SEMIFIVE's Partnership with HyperAccel for AI Chip Development
SEMIFIVE, a renowned design solution provider, has reached an exciting milestone in its mission to innovate custom silicon solutions. The recent agreement with HyperAccel marks a significant step towards developing a generative AI chip, named Bertha. This project employs advanced 4nm process technology and aims to initiate mass production in early 2026.
Revolutionizing AI with the New Chip
HyperAccel's innovation portfolio includes a unique Large Language Model Processing Unit (LPU), specifically crafted for handling transformer-based LLM tasks. This innovative chip is designed to outperform traditional Graphics Processing Units (GPUs) by offering enhanced performance, delivering speed improvements of up to double and presenting a price/performance ratio that is 19 times superior to that of conventional supercomputers.
SEMIFIVE's Commitment to SoC Design Solutions
As experts in system-on-chip (SoC) platforms, SEMIFIVE is taking purposeful strides toward reinforcing its commitment to developing AI-centric SoC design solutions. They are continuously optimizing the design process to fulfill customers' increasing demands for specialized AI hardware. Their proactive development of SoC chiplet platforms, in close collaboration with industry leaders, positions them at the forefront of this thriving market.
Insights from HyperAccel's Leadership
Joo-Young Kim, CEO of HyperAccel, expressed enthusiasm about partnering with SEMIFIVE, stating, "Collaborating with SEMIFIVE allows us to deliver AI semiconductor technologies that are not only cost-effective but also power-efficient. This innovation is set to markedly lower data center operational costs and open new business opportunities across various industries that increasingly rely on LLMs."
Remarks from SEMIFIVE's CEO
Brandon Cho, the CEO and co-founder of SEMIFIVE, shared his perspective on the partnership: "HyperAccel has established itself as a leader in LPU technology for LLMs. As the global demand for LLM computation surges, this collaboration positions both firms to lead a revolution in processor infrastructure. We're excited to be a key player in bringing the Bertha chip to life."
About SEMIFIVE
Founded on the principles of innovation, SEMIFIVE is dedicated to providing pioneering SoC design services that transform visionary ideas into complex custom silicon products. Their platforms are designed to be adaptable and are supported by pre-validated IP pools, enabling faster turnaround times while mitigating risk and cost for applications such as data centers and AI-enhanced IoT solutions. Through a strategic partnership with Samsung Foundry, SEMIFIVE is proud to be a leading SAFE™ DSP partner, offering comprehensive assistance for any SoC design challenges.
Frequently Asked Questions
What chip is SEMIFIVE developing with HyperAccel?
SEMIFIVE is partnering with HyperAccel to develop a generative AI chip called Bertha, utilizing 4nm process technology.
When is the mass production of the AI chip expected to start?
The mass production of the Bertha AI chip is targeted to commence in the first quarter of 2026.
What advantages does the LPU offer over traditional GPUs?
The LPU is designed to provide significant performance improvements, offering speed up to double that of conventional GPUs and a much more favorable price/performance ratio.
How is SEMIFIVE positioned in the AI chip market?
SEMIFIVE specializes in creating SoC platforms and ASIC design solutions, focusing heavily on AI chip development and responding to growing market demands.
Who are the key executives involved in this partnership?
The partnership features Joo-Young Kim, CEO of HyperAccel, and Brandon Cho, CEO and co-founder of SEMIFIVE, both of whom see great potential in their collaboration.
