Who Can Challenge NVIDIA?

Nvidia is undoubtedly dominant in the field of artificial intelligence semiconductors. Valuations fluctuate, but the company is estimated to have a market share of more than 80% in the data center chip space, as well as chips for products such as ChatGPT and Claude.

This enviable dominance dates back nearly two decades, when researchers began to realize that the intensive computing that made complex, visually stunning video games and graphics possible could also enable other types of computing.

The company began building its famous software stack, the Computing Unified Device Architecture (CUDA), 16 years before ChatGPT was launched. For most of that time, the company was in the red. But CEO Jensen Huang and a group of committed supporters see the potential of graphics processing units (GPUs) to empower AI. Today, Nvidia and its products dominate the majority of AI applications around the world.

Thanks to the foresight of Nvidia's leadership, the company has a huge lead in AI computing, but challengers are struggling to catch up. Some of them are competitors in gaming or traditional semiconductors, while others are starting from scratch.

AMD

AMD is NVIDIA's biggest competitor in the data center AI computing market. Under the leadership of a strong CEO, Lisa Su, AMD launched its own data center GPU in 2024, called the MI300, which is more than a full year after the shipment of Nvidia's second-generation data center GPUs.

Although experts and analysts alike rave about the chip's design and architecture specifications and potential, the company's software still lags behind Nvidia, making it somewhat difficult to program these chips and reach their full potential.

Analysts predict that the company has a market share of less than 15%. But AMD executives insisted that they are committed to improving their software and believe that the company's future expectations for accelerated computing market growth - specifically, the adoption of artificial intelligence into so-called edge devices such as mobile phones and laptops - will benefit the company.

Qualcomm, Broadcom and custom chips

Application-specific integrated circuits (ASICs) also pose challenges for NVIDIA. These custom-designed chips are not as functional as GPUs, but they can be specifically designed for specific AI computing workloads at a much lower cost, making them a popular choice for hyperscale computing enterprises.

While multi-purpose chips like graphics processors from Nvidia and AMD are likely to account for the lion's share of the AI chip market for a long time, custom chips are growing rapidly. Morgan Stanley analysts expect the ASIC market to double in size by 2025.

Companies specializing in ASICs include Broadcom and Marvell, as well as Asia-based companies Alchip Technologies and MediaTek.

0021-02983 TXZ Inner Shield

Marvell is partly responsible for Amazon's Trainium chips, while Broadcom is responsible for building Google's tensor processing units, among other things. OpenAI, Apple, Microsoft, Meta, and TikTok's parent company, ByteDance, have all joined the race for competing ASICs.

Amazon and Google

Major cloud providers such as Amazon Web Services and Google Cloud Platform, often referred to as hyperscalers, are also working to design their own chips, often with the help of semiconductor companies, as Nvidia's main customers.

Amazon's Trainium chips and Google's TPUs are the largest of these efforts, offering cheaper alternatives to Nvidia chips and are primarily used for AI workloads within companies. However, these companies have also made some progress in attracting customers and partners to use their chips. Anthropic has committed to running some workloads on Amazon silicon, and Apple has also committed to running some workloads on Google silicon.

Intel

Intel used to be a giant in chip manufacturing in the United States, but in the era of artificial intelligence, it has fallen far behind its competitors. The company does have a line of AI chips called Gaudi, though, and some reports say that the chip is comparable to Nvidia's chips in some ways.

Intel appointed a new CEO in the first quarter of 2025, semiconductor veteran Lip-Bu Tan, and one of his first actions was to streamline the organizational structure so that AI chip operations would report directly to him.

Huawei

While there are many promising challengers for Nvidia in the United States, China's Huawei is perhaps the most worrisome competitor for Nvidia and all those concerned about the continued U.S. dominance in AI.

Huang himself has called Huawei the "most powerful" technology company in China. There are more and more reports that Huawei's AI chip innovation is catching up. The Biden and Trump administrations' new restrictions on low-power GPU exports to China have further encouraged Huawei to catch up and serve the Chinese AI market. Analysts say further restrictions being considered by the Trump administration are unlikely to hinder Chinese's progress in artificial intelligence at this time.

Start-ups

Nvidia has also been challenged by many startups, offering new chip designs and business models for the AI computing market.

These companies started at a disadvantage because they lacked the well-established sales and distribution systems that have come with decades of chip sales in other technology areas. But some of these companies have maintained their advantage by looking for use cases, customers, and distribution methods to attract customers with faster processing speeds or lower costs. These new AI companies include Cerebras, Etched, Groq, Positron AI, Sambanova Systems, and Tenstorrent, among others.

In addition, China's Cambrian, Haiguang, Loongson, Moore Threads, Muxi, Bichen Technology and Tiantian Zhixin have also challenged Nvidia on this track.