.

Real-World Impacts and Stakeholder Perspectives

For healthcare, localized AI enables privacy-preserving diagnostics; in quantum control, low-latency processing. Industry stakeholders like Synopsys see it boosting Europe's AI/HPC autonomy. Bavarian Minister Markus Blume called it a "big breakthrough" for performance, efficiency, security.

Students gain actionable skills: from RTL to verification, preparing for roles at Intel, TSMC Europe outposts, or startups.

Future Horizons: Scaling and Beyond

TUM targets three annual designs; ETH integrates Synopsys IP for bandwidth gains. Horizon: sub-7nm academia tape-outs, quantum-AI hybrids. Challenges include PDK access and fab capacity, addressed via Chips Act's €1.3B design investments.

Optimistic outlook: Europe's open-source ethos (RISC-V) fosters collaborative scaling, positioning universities as innovation hubs.

Discover European higher ed opportunities or professor jobs in tech-forward institutions.

Photo by Florian Schmid on Unsplash

Career Opportunities in Europe's AI Chip Revolution

This surge opens doors: research assistant roles in neuromorphic design, postdocs in RISC-V acceleration. With fabs like ESMC rising, demand for PhDs surges. Actionable insight: Master Synopsys/ Cadence tools via university programs; contribute to PULP GitHub for portfolio boosts.

Position yourself at the forefront—visit Rate My Professor for insights on programs at ETH/TUM, or browse higher ed jobs.