Intel articulates a renewed AI chip strategy after years of competitive setbacks against Nvidia.
Intel's AI chip strategy under CEO Lip-Bu Tan is crystallizing around inference and agentic AI workloads powered by GPUs working in tandem with other processor types, marking a strategic pivot after years of costly missteps in the Nvidia-dominated accelerator market. The company has spent billions of dollars over the past 15 years attempting to build competitive accelerator chips, but has repeatedly failed to gain traction. This pattern of abandonment has created skepticism among channel partners; Dominic Daninger, vice president of engineering at Intel systems integration partner Nor-Tech in Burnsville, Minnesota, told CRN that his company previously invested in products like the Xeon Phi processors—which competed with Nvidia and AMD GPUs—only for Intel to retract from the category years later. "One of the things with Intel we've seen so many times [was] that they were out there a year or two with a particular product line, and if it's not sticking, it's gone," Daninger said, noting that for new solutions to succeed, they would "probably have to have a heck of a lot more performance than anything the competitor would have to make people interested—and a reasonable price."
Intel's current strategy stems from years of development that began under former CEO Brian Krzanich, but the company's accelerator efforts have included missteps with ASICs, most notably the Gaudi chips originally designed for AI training workloads that fell short of modest sales expectations even after Intel pivoted the focus to inference. Anil Nanduri, Intel's vice president of product management and go-to-market for data center AI accelerators, acknowledged the failures during CES 2026 in January, stating: "We did not meet the needs of the frontier AI training market, and we didn't meet the market needs." Under the new strategy, Nanduri said the chipmaker has "a plan and a road map" to meet growing needs for inference and agentic AI workloads "with our CPUs, our GPUs, from the client, to the edge, to the data center," adding: "We really want to go focus on this new world of AI inferencing, start leading with AI agents, and we want to be here to support you in that journey."
When Tan took the helm as Intel's CEO last year, he prioritized AI accelerator chips for data centers and openly stated he was not "happy with our current position" in the market, vowing to learn "past mistakes and work towards a competitive system," though cautioning: "It won't happen overnight, but I know we can get there." In October, Intel revealed a 160-GB, energy-efficient data center GPU code-named "Crescent Island," positioned as part of a new road map with an annual release cadence mirroring pushes by Nvidia and AMD. The company expects to start sampling Crescent Island with customers in the second half of 2026, with a launch to follow. The GPU features Intel's Xe3P microarchitecture optimized for performance-per-watt, 160 GB of LPDDR5X memory, and support for a broad range of data types, all designed as "power- and cost-optimized" for inference workloads running on air-cooled enterprise servers.
The Xe3P microarchitecture builds on the Xe GPU architecture Intel first disclosed in 2019 with its "Ponte Vecchio" GPU for high-performance computing and AI workloads in data centers. Sachin Katti, Intel's AI leader at the time of Crescent Island's September briefing, outlined a vision centered on heterogeneous computing—systems using multiple distinct types of processors to boost performance and efficiency—based on open systems and software architecture delivering "right-sized" and "right-priced" compute for future agentic AI workloads. Katti, whom Tan appointed a year prior to lead Intel's AI strategy, unexpectedly departed in November for a position at OpenAI.