» Current | 2021 | 2020 | 2019 | Subscribe

Linley Newsletter

Year in Review: Data-Center Silicon Diversifies

December 24, 2020

Author: Bob Wheeler

Even as pandemic-related macroeconomic concerns persisted, public-cloud data centers drove through 2020, propping up semiconductor demand. First-half sales benefited from a return to normal consumption following 2019’s inventory burn, whereas second-half growth slowed as customers became more conservative in their capital expenditures. In an unpredictable year, data-center sales fared better than most other chip segments.

Well-chronicled execution problems marred Intel’s 2020, but sales to data-center customers remained robust. The company continued its xPU strategy of diversifying its portfolio to include GPUs, FPGAs, and AI accelerators. The Xeon line hit a fork in the road as Intel repositioned Cooper Lake for high-end systems while continuing Cascade Lake for mainstream 2S servers. Meanwhile, AMD ramped second-generation Epyc shipments, continuing its slow but steady share gains. It also introduced a new GPU intended for high-performance computing (HPC), and it announced a deal to acquire FPGA leader Xilinx.

Nvidia made waves with new products and acquisitions. It debuted the impressive 7nm Ampere GPU for AI and HPC, and it closed the Mellanox acquisition, adding about $2 billion in annual revenue to its data-center business. Nvidia’s plan to acquire Arm met with skepticism, and regulatory hurdles remain before that deal’s closing. Regardless, with its fortunes rising and Intel’s declining, Nvidia’s market value soared well beyond Intel’s thanks to a push from Wall Street.

Startups abound in the deep-learning-accelerator (DLA) market, but Qualcomm also joined the fray by announcing its first data-center inference accelerator. Shipping their first products, startups Groq and Tenstorrent entered the inference segment, whereas Graphcore unveiled its second-generation training accelerator. At the same time, leading cloud-service providers (CSPs) continue to develop internal solutions, including Google’s TPUv4, Baidu’s Kunlun, and Amazon’s forthcoming Trainium.

Subscribers can view the full article in the Microprocessor Report.

Subscribe to the Microprocessor Report and always get the full story!

Purchase the full article

Free Newsletter

Linley Newsletter
Analysis of new developments in microprocessors and other semiconductor products
Subscribe to our Newsletter »


Linley Spring Processor Conference 2021
Coming the week of April 19, 2021
Virtual Event
More Events »