Axelera Raises $250M in Largest EU Semi Funding Round Ever

//php echo do_shortcode(‘[responsivevoice_button voice=”US English Male” buttontext=”Listen to Post”]’) ?>

Netherlands-based Axelera has completed a funding round over $250 million, bringing the AI chip company’s total raised to $450 million. This is the largest ever funding round for an AI chip company in the EU, according to the company.

Axelera’s record-breaking funding round is a vote of confidence for the company’s technology and market, but it also indicates that investors’ confidence in European chip companies is growing, Axelera CEO Fabrizio Del Maffeo told EE Times.

“Investors increasingly recognize that Europe has world-class AI talent,” Del Maffeo said. “Many of the people building Axelera helped create globally successful semiconductor companies over the past few decades—that capability has always existed.”

As sovereign AI infrastructure gains strategic importance, companies like Axelera are at the center of discussions, he added. Axelera is receiving funding as part of the DARE project, an EU-funded initiative to develop a European HPC stack, for which Axelera is developing an AI accelerator chiplet. This project is on track at the current architectural co-ordination and software interoperability phase, Del Maffeo said.

The company’s immediate focus in the coming months will be commercial expansion.

“This investment will allow us to scale faster while still prioritizing product execution, software, and customer success,” Del Maffeo said. “We are supporting customers worldwide who use edge AI to solve real business problems, while expanding our product roadmap to encompass larger AI systems and supercomputing-class infrastructure with [Axelera’s second-gen chip] Europa.”

Axelera has around 500 customers today. While the company’s initial customer base was a mix of hobbyists and those working on developing commercial products, that mix is evolving, Del Maffeo said.

“That mix has evolved significantly,” Del Maffeo said. “Early adoption of [Axelera’s first-gen chip] Metis naturally included developers and innovators experimenting with the platform. Today, we are seeing a clear shift toward commercial deployments and customers building production systems. Many of our customers are now integrating Metis into industrial products, infrastructure, and enterprise solutions.”

Axelera’s first-gen Metis chip was aimed largely at vision, while second-gen Europa can also handle bigger workloads like generative AI.

“Metis and Europa are tackling different challenges in the market and while they do attract different types of customers, we also have customers that will deploy both depending on which part of their business they are addressing,” Del Maffeo said.

Second generation

Europa is designed to run power- and cost-efficient edge AI, Del Maffeo told EE Times in an earlier conversation. The second-gen chip will double the number of AI cores as first-gen Metis to eight, and add two clusters of eight RISC-V vector cores for pre-and post processing. There is also an H.265 video decoder on-chip. Support for INT16 has been added (Europa supports INT16/8/4) and total compute is now 629 TOPS (INT8). There is also a quad-core RISC-V system controller that can run an operating system if needed. Overall, Europa will offer 3× to 5× the performance of Metis.

Europa’s on-chip memory has been upgraded to 128 MB and its memory interface has been upgraded from LPDDR4 to LPDDR5 at 200 GB/s. Europa’s power envelope is 45 W and up to four chips will be available on a PCIe card.

Europa can handle generative AI in the device, allowing Axelera to expand into markets like automotive, robotics, and edge servers.

“Europa will be cross-sold to existing customers,” Del Maffeo said. “If you are in surveillance today and you run a convolutional neural network to detect people or objects, tomorrow you want to understand the context. Metis could run small VLMs, but if you want to run a more sophisticated algorithm and really understand what’s going on, Europa fits the bill. It’s more powerful and you can have more camera streams.”

Large parts of Axelera’s software stack have been open-sourced (the company retains control of some areas related to Axelera’s IP, Del Maffeo said), in order to better support a growing customer base.

The company’s Partner Accelerator Network is a key part of Axelera’s strategy to help edge AI applications actually reach production; the network includes system integrators, OEMs, ISVs, and framework and model providers.

“Many of these partners are building applications, including intelligent video analytics, quality inspection systems, inventory optimization, and advanced computer vision platforms,” Del Maffeo said.

Combining a strong ecosystem with tightly integrated hardware and software is the best way to avoid fragmentation in edge AI, he added.

At 629 TOPS, Europa is within range to compete with Nvidia products like Jetson and Thor. Edge customers often don’t have the resources to start designs from scratch, Del Maffeo said, so they need to rely on legacy CPUs and code.

“A lot of customers in this space have legacy applications, they already have a CPU from NXP or Intel or AMD or TI,” he said. “Jetson is a system on chip, whereas they can just plug in our accelerator. We try to give tools which allow customers to keep their legacy designs and get to market faster by adding AI to any kind of existing application.”

Axelera’s third-generation chip, Titania, will be a chiplet-based architecture that will scale for edge and data center applications based on the number of chiplets in a package.

Meanwhile, Europa samples are due in Q2 2026.