Nvidia dethroned Intel this week as the top U.S. chipmaker. Nvidia's stock price soared to $424 per share Friday morning, pushing its market cap to $258.7 billion, outpacing Intel at nearly $250.4 billion.

Nvidia has enjoyed concurrent successes in recent years driven by a pivot to artificial intelligence (AI) and high-performance computing. While Intel has also enjoyed steady growth over the past five years and is by no means performing poorly, the chipmaker has continued to battle product delays, security vulnerabilities, and setbacks to its core product line.

Intel is also increasingly under pressure not just from longtime rivals like AMD, but from chipmakers like Marvell and Ampere, all of which have launched chips directly challenging Intel's Xeon Scalable processors. What's more, Amazon and Apple, while still very much reliant on Intel technology, have launched their own Arm-based chip lines. Apple plans to phase out Intel-based Macs during the next two years.

Nvidia's AI HPC Surge

Perhaps best recognized as a GPU manufacturer, Nvidia's innovations of late have been driven less by consumer demand, but rather high-performance computing and an emphasis on AI.

Eight of the top 10 supercomputers, according to Top500, are now powered in part by Nvidia GPUs to accelerate AI, machine learning (ML), and data analytics workloads.

The company's Volta GPU's have seen widespread deployment across public clouds in recent years, and the company's latest generation of GPUs, based on the company's new Ampere architecture, are even more powerful.

The company's A100 GPU, launched in mid-May, boasts 250% higher performance in 16-bit floating-point calculations compared to the Volta-based V100 GPUs, and nearly 500% higher performance when taking advantage of sparsity. That performance advantage balloons in 32-bit floating-point equations where the A100 reportedly offers a 1,000% advantage over the older chip.

These new chips were used to build the seventh most powerful supercomputing cluster, called Selene, in less than a month.

Intel's AI Ambitions, Floundering Fab

Nvidia isn't the only vendor chasing AI glory. Intel is aggressively pursuing AI technologies and integrating AI functionality into its chips.

Intel last month launched its first AI-optimized field-programmable gate array (FPGA), which it claims offers up to 9.5 times the performance of rival Nvidia's previous-generation V100 chips.

But Intel's AI ambitions haven't been without pitfalls. The company abandoned its Nervana neural networking processor line after less than a year in favor of Habana Lab's technology which the company acquired shortly before last Christmas.

Since then, the chipmaker has begun sampling the first generation of inference and training processors to customers. It should, however, be noted that these chips were designed and announced prior to Intel’s acquisition.

Meanwhile, Intel continues to pump out chips based on its aging 14-nanometer process. This includes the company's third-generation Xeon Scalable chips launched last month. Rival chipmakers including AMD, Marvell, Ampere, and Amazon already have 7-nanometer chips on the market with plans for 5-nanometer chips in the near future. By contrast, Intel aims to introduce 10-nanometer scalable chips sometime in 2021.

Intel Faces Pressure From All Sides

Nvidia's market victory comes as Intel faces increased pressure from rival chipmakers, particularly in its lucrative data center group which, according to CEO Bob Swan, now accounts for more than 50% of revenues each quarter.

AMD has made headlines month after month as its second-generation EPYC chips have seen adoption across public and private clouds. And, in a twist of fate, Nvidia is using two 64-core AMD EPYC chips in its DGX A100 AI accelerators.

Ampere, which was founded by former Intel President Renée J. James, in June announced a 128-core version of its Arm-based Altra data center chip.

The company is gunning for a piece of the data center market and has already seen adoption in Oracle's Gen 2 Cloud and Equinix Packet. Cloudflare and Scaleway have announced plans to source Ampere's chips.

Arm-based data center chips are not an anomaly. In pursuit of higher efficiency, Amazon developed its own line of data center chips and released its second generation of that offering in June.

Meanwhile, Microsoft baked Marvell's ThunderX2 chips into its Azure cloud offering in November 2019, and Nvidia last fall announced support for Arm-based processors beginning with Marvell's ThunderX2.