L4 · Chips
Top AI ASIC & Accelerator Companies
GPUs, custom ASICs, semicap, foundry, EDA & IP, power semiconductors.
Last updated: · weekly
The chip layer is where AI compute is physically realized. It spans merchant GPUs, custom training and inference ASICs, the semiconductor capital equipment and foundries that fabricate them, and the EDA and IP stack that designs them. When people ask which companies matter for AI silicon, they are usually asking about several distinct roles at once.
This roster lists the companies the engine tracks in the chip layer (L4) of the AI-infrastructure stack, defined by the V15.1 taxonomy charter as compute devices plus the equipment, foundry, and design stack behind them. It is a directory of who operates in the layer — not a market-share ranking. For an activity-based ranking and the underlying signals, follow the leaderboard and index links below.
Companies in this layer
17Source: engine canonical_companies registry, primary_layer = L4 (V15.1 taxonomy)
| Company | Segment | Ticker | Status | HQ |
|---|---|---|---|---|
| Alphawave Semi | — | AWE.L | Public | — |
| AMD | Merchant GPU | AMD | Public | US |
| Arm Holdings | EDA & IP | ARM | Public | — |
| Broadcom Inc. | Custom ASIC | AVGO | Public | US |
| Cadence Design Systems | EDA & IP | CDNS | Public | — |
| Intel | Merchant GPU | INTC | Public | US |
| Marvell Technology, Inc. | Custom ASIC | MRVL | Public | US |
| NVIDIA | Merchant GPU | NVDA | Public | US |
| Semtech | Power Semiconductors | SMTC | Public | US |
| Siemens EDA | EDA & IP | — | Public | — |
| Synopsys | EDA & IP | SNPS | Public | — |
| Cerebras Systems | Inference Startup | — | Private | US |
| Etched | Inference Startup | — | Private | — |
| Groq | Inference Startup | — | Private | US |
| Rebellions | Inference Startup | — | Private | KR |
| SambaNova Systems | Inference Startup | — | Private | US |
| Tenstorrent | Inference Startup | — | Private | CA |
How this list is built
The silicon: compute devices plus the equipment, foundry, and design stack behind them. Companies are sourced from the engine's canonical company registry, filtered to the L4 layer of the V15.1 taxonomy. This is a roster of who operates in the layer, not a subjective ranking. For an activity-based ranking (hiring + business signals) see the L4 leaderboard and the layer index on /indices.
Topics in this layer
FAQ
- What is the AI chip (L4) layer?
- In the V15.1 AI-infrastructure taxonomy, L4 (Chips) is the silicon layer: GPUs, custom ASICs, semiconductor capital equipment, foundry, EDA & IP, and power semiconductors — the compute devices plus the equipment, foundry, and design stack behind them.
- How is this company list built?
- Companies are sourced from the engine's canonical_companies registry, filtered to primary_layer = L4 under the V15.1 taxonomy. It is a roster of companies operating in the layer, not a subjective ranking; for an activity-based ranking see the L4 leaderboard.
- Where do the AI memory and networking companies sit?
- High-bandwidth memory and advanced packaging sit one layer down at L3, and the optical/networking fabric at L2. See the related stack pages linked on this page.