Axelera AI

Axelera AI is a Dutch fabless semiconductor startup that designs AI Processing Units (AIPUs) for edge inference. Founded in 2021 out of Bitfury's AI incubation program and imec's deep-tech ecosystem in Eindhoven, the company's core innovation is a digital in-memory computing architecture — matrix-vector multiplications are executed inside SRAM arrays, dramatically reducing the data movement that bottlenecks conventional neural-network accelerators. The first commercial silicon, Metis, delivers up to 214 TOPS (INT8) at roughly 15 TOPS/W and is offered in M.2 and PCIe form factors; multi-chip boards stack four Metis dies for up to ~856 TOPS. Target workloads span computer vision, object detection, robotics, industrial automation, defense, and agritech — all areas where edge latency and power budgets exclude cloud GPU solutions. In 2025 Axelera unveiled Europa, the next-generation AIPU reaching 629 TOPS INT8 with PCIe accelerators shipping in H1 2026. The company raised its latest equity round of over $250 million in February 2026 — the largest ever for an EU AI semiconductor startup — led by Innovation Industries with BlackRock and Samsung Catalyst Fund among investors, bringing total capital (equity + grants + venture debt) above $450 million. Separately, a €61.6 million EuroHPC/DARE grant funds the third-generation Titania chip targeting generative AI and computer vision. Axelera is headquartered at the High-Tech Campus AI Innovation Center in Eindhoven and has grown to serve its 500th global customer.

Axelera is the leading European bet on purpose-built edge AI silicon: its in-memory computing approach delivers an order-of-magnitude better TOPS/W than GPU-based alternatives, and the February 2026 round — the largest ever for an EU AI chip company — with backers including BlackRock and Samsung validates both the technology and Europe's ambition to own a slice of the AI chip stack.

axelera.ai Edge AI

Netherlandsprivate

At a glance

Total raised

$450M

last round series c

Headcount

—

no public estimate on record

Headquarters

Eindhoven

Netherlands

Founded

2021

5 years active

Status

Private

Edge AI

Subsector

Edge AI

For subscribers

Subscribe to see the full profile

Including Funding rounds, Bull / Bear thesis, Stock + earnings, Roster changes, Patents, News, and Open roles.

Founding

50% off

/month

91 of 100 spots left

Full

$10

/month

Available once all 100 Founding Member spots are claimed.

Get full access

Already subscribed? Sign in →

Funding history

Total raised

$450M

Rounds tracked

Latest stage

Series C

Date	Stage	Amount	Valuation	Lead investors
Feb 24, 2026	Series C	$250M	—	Innovation Industries
Mar 6, 2025	Grant	$68.2M	—	EuroHPC Joint Undertaking (DARE project)
Jun 27, 2024	Series B	$68M	—	Invest-NL Deep Tech Fund, European Innovation Council Fund
May 22, 2023	Series A Extension	$23M	—	Innovation Industries
Oct 25, 2022	Series A	$27M	—	Innovation Industries
Sep 15, 2021	Seed	$12M	—	Bitfury Group

How they make money

Hardware + recurring servicesPattern

Axelera AI sells physical AI Processing Units — the Metis AIPU in M.2 module and PCIe card form factors, and the forthcoming Europa PCIe accelerator cards — to OEMs, system integrators, defense primes, and industrial automation vendors. Revenue is primarily hardware-driven: customers purchase Metis or Europa boards and cards, which are manufactured by contract fabs and assembled into customer solutions. This is a classic fabless semiconductor model where margin depends on silicon yield, ASP, and volume ramp. The recurring-services layer comes through the Axelera SDK and partner ecosystem: ongoing software support, model compilation tooling, quantization services, and a model zoo that simplifies deployment of standard vision networks onto Axelera silicon. As the installed base grows, SDK and support contracts provide a growing software attach rate on top of each hardware sale. The company has also secured grant funding (€61.6M EuroHPC/DARE for the Titania chiplet) which supplements equity capital for R&D rather than contributing to product revenue. With Europa shipping H1 2026 and 500+ customers already on Metis, the near-term revenue mix is overwhelmingly hardware, with software and services expected to grow as a percentage of revenue as the platform matures.

