Axelera AI reels in $27M for its edge AI chip
Netherlands-based chip startup Axelera AI B.V. today announced that it has secured a $27 million early-stage funding round.
The Series A investment was led by venture capital firm Innovation Industries. The Federal Holding and Investment Company of Belgium participated as well along with imec.xpand, a venture fund associated with the IMEC nanotechnology research lab. Axelera AI detailed today that it has also secured a $6.7 million loan from the Netherlands Enterprise Agency.
Founded in 2021, Axelera AI is developing a chip optimized to run artificial intelligence software at the edge of the network, or outside data centers. The startup envisions its silicon finding use in connected devices such as industrial robots and smart city systems.
Axelera AI’s chip is based on an internally developed processing technology that the startup describes as an in-memory computing engine. The startup manufactured a test processor, dubbed the Thetis Core, in 2021 to evaluate the capabilities of the in-memory processing engine.
Modern processors comprise two main types of circuits: logic circuits that perform computations and memory circuits, which store the data on which the computations are carried out. In-memory computing is an approach that uses the memory circuits to carry out processing. For some types of software, this approach can provide faster performance than traditional chip technology.
The startup claims that the Thetis Core can provide performance of 39.3 TOPS, or 39.3 trillion computations per second. Speeding up the chip’s clock frequency can increase its peak performance to 48.16 TOPS.
According to Axelera AI, the Thetis Core carries out processing with a high degree of energy efficiency. The startup’s internal analysis determined that the Thetis Core provides 14.1 TOPS of performance per watt. Depending on the type of AI software that the processor runs, the Thetis Core’s peak energy efficiency can more than double to 33 TOPS per watt.
Energy efficiency is an important consideration in the connected device market that Axelera AI is targeting with its technology. Many of the connected devices that companies deploy at the edge of the network are powered by batteries. As a result, chipmakers must optimize their processors to use a limited amount of electricity.
Axelera’s first commercial chip will reportedly provide performance equivalent to “hundreds” of TOPS. It’s expected to feature several enhanced versions of the in-memory computing engine that powers the startup’s current Thetis Core processor. According to the company, the chip will begin shipping to an initial set of customers and partners early next year.
“The closing of this investment round comes closely after our Thetis Core announcement, proving our technology’s advantages to the AI community,” said Axelera AI co-founder and Chief Executive Officer Fabrizio del Maffeo. Axelera AI announced the Thetis Core in May.
The new funding will help the startup begin mass production of its chip. In parallel, the startup will expand its workforce to support product commercialization initiatives.
