Connect your agent to akta.pro and build company lists like this across sectors and geos.
The closest NVIDIA alternatives and competitors across AI accelerator chips, GPUs and AI computing hardware — direct peers designing and selling their own AI training and inference silicon, with products, business model, funding stage and named customers for each.
NVIDIA designs GPUs and AI computing systems that power most of today's AI model training and inference.
It sells chips, rack-scale systems and networking to cloud providers and enterprises, backed by its CUDA software platform, and also makes GeForce gaming GPUs.
Its lead in AI accelerators makes it the benchmark for the alternatives listed below.
Key products DGX Platform HGX GeForce RTX CUDA / NVIDIA AI Enterprise DRIVE Omniverse Founded 1993 HQ Santa Clara, USA Category AI Computing Hardware (GPUs and Accelerated Systems) Latest round Public Business model Sells AI accelerator hardware and rack-scale systems to hyperscalers and enterprises under multi-year contracts, alongside subscription AI Enterprise software, usage-based DGX Cloud, and direct-to-consumer GeForce gaming hardware. Named customers Mercedes-Benz, Eli Lilly, IQVIA, Foxconn, LG| # | Company | Category | Products & Services | Business Model | Funding & Stage | Named Customers |
|---|---|---|---|---|---|---|
| 1 | AMD↗ Direct peer 1969 · Santa Clara · USA | Semiconductor Design and Manufacturing | EPYC CPUs Instinct GPUs Ryzen Radeon Versal Adaptive SoCs ROCm Designs high-performance EPYC server CPUs and Instinct data-center GPU accelerators for AI training and inference, alongside adaptive SoCs and consumer Radeon graphics, unified by the open-source ROCm software platform. | Enterprise Field Sales Channel Partners Hardware Sales Hardware-centric sales to hyperscalers via direct enterprise field sales (OpenAI, Meta, Oracle multi-gigawatt commitments) and deep OEM partnerships with Dell, HPE and Supermicro. | – Public |
|
| 2 | Intel↗ Direct peer 1968 · Santa Clara · USA | Semiconductors | Xeon Processors Gaudi AI Accelerators Core Ultra Arc Graphics Intel Foundry Services OpenVINO Designs Xeon server processors and Gaudi AI accelerators for data-center AI workloads, Core Ultra client chips, Arc discrete graphics, and offers contract chip manufacturing through Intel Foundry Services. | Enterprise Field Sales OEM Embedded Hardware Sales High-volume OEM-embedded processor sales to PC makers (Dell, HP, Lenovo) supplemented by direct enterprise sales and a growing foundry business for external customers including Microsoft and Google. | – Public | |
| 3 | Moore Threads↗ Direct peer 2020 · Beijing · CHN | GPU Accelerators and AI Compute Hardware | MTT S5000 GPU KUAE Cluster MUSA Software Stack SGX5000 Servers MTT Gaming GPUs Chinese full-function GPU designer founded by a former NVIDIA China GM, building proprietary MUSA-architecture GPU accelerators, CUDA-compatible software, and the 10,000-card KUAE intelligent-computing cluster. | Enterprise Field Sales Channel Partners Hardware Sales One-time hardware sales of GPU cards and cluster solutions to Chinese cloud, government and finance customers, plus consumer GPU sales through JD.com and channel partners. | $528.7M Public Last · Nov 2025 |
|
| 4 | MetaX↗ Direct peer 2020 · Shanghai · CHN | GPU Chips and Accelerators | N-Series Inference GPU C-Series Training GPU G-Series Graphics GPU MXMACA Software Platform Shanghai-based fabless GPU designer building proprietary GPGPU chips across inference, training, graphics and scientific-computing product lines, paired with a CUDA-compatible MXMACA software platform. | Enterprise Field Sales Hardware Sales Sells proprietary GPU chips directly to Chinese government entities, state-owned enterprises, cloud providers and AI foundation-model developers as a domestic NVIDIA alternative. | $391M Public Last · Jul 2022 | |