Generated by editorial-sourced · June 4, 2026Analytical summary · not investment advice

Founder backgrounds

Fabrizio Del Maffeo

Co-founder & CEO

Fabrizio Del Maffeo holds a master's degree in telecommunications engineering from Politecnico di Milano and completed executive education in mergers and acquisitions at London Business School. He began his career in industrial computing, serving as Country Manager and Sales Director across Europe at Advantech, the world's largest industrial IoT computing company. He then became Vice President and Managing Director of AAEON Technology Europe (within the ASUS Group), where he founded the UP Bridge the Gap product line, which became a leading reference platform for AI and IoT on Intel silicon. Prior to co-founding Axelera AI, he served as Head of AI at the Bitfury Group, where the company was incubated. Del Maffeo co-founded Axelera AI in 2021 and serves as CEO, having grown the company from inception to over 200 employees and 500 global customers in four years.

Evangelos Eleftheriou

Co-founder & CTO

Evangelos Eleftheriou holds a BSc in Electrical and Computer Engineering from the University of Patras, Greece, and both a Master of Engineering and PhD in Electrical Engineering from Carleton University, Canada. He spent over 35 years at IBM Research – Zurich, holding multiple senior research and management positions and becoming an IBM Fellow — the company's highest technical honor — in recognition of his pioneering work in data storage, neuromorphic computing, and AI systems. He was elected a Fellow of the IEEE in 2002 and inducted as a Foreign Member of the US National Academy of Engineering in 2018. He has authored or co-authored over 250 publications and holds more than 160 patents. After retiring from IBM, he co-founded Axelera AI in 2021 to commercialize the digital in-memory computing paradigm he helped develop.

Bram Ernst Verhoef

Co-founder & Head of AI

Bram-Ernst Verhoef earned his PhD from KU Leuven in 2010, with a background spanning statistics, psychology, and computational neuroscience. He conducted postdoctoral research at Harvard University and the University of Chicago, studying the computational neuroscience of attention mechanisms. In 2017 he joined imec (Eindhoven/Leuven) as a Principal Member of Technical Staff, where he led algorithm development for a novel analog compute-in-memory deep learning chip — work that directly prefigured Axelera's architecture. He co-founded Axelera AI in 2021 and leads machine learning algorithm development, including quantization and compiler integration for the company's AIPU family.

Ioannis Papistas

Co-founder

Ioannis Papistas holds a degree in Computer Engineering from Aristotle University of Thessaloniki (2008–2014) and a PhD from the University of Manchester (2014–2018). Following his doctorate he held a postdoctoral research fellowship at the University of Manchester before joining imec as a Research and Development Engineer, where he worked on advanced AI hardware design. He co-founded Axelera AI in 2021, contributing deep expertise in chip microarchitecture and R&D engineering to the development of the Metis and Europa AIPU platforms.

Marco Barbato

Co-founder & Director of AI Integrated Systems

Marco Barbato has over 20 years of professional experience in AI and Industrial IoT. He began his career at Advantech Europe in 2001, and in 2016 joined AAEON Europe as AI and IoT Product Development Director. In 2020 he moved to the Bitfury Group as Director of AI Integrated Systems, where Axelera AI was incubated and where he worked alongside Fabrizio Del Maffeo. He co-founded Axelera AI in 2021, bringing his system-integration and product expertise to bear on the company's edge AI platform and partner ecosystem.

AI bull / bear

Bull case

Axelera's digital in-memory computing (d-IMC) architecture attacks the core bottleneck of neural-network inference — memory bandwidth — by executing matrix-vector multiplications inside SRAM arrays, delivering roughly 15 TOPS/W on the shipping Metis chip at a time when GPU-based edge alternatives consume an order of magnitude more power per operation. That efficiency advantage maps directly onto real deployment constraints: factory-floor automation, robotics, and defense platforms cannot run 300W server GPUs. The company has already shipped to 500+ global customers, booked the largest equity round ever for an EU AI semiconductor startup ($250M Series C in February 2026 backed by BlackRock, Samsung Catalyst Fund, and Innovation Industries), and added a €61.6M EuroHPC grant for the third-generation Titania chiplet — giving it a funded roadmap through at least 2028 with no obvious capital gap. The geopolitical tailwind is structural: European governments and enterprises are under pressure to source AI infrastructure outside US hyperscalers and Chinese supply chains, and Axelera is the continent's best-funded, most commercially advanced native edge-AI silicon vendor. The progression from Metis (214 TOPS) to Europa (629 TOPS, H1 2026) to Titania (targeting generative AI and HPC, 2028) shows deliberate market expansion from embedded modules toward edge-server and rack deployments — broadening TAM without abandoning the efficiency positioning.

