A guide to Nvidia's competitors: AMD, Qualcomm, Broadcom, startups, and more are vying to compete in the AI chip market

Nvidia is undoubtably ascendant successful the AI semiconductor space. Estimates fluctuate, but the institution has much than 80% marketplace stock by immoderate estimates erstwhile it comes to the chips that reside wrong information centers and marque products similar ChatGPT and Claude possible.

That enviable dominance goes backmost astir 2 decades, erstwhile researchers began to recognize that the aforesaid benignant of intensive computing that made complex, visually stunning video games and graphics possible, could alteration different types of computing too.

The institution started gathering its celebrated bundle stack, named Compute Unified Device Architecture oregon CUDA, 16 years earlier the motorboat of ChatGPT. For overmuch of that time, it mislaid money. But CEO Jensen Huang and a squad of existent believers saw the imaginable for graphics processing units to alteration artificial intelligence. And today, Nvidia and its products are liable for astir of the artificial quality astatine enactment successful the world.

Thanks to the prescience of Nvidia's leadership, the institution had a large caput commencement erstwhile it came to AI computing, but challengers are moving accelerated to drawback up. Some were competitors successful the gaming oregon accepted semiconductor spaces, and others person started up from scratch.

AMD

AMD, led by CEO Lisa Su, lags down Nvidia but remains its astir superior competitor. I-Hwa Cheng/AFP via Getty Images

AMD is Nvidia's apical rival successful the marketplace for AI computing successful the information center. Helmed by its formidable CEO Lisa Su, AMD launched its ain GPU, called the MI300, for the information halfway successful 2024, much than a afloat twelvemonth aft Nvidia's 2nd procreation of information halfway GPUs started shipping.

Though experts and analysts person touted the chip's specifications and imaginable based connected its plan and architecture, the company's bundle is inactive somewhat down that of Nvidia, making these chips somewhat harder to programme and usage to their afloat potential.

Analysts foretell that the institution has nether 15% marketplace share. But AMD executives importune that they are committed to bringing its bundle up to par and that aboriginal expectations for the improvement of the accelerated computing marketplace volition payment the institution — specifically, the dispersed of AI into alleged edge devices similar phones and laptops.

Qualcomm, Broadcom, and customized chips

Also challenging Nvidia are application-specific integrated circuits oregon ASICs. These custom-designed chips are little versatile than GPUs, but they tin beryllium designed for circumstantial AI computing workloads astatine a overmuch little cost, which person made them a fashionable enactment for hyperscalers.

Though multipurpose chips similar Nvidia's and AMD's graphics processing units are apt to support the largest stock of the AI-chip marketplace successful the agelong term, customized chips are increasing fast. Morgan Stanley analysts expected the marketplace for ASICs to treble successful size successful 2025.

Companies that specialize successful ASICs see Broadcom and Marvell, on with the Asia-based players Alchip Technologies and MediaTek.

Marvell is successful portion liable for Amazon's Trainium chips portion Broadcom builds Google's tensor processing units, among others. OpenAI, Apple, Microsoft, Meta, and TikTok genitor institution ByteDance person each entered the contention for a competitory ASIC arsenic well.

Amazon and Google

While besides being salient customers of Nvidia, the large unreality providers similar Amazon Web Services and Google Cloud Platform, often called hyperscalers, person besides made efforts to plan their ain chips, often with the assistance of semiconductor companies.

Amazon's Trainium chips and Google's TPUs are the astir scaled of these efforts and connection a cheaper alternate to Nvidia chips, mostly for the companies' interior AI workloads. However, the companies person shown immoderate advancement successful getting customers and partners to usage their chips arsenic well. Anthropic has committed to moving immoderate workloads connected Amazon's chips, and Apple has done the aforesaid with Google's.

Intel

Intel and its Gaudi enactment of AI chips has struggled to vie with Nvidia and AMD. Mandel Ngan/AFP via Getty Images

Once the large American sanction successful chip-making, Intel has fallen acold down its competitors successful the property of AI. But, the steadfast does person a enactment of AI chips called Gaudi that immoderate reports person said tin basal up to Nvidia's successful immoderate respects.

Intel installed a caller CEO, semiconductor seasoned Lip-Bu Tan, successful the archetypal 4th of 2025 and 1 of his archetypal actions was to flatten the enactment truthful that the AI spot operations reports straight to him.

Huawei

Though Nvidia's American hopeful challengers are many, China's Huawei is possibly the astir concerning rival of each for Nvidia and each those acrophobic with continued US supremacy successful AI.

Huang himself has called Huawei the "single astir formidable" tech institution successful China. Reports that Huawei's AI spot innovation is catching up are expanding successful frequency. New restrictions from the Biden and Trump administrations connected shipping adjacent lower-power GPUs to China person further incentivized the institution to drawback up and service the Chinese markets for AI. Analysts accidental further restrictions being considered by the Trump medication are present improbable to hamper China's AI progress.

Startups

Also challenging Nvidia are a big of startups offering caller spot designs and concern models to the AI computing market.

These firms are starting retired astatine a disadvantage, arsenic they don't person the full-sized income and organisation machines decades of spot income successful different types of tech bring. But respective are holding their ain by uncovering usage cases, customers, and organisation methods that are charismatic to customers based connected faster processing speeds oregon little cost. These new AI players see Cerebras, Etched, Groq, Positron AI, Sambanova Systems, and Tenstorrent, among others.

Read next