It seems AI chips are all the rage as Dutch semiconductor startup Axelera AI sees potential sales pipeline now reaching as much as $1.5 billion even as signed agreements remain a fraction of that figure.
CEO Fabrizio Del Maffeo told Reuters on Tuesday that Axelera has contracts worth tens of millions of dollars and more than 600 customers, while the company is pursuing potential sales of up to $1.5 billion. Reuters also reported that Axelera’s second-generation Europa accelerator will be integrated into systems from Dell Technologies and Super Micro Computer.
The Eindhoven-based company has raised more than $450 million in equity, grants, and venture debt since its 2021 incorporation, including a funding round of more than $250 million announced in February. Axelera said at the time that it had shipped products to its 500th customer.
Europa represents Axelera’s attempt to move beyond the lower-power edge applications targeted by its first-generation Metis platform.
The processor delivers up to 629 TOPS at INT8 precision, carries a 45-watt thermal design power, and can support as much as 64 GB of memory per chip, according to Axelera. The company is positioning the hardware for enterprise servers, generative AI inference, robotics, computer vision, and other workloads that require more computing power than traditional edge deployments.
Axelera claims Europa can deliver three to five times better performance per dollar and two to three times better performance per watt than competing products. Those comparisons are company benchmarks. Axelera’s own website also identifies its published 53.5-token-per-second result for the Qwen3-30B-A3B model as a pre-silicon simulation rather than measured production performance.
Europa itself was first announced in October 2025, when Axelera originally expected shipments to begin during the first half of 2026.
Axelera is also gaining a foothold in publicly backed European computing projects that are otherwise heavily dependent on Nvidia hardware.
The European High Performance Computing Joint Undertaking said in April that Italy’s €290 million IT4LIA AI Factory would use Nvidia Grace CPUs and Blackwell GPUs for its primary architecture while incorporating a separate inference partition using Axelera accelerators and European-designed SiPearl processors.
Germany’s HammerHAI project similarly includes Axelera accelerators alongside Nvidia-based infrastructure.
The architecture illustrates Axelera’s immediate competitive position. Rather than replacing Nvidia across an entire AI data center, the company is finding deployments in power-sensitive inference workloads where customers can add alternative accelerators alongside larger GPU installations.
The reported Dell and Supermicro integrations could broaden that route to market. Axelera already sells Metis-based configurations using Dell systems, while its commercial challenge now shifts toward converting Europa deployments and AI-factory opportunities into firm orders closer to the size of the $1.5 billion pipeline it is citing.