Intel's 18A process has recently become a hot topic in tech circles. On one side, there's a huge curiosity about Intel's progress, as TSMC's dominance in the semiconductor industry tends to stifle competition. On the other hand, Intel's seasoned enterprise observers understand the potential breakthrough that the 18A represents. Over the past six months, NVIDIA, Broadcom, Faraday Technology, along with several ASIC customers, have expressed significant interest. Recent updates from the supply chain have highlighted a breakthrough in the advanced 1.8-nanometer process, with chip sampling results gaining favorable industry attention. Intel appears poised to revitalize its foundry services (IFS) with the 18A process.

In the past few years, Intel ambitiously launched a "four years and five nodes" strategy aiming to catch up in process technology, but the feedback from the market was lukewarm, and the foundry business struggled to gain traction. However, the introduction of the 18A process has transformed the scenario. Intel plans to commence mass production of the 18A process in the latter half of 2025, with initial products including Panther Lake processors for mobile devices and Clearwater Forest chips for servers. These chips were successfully powered and running operating systems smoothly as of last year. Intel also announced that the first chip designs for external clients could begin fabrication by mid-2025.

A prominent feature of the 18A process is its incorporation of two pioneering technologies: RibbonFET and PowerVia. RibbonFET, a gate-all-around transistor, enhances current control via a nanosheet structure, thereby shrinking transistor size, minimizing leakage, and elevating the energy efficiency of high-density chips. PowerVia involves relocating power lines to the wafer's backside, which not only opens up primary side real estate for signal interconnects but also lessens resistance. This architectural innovation translates into a 5%-10% density increase and boosts performance by up to 4%. Compared to Intel's "3 process," the 18A showcases approximately 30% higher transistor density and a 15% improvement in power-performance ratio. Remarkably, its SRAM density matches TSMC's 2nm N2 process and even slightly excels in power-performance balance.

There is a tangible interest in Intel's 18A from external clients. NVIDIA and Broadcom are actively evaluating ASIC samples made with this process, with initial results showing a promising outlook. ASIC providers, like SmartPlanet, have received early samples and reported positive feedback. Additionally, collaborations with companies like IBM and Arm are underway to ensure the 18A process complies with industry standards. Analysts indicate that major players like NVIDIA are keen to diversify their supply chains and lessen reliance on TSMC, with Intel's U.S.-based foundry facilities offering a strategic alternative. This trend notably aligns with the climbing demand for AI chips, wherein Intel's 18A process stands out for its performance and supply chain advantages.

Interestingly, Intel aims to have 70% of its computing modules lever the 18A process. Nonetheless, owing to production limits and yield concerns, some high-end products, such as the next-generation Nova Lake desktop processors, might be outsourced partially to TSMC's 2nm process. Yet, Intel remains bullish on 18A, with CEO Lip-Bu Tan consistently stating that high-volume manufacturing (HVM) will peak by the end of 2025, potentially attracting more premium clients.

Reflecting on past hurdles, the journey to 18A's mass production hasn't been smooth. Reports previously indicated that yields were a meager 20%-30%, far from the over-70% standard necessary for mass production, primarily due to the complexities arising from RibbonFET and PowerVia. Nevertheless, Intel countered that Panther Lake's yield surpasses the same-period metrics of Meteor Lake, maintaining their mass production timeline. According to market analyst Gartner, TSMC will capture 68% of the global AI chip foundry market by 2024, with Intel only holding a fractional grip at 5%. To secure a competitive edge, Intel must ensure 18A's yield and performance live up to expectations.

Intel plans to showcase the prowess of 18A at the VLSI Symposium in 2025, demonstrating that at a 1.1V voltage, the process's ARM cores increase performance by 25% while cutting power consumption by 36%. Even at a lower 0.75V, performance climbs 18%, with a 38% reduction in power use. Compared to "Intel 3," the 18A process reduces the area to 0.72x while enhancing transistor density, underscoring its competitive stature in client and data center arenas.

When compared to TSMC's N2 process, Intel's 18A holds distinct advantages. Although TSMC's N2 leads in high-density standard cell transistor density reaching 313 MTr/mm² against 18A's 238 MTr/mm², and it touts a smaller SRAM cell size (0.0175 µm² vs. 0.021 µm²), 18A shines in the realms of performance and power efficiency thanks to PowerVia. TSMC anticipates mass-producing N2 by the end of 2025, with consumer releases set for mid-2026, marginally trailing 18A. Samsung's 2nm process (SF2) is also set for 2025 mass production, but with yields currently around 40%, its market potency remains unproven.

Intel’s 18A initiative is pivotal for rejuvenating its foundry business. With Panther Lake and Clearwater Forest nearing mass production, combined with vigorous customer involvement from NVIDIA and Broadcom, 18A is set to carve new market territories for Intel. Over the coming years, Intel is poised to introduce the 14A process aiming to enhance performance and density, persistently contesting TSMC’s market lead. In this race for semiconductor supremacy, tech enthusiasts eagerly await to see if Intel's bold venture with 18A will redefine industry standards.