Intel Spares No Effort in Building Artificial Intelligence Hardware

2023-07-14

NVIDIA has profited from the AI boom, while Intel has been forced to compete for the second position. However, compared to its competitor AMD, Intel finds itself in a predicament. Earlier this year, it shut down its server business and recently closed its mini PC business. According to reports, Intel has stopped investing in its mini PC product line, Next Unit of Computing (NUC). These small NUC PCs are widely used in the enterprise sector, but now Intel is handing them over to manufacturing partners. This move aligns with Intel's efforts to focus its resources on the booming AI computing market. Since acquiring AI chip manufacturer Habana Labs for $2 billion in 2019, Intel has been striving to enter the AI computing market. As NVIDIA's software barriers become stronger, Intel's market share is shrinking. Will doubling down on AI hardware efforts bring it the momentum it needs? Facing off against giants Exiting the mini PC market is primarily due to the company's reluctance to compete with its original equipment manufacturers, but it also comes at a time when the company is cutting costs across the board. Intel also plans to cut $10 billion in costs by 2023 as part of its IDM 2.0 strategy implemented in 2021. Even the server business has reportedly been sold off to focus on its new strategy. This strategy seems to be backfiring as Intel's stock price has plummeted, with its market value dropping 69% from its 2021 peak. While its competitors AMD and NVIDIA have experienced tremendous growth during the same period due to the flourishing of AI applications, Intel's AI products have fallen behind. Even comments from Intel's management echo this sentiment. Habana's Chief Business Officer Eitan Medina stated in 2020, "We have to realize that we start from zero, and NVIDIA is 100%. This tough battle or process of gaining market share has to go through convincing the end developers to give it a try." Despite the performance improvement, Habana chips still face competition from NVIDIA's Grace Hopper chip. This new product line thoroughly outperforms Intel's AI accelerator, and ironically, it uses Intel's Xeon CPU to achieve this. However, from a hardware perspective, Habana remains competitive. However, in a crucial aspect of AI computing, software, Intel falls short. Intel's One API performs well on basic AI workloads, but NVIDIA's software stack serves nine different verticals, from enterprise AI to Omniverse. The need for strategic adjustments As part of its cost-cutting measures, Intel is also turning to a disaggregated CPU architecture. This design first appears in the long-delayed Meteor Lake chip series, where Intel adopts chip modular design, which not only reduces manufacturing costs but also better competes with its rivals. Leveraging Habana's AI accelerator technology, Intel is gradually gaining a foothold in the AI computing market on devices. While Gaudi competes with NVIDIA's top-tier Grace Hopper and H100 chips, the new Meteor Lake chips will compete with NVIDIA's GPUs in laptops. Intel's integrated solution will provide a low-power alternative for everyday laptops, a strategy that has already seen some success in Apple's M-series chips. At first glance, this market seems to be a better field to compete with NVIDIA, especially in building software stacks. Take CUDA, for example, which was built through continuous feedback from developers using NVIDIA chips for acceleration. While Intel doesn't have such a long time span, delivering AI chips on laptops can greatly accelerate the creation of a stable software stack. The third-generation Gaudi chip is about to be released and will compete with products from other specialized chip manufacturers such as Graphcore, SambaNova, Tenstorrent, and Cerebras. While NVIDIA continues to enjoy its market dominance thanks to its powerful software stack, as the manufacturing cost of AI accelerators becomes lower, Intel may gradually gain momentum and begin to offer end-to-end AI solutions, including CPUs, GPUs, and AI accelerators. Combined with its software stack, this could make Intel the third major player in the market alongside NVIDIA and AMD. Source: https://analyticsindiamag.com/intel-is-giving-ai-hardware-everything-its-got/

RECENT AI NEWS

RECENT AI TOOLS