See where your brand stands in AI recommendations.

Products

Brands
Markets
Work with us
Pricing

Resources

Research
Methodology
Blog
MCP

© 2026 Parse. All rights reserved.

LegalPrivacy Policy Terms of Service

Sign in Check your brand

Best AI Semiconductor and Accelerator Vendors for custom AI chip design, According to AI | Parse

Markets
Semiconductor
AI Semiconductor and Accelerator Vendors
Custom AI chip design

AI Semiconductor and Accelerator VendorsCustom AI chip design

Best custom AI chip design, according to AI

Data as of Apr 11, 2026 · Based on 69 AI answers · A buyer need in AI Semiconductor and Accelerator Vendors. · See how Parse measures this

leads this need with 9 recommendations across both AI platforms, anchored by repeated mentions of the Azure Maia 100 accelerator and its Braga successor. follows closely at 8 recommendations, driven by the MTIA family. The gap between and is under 2 points, making the top of this list contested in the March to April observation window.

Microsoft

Where a different pick wins:

For wafer-scale trainingCerebras Systems
AI consistently names Cerebras WSE-3 when the buyer asks for the biggest, fastest training chip.
For RISC-v based custom acceleratorsTenstorrent
RISC-V questions consistently route to Tenstorrent and Jim Keller's Wormhole/Blackhole designs.
For edge and embedded npusHailo
Edge-focused buyers are directed to Hailo-8/10H processors for robotics, cameras, and automotive.
For transformer-specific asicsEtched
AI names Etched as the specialized transformer ASIC startup when buyers need language-model optimized silicon.
For mobile system-on-chip npusQualcomm
Snapdragon Hexagon NPU answers surface whenever mobile or on-device AI silicon is mentioned.
For custom RISC-v IP and NPU integrationSemidynamics
Semidynamics is referenced for customizable RISC-V and NPUs that other companies integrate into their own designs.

AnthropicIntel Corporation

Recommendation share

By platform

How AI answers questions about Custom AI chip design

#BrandRecommendation share

1

Shows up for Azure Maia 100 and the upcoming Braga accelerator aimed at cloud AI workloads in Azure.

2

Surfaces through MTIA accelerators with a focus on recommendation workloads and running Llama models internally.

3

Appears as AWS Annapurna Labs designing Trainium for training and Inferentia for inference, optimizing cost-effective cloud performance.

4

Known as the co-design partner for Google, Meta, and OpenAI, Broadcom is named as 'King of ASICs' for custom silicon.

5

Cerebras Systems6%

Recommended for Wafer-Scale Engine WSE-3 chips aimed at massive, fast training and inference workouts.

6

Cites RISC-V based AI accelerators like Wormhole and Blackhole led by Jim Keller, offering open-source flexibility.

7

Google's TPU and Ironwood designs surface for data center training and inference, now in their 6th generation.

8

Appears with Apple Neural Engine in M-series and the project Baltra custom server chip developed with Broadcom.

9

Named as a startup designing specialized transformer-optimized ASICs for large language models.

10

Specializes in Language Processing Unit or LPU architecture focused solely on rapid LLM inference.

11

Mentions partnerships with Broadcom and TSMC for its first in-house AI chip targeting 2026 mass production.

12

Brings Reconfigurable Dataflow Units (SN50 RDU) aimed at large generative AI models and enterprise workloads.

