Axelera AI raises $68 million Series B to scale Metis accelerators and push into global markets
- ›Axelera AI announced a $68 million Series B round, supported by institutional investors including the EIC Fund.
- ›The company says its Metis AI Platform delivers major gains in inference performance and energy efficiency and will enter full production in the second half of 2024.
- ›Axelera reports a pipeline exceeding $100 million and a team of 180 plus staff including more than 55 PhDs.
- ›Products include Metis AIPU in PCIe and M.2 form factors and a larger Europa AIPU for edge servers, with claimed metrics up to 214 TOPS per AIPU and 15 TOPS per watt.
- ›Independent benchmarks and marketing claims should be treated cautiously because real world performance depends on system integration, model mix, and manufacturing scale.
- ›Key near term challenges include ramping production, securing advanced packaging and foundry access, and building software and partner ecosystems to capture vertical markets.
Axelera AI secures $68 million Series B to scale Metis accelerators
Axelera AI, an EIC Scaling Club member active in Next-Gen Computing, announced a $68 million Series B round in July 2024. The company positions the funding as a step to commercialise its Metis AI Processing Unit and expand into North America, Europe and the Middle East, and into verticals such as automotive, digital healthcare, Industry 4.0, retail, robotics and surveillance. The round is described internally as oversubscribed and as Europe’s largest Series B in the semiconductor industry, a claim that depends on how comparable transactions are defined.
Company snapshot and commercial claims
Founded within the recent deep tech wave targeting edge inference acceleration, Axelera reports rapid growth. The company says it now employs more than 180 people including over 55 PhDs, with the academic output of the team totalling more than 40,000 citations. Axelera states it has visibility on a commercial pipeline that exceeds $100 million, and that production of the Metis AI Platform will be in full swing in the second half of 2024.
Metis platform claims and what they mean
The Metis family is presented as an AI Processing Unit or AIPU built for inference tasks, especially computer vision. Axelera publishes headline numbers that aim to highlight compute density, energy efficiency and price performance. These include up to 214 TOPS at INT8 for a single Metis AIPU, up to 3200 frames per second on ResNet-50 benchmarks, efficiency of 15 TOPS per watt, and a claimed price performance of 16.4 frames per second per dollar on ResNet-50 for the PCIe configuration. The company advertises multiple form factors including M.2 cards, single AIPU PCIe cards, quad-AIPU PCIe cards, and a larger Europa AIPU for edge servers.
Technology approach and context
Product forms and published specifications
Axelera publishes multiple hardware SKUs meant to cover a broad set of edge and server use cases. Notable product variants include M.2 cards for embedded systems, a M.2 Max version with higher dedicated memory for larger models, single Metis AIPU PCIe cards, quad-AIPU PCIe cards that aggregate compute, and Europa which targets edge servers with higher aggregate TOPS. The company also offers a Metis Compute Board that pairs a quad-core Metis AIPU with an RK3588 host for integrated single-board solutions.
| Product | Claimed Peak TOPS | Typical power | Target use case |
| Metis single AIPU PCIe card | Up to 214 TOPS @ INT8 | Typical system power around 10 W for small systems | Edge vision inference, PCIe hosts |
| Metis quad-AIPU PCIe card | Up to 856 TOPS | Higher system power depending on configuration | High throughput edge and on-prem inference clusters |
| M.2 AIPU | Single AIPU performance in M.2 form factor | Low power, few Watts | Embedded and edge devices |
| M.2 Max | Single AIPU with up to 16 GB DRAM | Low power | Larger models on embedded devices, LLM/VLM inference at edge |
| Europa AIPU | Company cites larger scale option, up to 629 TOPS in marketing | Server class power envelope | Edge servers and facility consolidation workloads |
Benchmarks and independent testing
Axelera cites third-party testing by HotTech Vision and Analysis that reportedly placed Metis favourably on performance, power efficiency and detection accuracy for computer vision workloads. Published benchmark numbers such as TOPS, frames per second on ResNet-50, TOPS per watt and frames per dollar are useful starting points for comparison. Observers should note that benchmark conditions, quantisation formats, model variants and batch sizes all shape results. End users should verify claims in their own workloads and with system level tests, because integration choices can change outcomes significantly.
Market opportunity and the financing context
Axelera references market research from IDC that forecasts IT infrastructure spending for AI semiconductors reaching $193 billion by the end of 2027. Forecasts of this kind highlight a large addressable market, but they do not guarantee company specific uptake. Axelera aims to capture segments of edge inference and to move into datacenter and HPC markets. The Series B is positioned to help the company pursue those moves.
| Item | Published value |
| Series B size | $68 million |
| Reported team size | 180+ employees |
| PhDs in team | 55+ |
| Pipeline visibility | Exceeds $100 million (company claim) |
| Production target | Full production in H2 2024 (company claim) |
Risks and what to watch next
Axelera’s announcements combine optimistic product metrics with large commercial ambitions. That mix is common in early scaling deep tech announcements. There are several practical risks and dependencies to monitor as the company moves from prototype and sampling into volume shipments.
Why this matters for European deep tech
If Axelera scales as planned it would be an example of an EU-born hardware company successfully moving from research and prototyping into volume product shipments. Europe has been investing to strengthen its semiconductor and AI supply chains. Public backing through programmes such as Horizon and instruments such as the EIC Fund can be material for companies that must coordinate long development cycles and capital intensive manufacturing ramps. At the same time, success is not guaranteed and will depend on execution across engineering, supply chain and commercial partnerships.
Bottom line
Axelera’s $68 million Series B is a clear signal that investors see opportunity in energy efficient inference hardware aimed at the edge. The Metis platform presents compelling headline metrics and a broad product family. Independent verification of those metrics under customer workloads, and the company’s ability to scale manufacturing and software support, will determine whether those claims translate into market share. The EIC Fund involvement underscores the EU’s continuing interest in supporting native deep tech challengers in semiconductors and AI.
Further reading and sources
Key sources for this piece include Axelera AI product pages and benchmark claims, the EIC Scaling Club announcement of the Series B, public EIC Fund materials describing its investment role, and industry market forecasts such as those published by IDC. Readers evaluating technical claims should review independent benchmarks and product validation reports, and follow up on customer design wins and volume ship announcements.