The tech world is abuzz with the possibility of Nvidia entrusting the production of its gaming GPUs to an Intel foundry. If this noteworthy alliance materializes, it could mark a significant breakthrough for Intel's foundry business. Timothy Arcuri, an analyst at UBS, observes that should Intel manage to secure orders from Nvidia, a leading fabless chip design firm globally, it would represent a pivotal moment for Intel's foundry ambitions. Simultaneously, other prominent players in the artificial intelligence chip sector, like Broadcom and AMD, have shown substantial interest in Intel's 18A process technology.

According to Arcuri, Intel is earnestly pursuing foundry orders from Nvidia and Broadcom, with Nvidia seemingly favoring Intel's inclusion in its supply chain potentially as an additional supplier. In contrast, the involvement of Broadcom and AMD remains more ambiguous, yet their interest underscores the potential of Intel's 18A technology. Arcuri notes in his report that Intel is striving to cement collaborations with Nvidia or Broadcom, thereby enhancing its foundry's market stature.

The crux of the partnership is Intel's 18A process, which boasts a 1.8-nanometer scale that integrates the RibbonFET transistor architecture and PowerVia backside power technology. These innovations aim to enhance chip performance and energy efficiency. Intel plans to commence mass production of next-gen products built on the 18A process in the latter half of 2025. These include the Panther Lake processors for consumers and the Xeon 7 "Clearwater Forest" CPUs for data centers. Additionally, an enhanced 18A-P variant promises superior performance compared to the standard 18A process, offering crucial benefits for external clients prioritizing performance and energy efficiency.

For Intel, the 18A process transcends a mere technological upgrade; it symbolizes a cornerstone in its evolution as a prominent system foundry, propelled by the IDM 2.0 strategy. In August 2024, Intel celebrated Panther Lake and Clearwater Forest's successful delivery with operational systems, signaling strong progress in the 18A process's development. Expected timelines suggest initial external customer design tape-outs in the first half of 2025, advancing to volume production by early 2026. Though slightly trailing TSMC's planned 2nm production in late 2025, Intel asserts that the 18A process could rival or exceed the performance metrics of TSMC's 2nm and 3nm technologies.

Nvidia's interest is a calculated maneuver. Commanding the GPU market, Nvidia's gaming GPUs, such as the GeForce RTX series, demand robust manufacturing processes to deliver high performance for gamers while managing power consumption for optimal heat dissipation and costs. Historically reliant on TSMC and Samsung, Nvidia is now keen on diversifying its supply chain. If Intel's 18A process aligns with Nvidia's requirements, it could alleviate TSMC's capacity constraints while securing consistent external orders for Intel. Projections indicate Nvidia's 2024 GPU shipments might surpass 30 million units. Shifting even a portion of these to Intel's production could be mutually beneficial.

Broadcom's situation, however, presents more complexity. Renowned for networking chips and application-specific integrated circuits (ASICs), Broadcom's prior tests with Intel didn't fully reach mass production standards. The reception of subpar test wafers for the 18A process in late 2024 has made Broadcom cautious. Nevertheless, Intel maintains that both the performance and yield of 18A have met expectations, presenting the issues as typical optimization adjustments. Meanwhile, AMD's involvement bears watching. As Intel's direct competitor using TSMC for CPUs and GPUs, AMD's successful 18A testing could foster Intel's diversified foundry expansion.

Internally, Intel is diligently preparing for this pivotal juncture. New CEO Lip-Bu Tan is set to make his inaugural appearance at Intel's Vision Conference on March 31 in Las Vegas, an event poised to shed light on his strategies for chip design and foundry operations. Predecessor Pat Kissinger ardently championed the 18A process and foundry strategy, and the continuance of these initiatives remains anticipated under Tan's leadership. Amid high expectations, the industry is keen to see if Tan can steer Intel to a formidable standing within the fiercely contested foundry market.

In 2024, Intel's foundry division witnessed a 13 percent revenue decline year-over-year to $4.5 billion, with an operating loss reaching $2.26 billion, nearly doubling from 2023. Intel forecasts its foundry revenue could swell to $16.47 billion by 2025, with hopes to reach breakeven by 2027. To support anticipated 18A production volumes, Intel is augmenting its Fab 52 facility's capacity in Arizona, fortified by a $1.1 billion subsidy courtesy of the U.S. Department of Commerce, part of the broader $7.86 billion CHIPS Act funding package.

Ultimately, this potential collaboration signifies more than just Nvidia and Intel's futures; it embodies significant shifts within the global chip supply chain. The question of whether Intel's 18A process can deliver as promised and become a trusted option for Nvidia, Broadcom, and other industry players will find its answer come 2025.