The generative AI boom is driving demand for AI chips specifically designed to coach and run generative AI models. And the massive players, from enterprise capitalists to startups, try to get in on the motion from the beginning.
Masayoshi Son of SoftBank is According to reports 100 billion dollars for a chip initiative to compete with technology giant Nvidia. OpenAI is now said are in talks with investment firms to start out an organization to fabricate AI chips.
AI chip startup Axelera has been relatively cautious. Nevertheless, the corporate has managed to draw supporters corresponding to Samsung, partially by specializing in a distinct segment within the emerging AI chip market: chips that implement AI on Edge devices.
“There is not any denying that the AI ​​industry has the potential to rework a wide selection of industries,” said Fabrizio Del Maffeo, co-founder and CEO of Axelera, in an interview with TechCrunch. “However, to actually leverage the worth of AI, corporations need an answer that gives high performance and efficiency at a balanced cost.”
Axelera – headquartered within the Netherlands and employing around 180 people in offices in Belgium, Switzerland, Italy and the UK – develops AI-powered chips and systems for security, retail, automotive and robotics applications, supplying them to partners that manufacture B2B edge computing and Internet of Things products.
Axelera emerged from an initiative led by Del Maffeo and a gaggle from the Belgian technology lab Imec, together with Evangelos Eleftheriou and a gaggle of IBM researchers from Zurich who desired to develop a highly efficient AI chip architecture. The founding team developed much of Axelera inside the Bitfury Group, a blockchain company specializing in Bitcoin hardware.
The defining features of Axelera’s AI hardware stack are the RISC-V instruction set architecture (ISA) and in-memory computing.
ISAs are a technical specification underlying chips that describes how software controls the chip's hardware. Chip designers typically license an existing ISA from a significant chip manufacturer corresponding to Arm or Intel, but RISC-V provides an open alternative with no royalties. In-memory computing refers to running calculations in a system's memory. R.A.M. to cut back the latency attributable to storage devices.
Axelera will not be the primary to try an in-memory and/or RISC-V-based architecture for AI chips.
NeuroBlade develops chips that mix compute power and memory in a single hardware block for data processing. MemVerge, GigaSpaces, Hazelcast and H20.ai also offer in-memory hardware solutions for AI and data analytics applications. Tenstorrent, backed by Hyundai Motor Group and Samsung, sells AI processors and other related IPs based on RISC-V.
Axelera has tried to distinguish itself from the competition by providing each chip hardware and software to administer and deploy AI models on that hardware. And by all appearances, the strategy appears to be working.
Axelera announced on Thursday that it has closed a $68 million Series B funding round, bringing the corporate's total capital raised to $120 million. Backers within the round included the European Innovation Council Fund, the Innovation Industries Strategic Partnership Fund, Invest-NL and the Samsung Catalyst Fund.
According to Del Maffeo, the brand new money might be used to expand into latest markets before Axelera's flagship AI platform Metis reaches full production within the second half of 2024. Axelera can also be eyeing the information center chip market and plans to fund research and development of chips for high-performance computing.
“Metis entered full production within the second quarter and can begin shipping in volume within the third quarter,” said Del Maffeo. “Axelera AI is currently developing a brand new generation of products for computer vision, large language models and enormous multimodal models. This latest product family might be unveiled later this 12 months and enter full production in 2025.”
The challenge might be to ship its AI chips at scale – and compete against the countless others within the AI ​​chip race. Many competitors have impressive backing; a Crunchbase report June shows that VC-backed chip startups have raised nearly $5.3 billion in only 175 deals thus far this 12 months.
However, the reward might be significant. According to data from Statista and Market.us, the The AI ​​chip market could generate as much as $67 billion revenue by 2027. Axelera has little probability of dethroning established providers corresponding to Nvidia within the near future, if in any respect. (Nvidia has a estimated 70 to 95 percent market share of the AI ​​chip market (in accordance with Mizuho Securities). But grabbing even a fraction of the market can be a big victory.
“The funding supports our mission to democratize access to AI, from the sting to the cloud,” Del Maffeo said, adding that Axelera has “dozens” of enterprise customers. “By expanding our product lines beyond the sting computing market, we’re capable of address industry challenges within the AI ​​inference space and support current and future AI processing needs.”