13

14

15

16

17

18

19

20IIntel Corporation1%

21

22NNVIDIA Corporation1%

23

24

25

26TTSMC1%

View 21 more brands

Google AI Overviews

1Microsoft
2Meta
3Amazon

ChatGPT

1Meta
2Etched
3Groq

Follow changes to AI Semiconductor and Accelerator Vendors in AI answers

Follow for free

Markets
Semiconductor
AI Semiconductor and Accelerator Vendors
Custom AI chip design

AI Semiconductor and Accelerator VendorsCustom AI chip design

Best custom AI chip design, according to AI

Data as of Apr 11, 2026 · Based on 69 AI answers · A buyer need in AI Semiconductor and Accelerator Vendors. · See how Parse measures this

We are evaluating neural processing units (NPUs). Who designs custom AI silicon?Microsoft Meta

Microsoft leads this need with 9 recommendations across both AI platforms, anchored by repeated mentions of the Azure Maia 100 accelerator and its Braga successor. follows closely at 8 recommendations, driven by the MTIA family. The gap between and is under 2 points, making the top of this list contested in the March to April observation window.

Amazon

Cerebras Systems

ChatGPT and Google AI both produce long lists of hyperscalers, chip firms, and startups. Microsoft, Meta, Amazon, Broadcom, Cerebras, and Tenstorrent are among the most consistently repeated across responses.

Where a different pick wins:

For wafer-scale trainingCerebras Systems
AI consistently names Cerebras WSE-3 when the buyer asks for the biggest, fastest training chip.
For RISC-v based custom acceleratorsTenstorrent
RISC-V questions consistently route to Tenstorrent and Jim Keller's Wormhole/Blackhole designs.
For edge and embedded npusHailo
Edge-focused buyers are directed to Hailo-8/10H processors for robotics, cameras, and automotive.
For transformer-specific asicsEtched
AI names Etched as the specialized transformer ASIC startup when buyers need language-model optimized silicon.
For mobile system-on-chip npusQualcomm
Snapdragon Hexagon NPU answers surface whenever mobile or on-device AI silicon is mentioned.
For custom RISC-v IP and NPU integrationSemidynamics
Semidynamics is referenced for customizable RISC-V and NPUs that other companies integrate into their own designs.

AnthropicIntel Corporation

Recommendation share

By platform

How AI answers questions about Custom AI chip design

#BrandRecommendation share

1

Shows up for Azure Maia 100 and the upcoming Braga accelerator aimed at cloud AI workloads in Azure.

2

Surfaces through MTIA accelerators with a focus on recommendation workloads and running Llama models internally.

3

Appears as AWS Annapurna Labs designing Trainium for training and Inferentia for inference, optimizing cost-effective cloud performance.

4

Known as the co-design partner for Google, Meta, and OpenAI, Broadcom is named as 'King of ASICs' for custom silicon.

5

Cerebras Systems6%

Recommended for Wafer-Scale Engine WSE-3 chips aimed at massive, fast training and inference workouts.

6

Cites RISC-V based AI accelerators like Wormhole and Blackhole led by Jim Keller, offering open-source flexibility.

7

Google's TPU and Ironwood designs surface for data center training and inference, now in their 6th generation.

8

Appears with Apple Neural Engine in M-series and the project Baltra custom server chip developed with Broadcom.

9

Named as a startup designing specialized transformer-optimized ASICs for large language models.

10

Specializes in Language Processing Unit or LPU architecture focused solely on rapid LLM inference.

11

Mentions partnerships with Broadcom and TSMC for its first in-house AI chip targeting 2026 mass production.

12

Brings Reconfigurable Dataflow Units (SN50 RDU) aimed at large generative AI models and enterprise workloads.

13

14

15

16

17

18

19

20IIntel Corporation1%

21

22NNVIDIA Corporation1%

23

24

25

26TTSMC1%

View 21 more brands

Google AI Overviews

1Microsoft
2Meta
3Amazon

ChatGPT

1Meta
2Etched
3Groq

Follow changes to AI Semiconductor and Accelerator Vendors in AI answers

Follow for free

We are evaluating neural processing units (NPUs). Who designs custom AI silicon?Microsoft Meta

Amazon

Cerebras Systems

ChatGPT and Google AI both produce long lists of hyperscalers, chip firms, and startups. Microsoft, Meta, Amazon, Broadcom, Cerebras, and Tenstorrent are among the most consistently repeated across responses.