Intel’s Struggle to Keep Up with the AI Revolution and the Road Ahead

Intel (NASDAQ: INTC) has faced significant challenges in the artificial intelligence (AI) market, a space where competitors like Nvidia and AMD have thrived. While Nvidia dominates the AI accelerator chip market, and AMD has made considerable strides with its Instinct data center GPUs, Intel’s efforts have faltered. Despite its goal of $500 million in AI accelerator sales for 2024, the company has yet to hit that target, underperforming in comparison to its competitors.
A Missed Opportunity
Intel’s struggles in AI are rooted in decisions made long ago. In 2008, the company was developing a discrete GPU project, codenamed Larabee, which could have been a strong contender for the AI market. However, Larabee was canceled, setting Intel up for its current struggles in the face of AI’s explosive growth and demand for specialized GPU accelerators.
Today, Intel’s AI accelerators come from its 2019 acquisition of Habana Labs, a company focused on AI chips. The Gaudi family of chips, while solid in terms of performance, has not taken off due to a lack of developer familiarity and an underdeveloped software ecosystem. This lack of software maturity continues to hinder Intel’s progress in both the GPU and AI sectors. For example, Intel’s foray into discrete GPUs in 2023 with its Arc Alchemist gaming cards was marred by buggy software and poor sales, though its second-generation cards have seen improvements.
Looking Ahead: Challenges in 2025
Intel’s path forward is far from clear. The company’s AI and GPU plans are complicated, with the Gaudi chips and the Max family of GPUs, which power the Aurora supercomputer, slated for discontinuation. Instead, Intel is focused on its next-generation Falcon Shores chips, which are expected to combine features from both traditional GPUs and Gaudi’s AI architecture. While Falcon Shores is set to launch by the end of 2025, the uncertain trajectory of Intel’s AI strategy makes it difficult to predict significant progress in the near term.
Although Gaudi 3 sales are expected to improve this year, software challenges will likely prevent the chip from reaching the same success as Nvidia or AMD’s offerings in the data center market. One bright spot for Intel, however, is its deal with IBM to integrate Gaudi 3 chips into IBM’s watsonx AI platform, which could help boost sales this year.
A Long-Term Strategy: Betting on AI Foundry Business
Intel’s future in AI may ultimately lie not in developing its own AI accelerators, but in leveraging its decades of manufacturing expertise. The company is making significant investments to build a foundry business capable of manufacturing chips for other companies, including some of the biggest names in tech. Intel’s upcoming 18A process, set to go into production next year, will play a key role in this strategy. Notable customers such as Microsoft and Amazon have already signed on to use Intel’s 18A process for their own chips, including an AI fabric chip for Amazon.
If Intel can prove the success of its 18A process with these early partnerships, it could position itself as a major player in manufacturing AI-related chips, especially in the years following 2025. This foundry business could eventually become a larger and more profitable segment of Intel’s business than its own AI accelerator efforts.