| 5 | Broadcom↗ Direct peer 1991 · Palo Alto · USA | Semiconductor and Infrastructure Software | Custom AI ASICs / XPUs Tomahawk Ethernet Switches Jericho PCIe Switches Optical Interconnects VMware Cloud Foundation Designs custom AI accelerator ASICs (XPUs) and Ethernet networking silicon co-developed with hyperscalers, holding a dominant share of the custom AI chip market serving Google, Meta and OpenAI. | Enterprise Field Sales Channel Partners Hardware Sales Direct enterprise engagements co-designing custom AI silicon with hyperscale customers, supplemented by channel and OEM-embedded networking hardware sales. | – Public |
|
| 6 | Google↗ Direct peer 1998 · Mountain View · USA | AI Accelerator Chips (Cloud TPU) | Cloud TPU Vertex AI Gemini Google Cloud AI Infrastructure Designs proprietary Tensor Processing Units (TPUs) for AI training and inference, offered internally and to external customers via Google Cloud as an alternative to GPU-based AI infrastructure. | Enterprise Field Sales API First Usage Based Subscription Recurring TPU capacity sold primarily as usage-based Google Cloud infrastructure to enterprise and AI-lab customers, alongside subscription Workspace and API-first developer products. | – Public | |
| 7 | Amazon Web Services↗ Direct peer 2006 · Seattle · USA | AI Accelerator Chips (Trainium / Inferentia) | AWS Trainium AWS Inferentia Amazon Bedrock Amazon SageMaker Amazon EC2 Designs custom Trainium and Inferentia AI accelerator silicon deployed exclusively within AWS cloud infrastructure, offered as an alternative to GPU instances for AI training and inference workloads. | Enterprise Field Sales Usage Based Subscription Recurring Custom AI silicon monetized purely as usage-based cloud compute instances (EC2, SageMaker, Bedrock) rather than sold as standalone hardware. | – Subsidiary of Amazon.com (Public) |
|
| 8 | Huawei↗ Direct peer 1987 · Shenzhen · CHN | AI Accelerator Chips (Ascend) | Ascend NPUs Atlas AI Servers CANN Software Stack Kunpeng CPUs Designs proprietary Ascend AI accelerator chips and Atlas AI server systems with the CANN software stack, positioned as a domestic AI-chip alternative for the Chinese market amid export restrictions on foreign GPUs. | Enterprise Field Sales Channel Partners Hardware Sales End-to-end ICT hardware sold through direct enterprise sales and channel partners to carriers, enterprises and governments across 170-plus countries. | – Private (Employee-Owned) |
|
| 9 | Cerebras Systems↗ Direct peer 2015 · Sunnyvale · USA | AI Semiconductor Hardware | WSE-3 Wafer-Scale Chip CS-3 System Cerebras Inference Cloud Condor Galaxy Designs the Wafer Scale Engine, the world's largest single-chip AI processor, and the CS-3 system, delivering wafer-scale AI training and pay-per-token inference as an alternative to distributed GPU clusters. | Enterprise Field Sales Product Led Growth Hardware Sales Usage Based Sells CS-3 hardware systems directly to enterprises and research institutions, plus pay-per-token inference and pay-per-hour training via the Cerebras Cloud platform. Went public on NASDAQ in 2026. | $4.67B Public Last · Apr 2026 |
|
| 10 | SambaNova↗ Direct peer 2017 · Palo Alto · USA | AI Inference Infrastructure | SN40L / SN50 RDU Chips SambaRack SambaCloud SambaStack Designs the Reconfigurable Dataflow Unit AI inference chip and a full-stack platform (SambaRack hardware, SambaStack on-prem software, SambaCloud inference service) for low-latency inference on large open-source models. | Enterprise Field Sales Hardware Sales Usage Based Sells SambaRack hardware and SambaStack licenses directly to enterprises and sovereign AI operators, plus usage-based inference via SambaCloud. | $2.49B Series F Last · Jul 2026 |
|
| 11 | Rebellions↗ Direct peer 2020 · Seongnam · KOR | AI Inference Accelerators (NPU Semiconductors) | Rebel100 NPU ATOM-Max RebelRack / RebelPOD RBLN SDK South Korean fabless designer of chiplet-based AI inference NPUs and vertically integrated rack-scale AI infrastructure, targeting large language model inference for hyperscalers and telecom operators. | Enterprise Field Sales Channel Partners Hardware Sales Licensing Royalties Sells NPU chips and rack-scale systems directly to Korean and regional hyperscalers, telecoms and enterprises under Korea's sovereign AI chip initiative. | $918M Series C Last · Mar 2026 | |
