Socionext Bets on Intel 18A-P as Japanese Chip Designer Diversifies Away From TSMC Monopoly
YOKOHAMA, Japan — Socionext, one of Japan's largest custom chip designers, has committed its next-generation data-center and AI silicon to Intel's 18A-P process, marking the biggest external customer win yet for Intel's foundry revival. The move, announced August 18, ends Socionext's exclusive reliance on TSMC for leading-edge production and signals a shift in how fabless companies hedge their manufacturing bets.
The Yokohama-based company, which designs system-on-chips for automotive, networking, and data-center markets, said it will use 18A-P — a performance-enhanced derivative of Intel's 18A node — to develop a high-performance compute chiplet targeting AI acceleration, high-performance computing, and what it calls "Physical AI" workloads. Socionext CTO Rajinder Cheema described the process as "an important step in advanced process technology, combining innovations in transistor architecture and power delivery."
A Foundry Strategy Shift
For years, Socionext's leading-edge roadmap ran through TSMC. In July, the company told Nikkei it was working with customers planning to adopt TSMC's A14 process for data-center AI SoCs. The Intel 18A-P addition expands that portfolio rather than replacing it, giving Socionext customers flexibility across two foundry ecosystems.
"By adding Intel's 18A-P to its technology portfolio, Socionext is expanding its foundry options and giving customers greater flexibility in chip development," Nikkei reported. The Japanese outlet noted the move marks a shift for a company that has "so far relied on TSMC for leading-edge production."
Intel positions 18A-P as an incremental upgrade over the base 18A node. At the VLSI Symposium in June, the company claimed up to 9 percent higher performance or 18 percent lower power consumption compared to 18A, with at least 20 percent better heat resistance. Crucially, 18A-P maintains full IP and design compatibility with 18A, meaning customers can migrate designs without a full re-spin.
Intel's Foundry Momentum Builds
The Socionext win follows Intel's June announcement that 18A-P had entered risk production — an early manufacturing phase where output data indicates the node will meet customer requirements upon final qualification. CEO Lip-Bu Tan, speaking on Intel's Q2 FY2026 earnings call in July, said the company had cut the cost of its primary Panther Lake SKU by roughly 50 percent year-to-date and targeted another 20 percent reduction by year-end.
"We also began risk production of 18A-P, providing additional performance and power advantages while maintaining IP and design compatibility with Intel 18A, positioning 18A-P as a competitive node for external customers," Tan said. He added that defect density and transistor performance on the follow-on 14A node were "all outpacing 18A development," with PDK 0.5 complete and PDK 0.9 on track for October delivery.
Intel's foundry revenue reached $5.8 billion in Q2, up 6 percent sequentially on higher fab volumes driven by 18A growth. External foundry revenue was $293 million — still a sliver of the total, but growing. The company said 18A output came in approximately 25 percent above target and up more than 50 percent quarter-over-quarter.
The Apple Question Looms
Socionext is not the only name linked to 18A-P. Reports from Wccftech and CNBC have tied the node to Apple's next-generation M-series chips, with speculation that Apple could tap Intel's 18A-P for M7 processors as early as 2027. Intel has not confirmed any Apple deal, but CEO Lip-Bu Tan told CNBC in May he expected commitments from multiple foundry customers in the second half of 2026.
Analyst Neil Shah of Counterpoint Research said yield rate remains the critical metric. "If they can commit to more than 90 percent yield rate in the first month, I think they can attract a few more customers," he told CNBC in June. Intel's 18A node has been in high-volume production for internal products since December at its Arizona Fab 52, but securing major external customers has been the missing piece.
Advanced Packaging as a Parallel Play
While process technology grabs headlines, Intel's EMIB (embedded multi-die interconnect bridge) packaging has emerged as a parallel competitive vector. TSMC's CoWoS advanced packaging remains capacity-constrained, creating what Shah called "a big opportunity right now, very low hanging opportunity for Intel." Intel's Q2 call revealed a growing EMIB-T backlog with yield and reliability hitting targets, aiming for high-volume customer ramps in 2027.
Socionext's press release highlighted "the combination of Socionext's Solution SoC development methodology and Intel Foundry's advanced process and packaging technologies" — a nod to the packaging angle. Intel Foundry VP John Zucker said the partnership "enriches the ecosystem for enabling product development for AI acceleration, high-performance compute, and Physical AI."
Roadmap to 14A and Beyond
Intel's 14A node remains on track for risk production of internal products in the second half of 2027, with high-volume ramp targeted for 2028. The company committed to the 2028 high-volume milestone in Q2, stepping up CapEx for 14A tooling. Tan said customer engagements for 14A were "increasing" and he was "increasingly confident that 14A will be a highly competitive process offering across key vectors of performance, power, density, cost, and schedule."
Capital expenditures for 2026 now exceed $20 billion, up substantially from earlier guidance, with the "vast majority" spent across Intel's U.S. network. The company said total U.S. tool and space spending from 2021 through 2026 approaches $100 billion.
What This Means for the Foundry Outlook
TSMC still commands more than 70 percent of global foundry revenue. Samsung accounts for roughly 7 percent. Intel's foundry business, while growing, remains a fraction of that scale. But the Socionext win — alongside earlier announcements from Hitachi for silicon quantum chips and reports of Intel building up to 90 percent of Nova Lake compute tiles on 18A — suggests the external customer pipeline is finally moving.
For fabless companies like Socionext, the calculation is straightforward: TSMC's advanced nodes are sold out through 2027, and having a qualified second source at the leading edge reduces supply-chain risk. Intel's 18A-P, with its 18A compatibility and performance uplift, offers a pragmatic alternative without requiring a full redesign.
The broader question is whether Intel can sustain yield improvements and cost reductions at scale. Tan's earnings-call comment that "yields continue to track ahead of expectations" and that the team "keeps raising the bar on the internal targets" will be tested as external tape-outs begin. Socionext's first 18A-P chiplet is expected to tape out in the coming quarters, with production timing dependent on Intel's qualification schedule.
For now, the Japanese designer has made its bet. The rest of the fabless ecosystem is watching closely.
Sources: TrendForce — Japan's Socionext Adds Intel's 18A-P to Foundry Mix Alongside TSMC, Socionext — Broadens SoC Roadmap with Intel 18A-P Process Technology, CNBC — Intel Begins Production of Most-Advanced Chip, Inching Closer to Possible Apple Deal, Yahoo Finance — INTEL CORP Q2 FY2026 Earnings Call Transcript
Internal links: AI, Cloud & Edge Computing
Keywords: Socionext, Intel Foundry, 18A-P, 18A, 14A, TSMC, semiconductor foundry, AI chips, chiplet, advanced packaging, EMIB, Lip-Bu Tan, Rajinder Cheema