Who makes motherboards for AI data centers
The non-GPU trays in a rack — CPU-only compute trays, host motherboards and the switch trays carrying scale-up switch silicon — built by the same board makers as the GPU trays.
This part covers the trays that round out a rack-scale system alongside the GPU baseboard (4.03): CPU-only compute trays for control and orchestration nodes, conventional server motherboards, and switch trays that carry the scale-up switch silicon (3.01) linking GPU trays together. The same ODM tier that builds GPU baseboards builds these, joined by motherboard specialists who don't necessarily compete for GPU-tray business. AI racks need more of this category than a conventional data center rack: a rack-scale cabinet carries dedicated switch trays purely to fan out the scale-up fabric, a tray type that barely existed before rack-scale GPU systems. For investors, this is a lower-margin, higher-volume counterpart to the GPU baseboard — less differentiated, more competitive — so the interesting question is less "who wins" and more whether a builder's exposure here dilutes or complements its GPU-tray economics.
AI delta: Rack-scale GPU systems require dedicated switch trays purely to fan out the scale-up fabric between GPU trays, a tray category that has little equivalent in conventional servers.
Companies on this part (5)
| Company | Role | Ownership | Ticker | Price | 1d | Confidence | Note |
|---|---|---|---|---|---|---|---|
| ASRock Rack | manufacturer | public | 3515.TW | 224 TWD | -0.9% | inferred | |
| ASUS | manufacturer | public | 2357.TW | 946 TWD | -3.7% | inferred | |
| Gigabyte | manufacturer | public | 2376.TW | 349 TWD | -2.4% | inferred | the same ODMs plus |
| MiTAC (Tyan) | manufacturer | public | 3706.TW | 80.50 TWD | -1.1% | inferred | |
| MSI | manufacturer | public | 2377.TW | 163 TWD | -0.6% | inferred |
Who competes here
The same table read the other way: every company above competes for motherboards, CPU trays, switch trays sockets. Confidence tags matter — an inferred placement is a lead, not a confirmed rivalry.
Seen in real machines
motherboards, CPU trays, switch trays is named in 3 of the rack teardowns and supercomputer breakdowns on this site. Each line below is a piece of that machine. The companies shown are the ones named on that line, not necessarily the maker of this part.
- Nvidia GB300 NVL72 · rack teardown
- 9× NVLink switch trays (2× NVSwitch each) — Nvidia
- AMD Helios (MI455X) · rack teardown
- MareNostrum 5 · supercomputer
- General-purpose compute node (Lenovo ThinkSystem SD650 V3 dual-node trays, 12 nodes per chassis, 72 nodes per rack; 6,480 nodes in total) — Lenovo
Nearby parts
- Previous in layer: GPU baseboards & compute trays
- Next in layer: High-layer-count PCBs
- Layer: Layer 4 — Servers, racks & integration
From the blog
Latest feed
- Gigabyte Z890M Aorus Elite WiFi7 Motherboard Review: Icy-white SFF board offers full-size features for Intel — Tom's Hardware, 2026-10-23
- IBM infrastructure revenue bounces back after Q2 drop — Data Center Dynamics, 2026-10-22IBM's CEO highlighted the positive momentum expected from sovereign AI initiatives boosting infrastructure revenue after a dip in the second quarter.
- [AI Infrastructure Insight] How AI infrastructure will be designed for the future — newsroom:sk-hynix, 2026-10-22Effective real-world AI application requires improvements across faster compute, wider memory bandwidth, higher-performance networking, better storage, and more stable power infrastructure.
- Amazon spends US$220B on AI as power delays squeeze cash flow — DigiTimes, 2026-10-22Amazon's planned $220B AI CAPEX for 2026 is hampered by power and construction delays, causing hardware like server racks to stockpile in warehouses.
- Supermicro lifts quarterly sales estimate as AI demand outpaces forecasts — DigiTimes, 2026-10-21Supermicro updated its sales forecasts upwards due to stronger demand for its AI platforms and a favorable product mix.
- Interview: SEMI CEO sees US$3 trillion chip industry by 2035, as trusted partners reshape supply chain — DigiTimes, 2026-10-16SEMI's CEO predicts the semiconductor industry will achieve $2 trillion in revenue by 2035, driven by AI and quantum computing growth.
- Nvidia unveils PTFE switch board at OCP, names Taiwan partners — DigiTimes, 2026-10-15Nvidia presented a PTFE switch board at OCP and announced partnerships with Taiwanese firms.
- OCP-recognized infrastructure spending forecast to hit US$800 billion by 2029, nearly triple prior estimate — DigiTimes, 2026-10-15IDC forecasts that global spending on Open Compute Project (OCP) recognized IT infrastructure equipment will reach approximately $800 billion by 2029.
- New motherboards, old RAM, plenty of life left in yesterday's PCs — The Register, 2026-10-14The article discusses the resurgence of using DDR4 memory and older hardware for computing due to rising costs and software alternatives.
- Cisco expands AI networking with new chips, SONiC, cross-site deployment — DigiTimes, 2026-10-14Cisco is expanding its AI networking capabilities by deploying new switches powered by high-performance silicon and the SONiC operating system across various data center sites.
Questions about motherboards, CPU trays, switch trays
- What are motherboards, CPU trays, switch trays?
- The non-GPU trays in a rack — CPU-only compute trays, host motherboards and the switch trays carrying scale-up switch silicon — built by the same board makers as the GPU trays.
- Which companies make motherboards for AI data centers?
- This catalog lists 5 companies on Motherboards, CPU trays, switch trays, 5 of them with listed equity: ASRock Rack, ASUS, Gigabyte, MiTAC (Tyan), MSI.
- How does AI change demand for motherboards, CPU trays, switch trays?
- Rack-scale GPU systems require dedicated switch trays purely to fan out the scale-up fabric between GPU trays, a tray category that has little equivalent in conventional servers.
- How reliable are these motherboards, CPU trays, switch trays supplier placements?
- The 5 placements on this part break down as 5 inferred. "disclosed" means the company or a filing says so; "reported" comes from trade press or a reference-design list; "inferred" is a taxonomy lead with no citation yet.
Sources: placements without a source URL come from the taxonomy seed and are tagged accordingly. How this data is built.· Markdown · JSON