CONNECT WITH US

Computing is the new currency: Axelera AI bags US$68 million, eyes for global AI expansion

Jerry Chen, DIGITIMES Asia, Taipei 0

Credit: Axelera AI

Netherlands-based Axelera AI, an AI acceleration hardware and software provider has announced the successful completion of their US$68 million Series B financing round.

This oversubscribed funding round brings the total amount raised by the company to US$120 million. Speaking to DIGITIMES Asia, Axelera AI CEO Fabrizio Del Maffeo says their funding benefited from the renewed interest in AI infrastructure driven by Nvdia's surge in 2023.

"The infrastructure always wins, there is no AI without computing," says Del Maffeo. "Before it was said that oil was the new currency, then data became the currency. But today, computing is the new currency, hence why investors start to look at us seriously," he added.

IDC predicts that by the end of 2027, the market for IT infrastructure supporting AI semiconductors will grow to US$193 billion.

Huge financial boost

Since its founding in 2021, Axelera AI has quickly grown to employ over 180 professionals, with backgrounds ranging from IBM and IMEC to Intel.

The official press release claims the company's series B round is Europe's largest oversubscribed funding in the fabless semiconductor sector. Institutional investors participating in this round include the Invest-NL Deep Tech Fund, the European Innovation Council Fund, the Innovation Industries Strategic Partners Fund, and the Samsung Catalyst Fund.

Existing investors such as Verve Ventures, Innovation Industries, Fractionelera, and the Italian sovereign fund CDP Venture Capital SGR also contributed.

In the statement, Marco Chisari, Head of Samsung Semiconductor Innovation Center expresses enthusiasm for supporting Axelera AI. Chisari praises the company for how they utilize digital in-memory computing to accelerate AI at the edge while addressing critical challenges such as the 'memory wall' and power consumption.

Del Maffeo says the funding they received will primarily support research and development, team expansion, and global sales efforts. Del Maffeo noted that AI is heavily software-driven, and the company plans to have more software personnel than hardware personnel by the end of next year.

The company says its flagship Metis AI Platform, the AI processing unit for inference, has been a key driver of its success, offering a 3-5x improvement in efficiency and performance. The company anticipates the unit will enter full production in the second half of 2024, positioning it as a leader in efficient and cost-effective AI processing units for inference.

Applications in various industries

The primary application of the company's proprietary chip is in computer vision, which is utilized across various sectors, including retail, surveillance, medical, robotics, drones, telecom, and industrial automation. "The surveillance market", Del Maffeo highlighted, "has shown strong traction."

The statement also revealed the funding will support Axelera AI's global expansion strategy, targeting growth opportunities in sectors such as automotive, digital healthcare, Industry 4.0, retail, robotics, drones, and surveillance. Additionally, the expansion will also support the design of AI accelerators with better performance, efficiency, and price advantages customized for exa-scale and peta-scale HPC centers.

Collaboration across the industry and regions

Beyond its current market of North America, Europe, and the Middle East, Del Maffeo says he also sees Taiwan, South Korea, and Japan as having the most potential.

Del Maffeo says the company is not targeting China at this stage of its operations. He highlights their strong integration with Taiwanese partners for design and system integration, with Advantech and ASE being key collaborators.

Del Maffeo emphasized the company's collaborative approach, working with other semiconductor companies and being part of the automotive ecosystem of IMEC. He envisions a future where AI accelerators enable creative content generation, similar to how Nvidia's graphics cards are used today.