Technology · Dev
Socionext Picks Intel 18A-P Node for Custom Compute Chiplets
Japanese SoC designer selects Intel Foundry's advanced process to power next-generation workloads across data centers, edge systems, and AI applications

KEY TAKEAWAYS
- ·Socionext will manufacture its next-generation custom compute chiplets using Intel Foundry's 18A-P process node for data center, edge, and AI applications.
- ·The design win provides Intel Foundry a foothold with a Japanese SoC designer as it competes against TSMC and Samsung for third-party customers in Asia.
- ·Success depends on Intel delivering competitive yield, cost, and time-to-market performance as 18A enters volume production later this year.
A Strategic Foundry Choice
Socionext, the custom System-on-Chip designer headquartered in Japan, has committed to Intel Foundry's 18A-P process node for its next-generation silicon development. The decision positions Intel's advanced manufacturing technology as a cornerstone for Socionext's upcoming portfolio targeting data centers, edge computing environments, and artificial intelligence workloads.
The partnership represents a tangible design win for Intel Foundry as it competes to attract customers beyond its internal chip divisions. Intel's 18A family of nodes, built on RibbonFET gate-all-around transistor architecture and PowerVia backside power delivery, aims to reclaim process leadership lost to TSMC and Samsung over the past half-decade.
Chiplet Architecture for Performance Flexibility
Socionext specializes in custom SoC solutions rather than high-volume commodity chips, serving clients in imaging, networking, automotive, and compute-intensive applications. The company's use of chiplet designs allows it to mix and match functional blocks, optimizing performance and power for specific customer requirements without redesigning entire monolithic dies.
By selecting 18A-P, Socionext gains access to a process node Intel positions as production-ready ahead of the full 18A rollout. The "P" designation typically signals a variant optimized for specific power or performance trade-offs, though Intel Foundry has not disclosed granular specifications for this sub-node publicly.
Intel's Foundry Ambitions in Asia
The announcement arrives as Intel Foundry works to secure third-party customers across Asia, a region where TSMC commands dominant market share in advanced logic manufacturing. Socionext's decision offers Intel a foothold with a Japanese design house that counts major electronics and automotive firms among its clients.
Intel has invested heavily in expanding foundry capacity in the United States and Europe, while also maintaining packaging and test operations in Malaysia and Vietnam. Winning customers like Socionext helps justify the capital expenditure required to bring 18A into high-volume manufacturing, expected later this year.
Socionext's roadmap includes chips for AI inference at the edge, where lower latency and power efficiency matter more than raw compute density. The 18A-P node's combination of transistor performance and power delivery innovation aligns with those design priorities, particularly for workloads that demand frequent data movement between compute cores and memory.
Competitive Landscape for Advanced Nodes
TSMC's N3 and N2 families remain the default choice for most high-performance compute and mobile SoC designs, supported by mature ecosystems of design tools and IP libraries. Samsung Foundry's 3nm and 2nm nodes compete primarily in mobile processors and select data center applications.
Intel Foundry must differentiate on more than process geometry alone. Its pitch emphasizes domestic supply chain security for U.S. and European customers, tighter integration with Intel's own design expertise, and the technical benefits of backside power delivery, which reduces voltage droop and improves clock frequency headroom.
Socionext's engagement suggests that at least some designers view Intel's technology as credible for performance-sensitive applications. Whether the partnership translates into volume production depends on yield ramp, cost competitiveness, and the ability to meet Socionext's time-to-market windows.
Implications for the Foundry Market
Intel's strategy hinges on converting design wins into sustained revenue. The company separated its foundry operations into a distinct business unit to provide transparency and attract external customers wary of competing with Intel's own product divisions.
Socionext's commitment adds to a growing list of announced 18A customers, including other chiplet-focused designers and government-backed projects under the U.S. CHIPS Act. The test for Intel will be execution: delivering on promised performance, power, and area metrics while ramping yield curves that make 18A economically viable at scale.
For Socionext, the partnership diversifies supply beyond Taiwan-based fabs and positions the company to serve customers seeking geographic risk mitigation. As data center operators and edge infrastructure providers scrutinize supply chain resilience, access to advanced nodes outside Taiwan becomes a competitive advantage.
The semiconductor industry is watching whether Intel can rebuild the foundry credibility it once held before falling behind in process technology. Deals like this one with Socionext offer early evidence, but sustained success will require flawless execution through 2027 and beyond.
RELATED STORIES
Spot something wrong? Email editor@briefasia.com. We log every correction publicly.



