Een nvidia GPU kan met een AMD cpu worden gebruikt. Verkoop van de een kan de ander ook toe doen nemen. En hosters hebben nog steeds hardware nodig. Als ze intel niet kunnen krijgen meer switchen ze wellicht naar AMD voor niet AI dingen.
AMD staat op de vierde plek voor AI hardware met 5 tot 10% marktaandeel tegenover nvidia 80 tot 85%, nog al een verschil.
Nvidia controls approximately 80% to 85% of the global data center AI accelerator market share by revenue. While its peak near-monopoly of ~92% has slightly decreased due to rising traction from AMD and custom cloud chips (ASICs), it remains overwhelmingly dominant. [1, 2, 3]
The metrics vary depending on whether you look at the
merchant chip market (standalone components sold to third parties) or the
broader value chain (including foundries, custom server builders, and memory providers). [4]
This breakdown tracks the merchant silicon and design market share for AI processors (GPUs, TPUs, and specialized accelerators): [2]
- Nvidia: Holds an estimated 80% to 85% market share. Its dominant positioning is driven heavily by massive hyperscaler clusters and its proprietary CUDA software ecosystem. [1, 2, 4]
- Google (TPU): Captures roughly 6% to 8% market share. While its custom Tensor Processing Units are mostly kept for internal workloads and Google Cloud ecosystem clients, it expanded its footprint via major multi-year tenant lease agreements. [2, 5]
- AMD: Captures approximately 5% to 10% market share. Driven by its Instinct MI300X and MI325X accelerator series, AMD has established itself as the leading alternative source for heavy inference workloads. [1, 2, 6]
- Amazon (AWS Trainium/Inferentia): Commands about 2% to 3% market share, scaling fast due to cost-effective inference options across AWS deployments. [2]
- Intel: Holds around 1% market share as it shifts enterprise focus toward Gaudi accelerator pipelines and upcoming node transitions. [7, 8]
For companies that provide the foundational layers, memory, or assembly rather than the processor architecture itself, market share is measured against their respective niches:
- TSMC (Taiwan Semiconductor Manufacturing Company): Commands a near-monopoly (>90% market share) for manufacturing the advanced physical silicon nodes (3nm, 2nm) used by Nvidia, AMD, Apple, and major custom-silicon providers.
- Broadcom: Dominates the custom AI ASIC co-design market and high-end networking switches, acting as the key backend partner for platforms like Google, Anthropic, Meta, and OpenAI.
- Micron Technology: Is a major player in the high-bandwidth memory (HBM) ecosystem, sharing the highly consolidated AI memory market alongside SK Hynix and Samsung.
- Super Micro Computer: Competes heavily in the server infrastructure and rack-scale liquid cooling assembly segment, capturing a major portion of turnkey GPU deployment buildouts. [9, 10, 11, 12, 13, 14, 15]
Would you like to narrow down this breakdown to look exclusively at the
custom ASIC (cloud chip) market, or compare the
financial revenue differences between Nvidia and AMD?
[1]
https://presenc.ai
[2]
https://presenc.ai
[3]
https://siliconanalysts.com
[4]
https://www.remio.ai
[5]
https://valueaddvc.com
[6]
https://commandlinux.com
[7]
https://en.unibetter-ic.com
[8]
https://commandlinux.com
[9]
https://www.sphericalinsights.com
[10]
https://finance.yahoo.com
[11]
https://bingx.com
[12]
https://www.theglobeandmail.com
[13]
https://www.tradingview.com
[14]
https://sqmagazine.co.uk
[15]
https://www.prnewswire.com