Bear case

Axelera is competing in a market where Nvidia, Qualcomm, Intel, and a cohort of well-funded startups (Hailo, Kneron, d-Matrix) are all pushing edge-inference silicon. Nvidia's Jetson Orin and Thor families enjoy an insurmountable software ecosystem advantage — CUDA, TensorRT, and thousands of pre-optimized models — making customer adoption of alternative silicon a real switching-cost problem regardless of raw TOPS/W. Axelera's own SDK and model zoo are still maturing, and every new customer currently requires dedicated integration effort, which compresses margins and limits scale. A second risk is timing. Edge AI inference demand is real but fragmented across verticals (industrial, defense, agritech, retail) that each require custom validation cycles and reference designs; revenue ramp has been slower than cloud AI infrastructure. The company has taken on $450M+ in combined equity, grants, and venture debt to fund three successive chip generations — a capital intensity that demands either a very fast revenue ramp or a strategic acquirer. If Europa's commercial traction disappoints or if Nvidia expands Jetson aggressively into the same form factors, Axelera's runway advantage may narrow faster than its next chip generation can print revenue.

Generated by editorial-sourced · June 4, 2026Refreshes quarterly · see disclosures

In the news

Axelera AI · last 30 days

No articles ingested yet for Axelera AI. Once the hourly news pipeline is live, every article the classifier tags as mentioning this company appears here with its one-line AI summary and sentiment.

Watch

Leadership

Fabrizio Del Maffeo

Co-founder & CEO & CEO & Co-founder

Evangelos Eleftheriou

Co-founder & CTO & CTO & Co-founder

—

Bram Ernst Verhoef

Co-founder

—

Ioannis Papistas

Co-founder

—

Marco Barbato

Co-founder & Director of AI Integrated Systems

—

Marta Ostroumoff

CFO

—

Alexis Crowell

CMO & General Manager, Americas

—

Products

Metis AIPU
Axelera's first commercial AI Processing Unit, built around a digital in-memory computing (d-IMC) core that executes matrix-vector multiplications directly inside SRAM — eliminating the memory-bandwidth bottleneck that limits GPU inference at the edge. Metis delivers up to 214 TOPS (INT8) at ~15 TOPS/W and is offered in M.2 modules, single-chip PCIe cards, and quad-chip cards scaling to ~856 TOPS. It targets computer vision, object detection, and embedded AI workloads across robotics, industrial, defense, and retail verticals, with Axelera SDK and pre-compiled model zoo for rapid deployment.
Europa AIPU
Next-generation AI Processing Unit announced in 2025, targeting edge servers, enterprise workstations, and rack-mount deployments. Europa delivers up to 629 TOPS at INT8 precision, significantly expanding Axelera's addressable market beyond embedded modules. PCIe accelerator cards are expected to begin shipping in H1 2026, bringing the company's in-memory computing architecture to higher-throughput inference scenarios including generative AI at the edge.
Titania (in development)
Third-generation chip under development with a €61.6 million EuroHPC Joint Undertaking DARE grant, targeting generative AI and advanced computer vision processing. Titania is intended to push Axelera's in-memory computing approach into the emerging on-device and near-edge generative AI workload tier, reinforcing Europe's sovereign AI semiconductor ambitions.
Axelera SDK & Partner Ecosystem
Software development kit providing model compilation, quantization tooling, and runtime support for Axelera AIPUs. Paired with a global partner program launched to help system integrators and OEMs build customer-ready edge AI solutions, the SDK ensures that deploying neural networks to Metis and Europa hardware is accessible without deep hardware expertise.