Intel stated that it will stick to the GPU business and Chen Liwu reiterated that it will continue to invest in building future product lines.

American chip giant Intel's direction in the GPU business has always been confusing. Whether it is the consumer market or the data center field, it has caused doubts about its strategic determination. However, in a series of recent public statements, Intel's current CEO Lip-Bu Tan clearly denied the statement of "exiting the GPU market" and instead released a signal to strengthen its self-research and production capabilities.

Since Chen Liwu took over, Intel's overall roadmap in the field of AI and accelerated computing has always been considered "not clear enough" by the outside world. In terms of accelerator products, the most clearly disclosed project at present is Crescent Island for reasoning tasks, while progress around the higher-end Jaguar Shores platform is still vague. On the consumer side, Intel will launch the first batch of Battlemage graphics cards at the end of 2024, and will supplement the Pro version in the second half of 2025. However, these products are regarded more as tests to "show potential" rather than as a complete, long-term product rhythm commitment.

At the recent Cisco AI Summit, Chen Liwu was asked whether Intel would "design CPUs, GPUs, build its own wafer factories to produce its own chips, and also OEM for other manufacturers, and continue to cooperate with other GPU companies." He gave a clear and affirmative answer. He emphasized that Intel will not only continue to build its own GPU product line, but also plans to produce these chips on a larger scale through its foundry business. He also revealed that GPU architect Eric Demers, recently poached from Qualcomm, is part of Intel's overall strategy to return to the GPU track. This appointment highlights the company's redeployment on graphics and AI routes.

Chen Liwu said at the scene that he had "just hired a chief GPU architect, and he is very good." He also said that he had put in a lot of effort to win this executive to join, and his words revealed his emphasis and confidence in the GPU business. The outside world generally regards this move as a signal for Intel to regroup its "core capabilities" in the fields of graphics and AI acceleration.

At last year's Intel Tech Tour event, Intel announced that it would shift to an "annual product rhythm" and disclosed that it was advancing the research and development of Crescent Island and a new generation of Jaguar Shores series, aiming directly at the expanding total serviceable market (TAM) of the AI ​​accelerator market. However, compared to Nvidia, which already dominates the field of high-performance GPUs and AI acceleration, and AMD, which continues to make efforts in data centers and game graphics, Intel needs to regain its footing not only in terms of competitiveness in performance and energy efficiency (TCO), but also in large-scale deployment and complete infrastructure solutions. "Not a single link can be lost."

In terms of consumer GPUs, Intel has not been without bright spots in recent years. The Xe3 "Celestial" integrated graphics card launched with the Panther Lake platform has demonstrated a performance level that can compete with AMD Strix Halo in some test scenarios due to its Arc B390 performance. At the same time, in the independent graphics field, the market is still waiting for the higher-end Arc B770 in the Battlemage series, which is said to be expected to be unveiled at this year’s Taipei International Computer Show (Computex). However, some plans for the high-end Xe2 Arc Battlemage BMG-G31 were reportedly canceled, which also made the outside world pay more attention to Intel's pace and choices in high-end independent graphics.

Based on the current signs, it can be said that Intel's GPU business has not only not "died down", but has been reorganized at the management and organizational structure levels in an attempt to create a clearer internal technology route and product rhythm. However, in the GPU and AI acceleration market where competition is extremely fierce and the update cycle is accelerating, if Intel wants to truly realize the transformation from "proving its potential" to "proving long-term sustainability", it still depends on the overall performance of subsequent generations of products in terms of performance, ecology, cost control and delivery capabilities.