Semifive Signed $52 Million AI Chip Contract
The South Korean company secured its largest-ever deal with a U.S.-based artificial intelligence firm.
Updated on Sept. 30, 2026 in Semiconductors
Live Poll
Do you believe custom-designed chips will significantly improve the AI services you use?
Semifive has finalized a $52 million chip development contract with a U.S. AI company, marking the largest deal in the firm's history. This agreement represents 60 percent of the company's $88 million in new order revenue from the first half of 2026.
Why it matters
The project represents Semifive's first Spec Hand-off engagement in North America, a model where the developer leads design and testing processes from an early stage. It aims to satisfy the surging demand for custom silicon optimized for specific customer workloads.
The new accelerator design utilizes LPDDR6 memory technology and the PCIe Gen5 interface. These specifications are part of a development process that includes comprehensive software development, packaging, and testing.
The players
Semifive
A South Korea-based design solution company that specializes in custom silicon and semiconductor development.
The details
Under the Spec Hand-off model, Semifive receives performance requirements from the client to oversee the design lifecycle rather than operating under traditional turnkey arrangements. This specific accelerator targets high-performance needs for hyperscalers and cloud service providers.
Timeline
The contract was reported as part of H1 2026 performance results.
The chip tape-out is expected in the first half of 2027.
Mass production of the new accelerator is scheduled for 2028.
The Tech Race
The transition to LPDDR6 memory and PCIe Gen5 interfaces reflects a broader industry shift toward high-bandwidth, energy-efficient architectures required for modern AI workloads. Companies are increasingly moving away from off-the-shelf components toward specialized, bespoke silicon to maintain a competitive edge in cloud infrastructure.
As hyperscalers and cloud providers adopt these custom accelerators, users may experience improved efficiency and performance in cloud-based AI services. The transition to advanced standards like LPDDR6 could lead to faster data processing speeds and lower power consumption for large-scale data center operations.
The takeaway
The move toward early-stage design hand-offs allows tech firms to better align custom hardware with specialized software needs. Developers should monitor these shifts as custom silicon continues to replace generalized components in high-demand cloud environments.
Further reading
Learn more about the latest innovations in the chip industry at /Semiconductors.
Source note: This article includes information reported by Datacenterdynamics.
Live Poll
Do you believe custom-designed chips will significantly improve the AI services you use?
