Intel faces intensifying competition from NVIDIA and AMD in datacenter GPU and accelerator market.

Intel's grip on the data center market, where it holds more than 95 percent of market share, is loosening as the company struggles with 10-nanometer manufacturing technology while competitors advance with more efficient process nodes. While Intel remains on 14-nanometer server chips due to manufacturing bugs with its 10-nanometer process, Advanced Micro Devices is preparing to release 7-nanometer chips ahead of Intel's 10-nanometer offerings. Nvidia, meanwhile, has pivoted faster into artificial intelligence, threatening to marginalize Intel's Xeon processors. These challenges come as Intel searches for a new chief executive following Brian Krzanich's dismissal, and as the company attempts to rebuild customer trust following the Meltdown and Spectre vulnerabilities.

Intel's roadmap aims to demonstrate continued competitiveness through annual product updates. The company will introduce Cascade Lake server chips before the end of the year, followed by Cooper Lake—the last 14-nanometer generation—using the same architecture as Intel's initial 10-nanometer designs. Cooper Lake will arrive in 2020 with Ice Lake, allowing customers to upgrade between the two generations. However, AMD's 7-nanometer processors could reach market more than a year before Intel's 10-nanometer chips, and industry analysts expect AMD to capture meaningful data center market share. Intel has acknowledged the risk, acknowledging it could lose as much as 20 percent market share to AMD and other competitors.

Despite manufacturing challenges, Intel's data center business remains highly profitable. Last year, the division reported $21.2 billion in revenues and $10.6 billion in profits, compared to AMD's annual revenue of $6.5 billion. The latest quarter showed data center revenues of $5.5 billion, representing 26 percent growth year-over-year, though results undershot Wall Street estimates and drove Intel shares down more than five percent. The addressable market itself has expanded significantly; Intel's estimate of the data center hardware market grew from $70 billion to $90 billion, with the company's data center sales increasing more than five percent this year.

Recognizing the shift toward artificial intelligence, Intel is aggressively positioning itself in the AI chip market. Last year, Xeon processors for artificial intelligence generated more than $1 billion in sales, driven by enhancements for both training and inferencing tasks. Navin Shenoy, vice president and general manager of the data center business, noted that "The step-function increase in performance led to a meaningful business impact for us." Intel estimates the AI chip market will grow from $2.5 billion to $10 billion by 2022, though the landscape is complicated by major customers including Google and Facebook developing custom chips. To strengthen its position, Intel created a dedicated artificial intelligence unit headed by Naveen Rao, founder and former chief executive of Nervana Systems, and is releasing Spring Lake, a custom neural network processor, in 2019. The company has also developed nGraph, a software tool enabling TensorFlow and other machine learning environments to work with any Intel processor, reducing dependence on hardware alone.