| 12 | Graphcore↗ Direct peer 2016 · Bristol · GBR | AI Accelerator Hardware | IPU (Intelligence Processing Unit) Bow Pod IPU-POD Poplar SDK Designs Intelligence Processing Units with an in-processor-memory architecture, sold as Bow Pod and IPU-POD scale-out systems with the Poplar SDK, positioned as an alternative to NVIDIA's GPU and CUDA stack. | Enterprise Field Sales Channel Partners Hardware Sales Licensing Royalties Sells IPU hardware and software licenses directly and through channel partners; acquired by SoftBank in 2026 while continuing to operate its IPU product line. | $1.14B Acquired (SoftBank) Last · May 2026 |
|
| 13 | d-Matrix↗ Direct peer 2019 · Santa Clara · USA | AI Inference Accelerators | Corsair DIMC Accelerator SquadRack JetStream I/O Aviator Software Designs Digital In-Memory Compute AI inference accelerators (Corsair) as PCIe cards and rack-scale SquadRack systems, targeting low-latency generative AI inference for hyperscalers and neoclouds. | Enterprise Field Sales Channel Partners Hardware Sales Sells Corsair accelerator cards and rack-scale reference designs directly to hyperscalers, neoclouds and frontier AI labs, with OEM integration through Arista, Broadcom and Supermicro. | $429M Series C Last · Nov 2025 |
|
| 14 | Etched.ai↗ Direct peer 2024 · San Jose · USA | AI Inference Semiconductors | Sohu ASIC Oasis Inference System Designs Sohu, the first transformer-specialized ASIC, purpose-built to accelerate inference for large language models, sparse Mixture-of-Experts and diffusion models, fabricated via TSMC. | Enterprise Field Sales Hardware Sales Direct B2B hardware sales of the Sohu transformer ASIC to hyperscalers and AI enterprises, positioned as a lower-cost-per-token alternative to general-purpose GPUs. | $2.13B Series D Last · Aug 2026 |
|
| 15 | Cambricon↗ Direct peer 2016 · Beijing · CHN | AI Accelerator Chips | Siyuan 370 (MLU370) MLU370-X8 Training Card MLU370-X4 Accelerator Card China's leading AI chip designer, selling MLU cloud training and inference accelerators to hyperscalers and server OEMs | Enterprise Field Sales Channel Partners Hardware Sales Licensing Royalties Sells accelerator chips and cards directly to Chinese hyperscalers and through server OEMs such as Inspur and Lenovo | $761.6M Public (SSE: 688256) Last · Oct 2025 |
This list covers the primary NVIDIA alternatives and competitors across AI accelerator chips, GPUs, and AI computing hardware. NVIDIA is the reference point and is not listed; direct peers are companies that design and sell their own AI training or inference silicon, competing on the same core workflow as NVIDIA's GPU and rack-scale systems business.
The peer set spans established diversified semiconductor makers (AMD, Intel, Broadcom), hyperscaler in-house silicon programs (Google TPU, AWS Trainium), sovereign and geography-specific chip makers (Huawei, Cambricon, Moore Threads, MetaX, Rebellions), and venture-backed AI-chip specialists (Cerebras Systems, SambaNova, Graphcore, d-Matrix, Etched.ai). Companies that only rent or resell NVIDIA GPU capacity, rather than designing their own chips, are excluded.
Firmographic, business model, funding, and customer data sourced directly from akta.pro Company Data. Figures shown as "–" reflect data not currently available in akta.pro, not a confirmed absence.
Peer tiersCompanies are tagged against standard industry codes and against akta.pro's internal taxonomy, so the same set can be rebuilt by code, by segment, or by a combination of both.
NAICSIndustry codes are provided for cross-referencing against public filings and procurement records. akta.pro classification is independent of them, so a company may carry an internal segment without a matching NAICS or SIC registration.