Which companies make server CPUs for AI data centers?

This catalog lists 5 companies on Server CPUs, 5 of them with listed equity: AMD, Ampere, Arm, Intel, Nvidia.

How reliable are these server CPUs supplier placements?

The 5 placements on this part break down as 2 disclosed, 3 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.

1.05

Who makes server CPUs for AI data centers

The host CPU in an AI server — from AMD, Intel, or Arm-based designs including Nvidia's own and hyperscaler in-house parts — that orchestrates the GPUs and handles work the accelerators don't do.

Even a GPU-dense AI server needs a CPU to boot the system, run the operating system, orchestrate data movement to the accelerators, and handle any workload that isn't matrix math. Server CPUs sit alongside merchant GPUs or custom ASICs (Merchant GPUs / accelerators, Hyperscaler custom ASICs) on the same board, feed data to them over NICs and DPUs (NICs / SuperNICs, DPUs), and are increasingly Arm-based rather than x86, both from merchant vendors and hyperscalers building their own (see the child part below). The category also includes CPU designs, like Nvidia Grace, built specifically as a tightly coupled companion to that vendor's own GPUs rather than a general-purpose part. For investors, this is a maturer, lower-growth-rate segment than the accelerator itself, but ride share matters: AI servers still buy a CPU per node even as GPU count per node rises, so unit volume tracks server shipments broadly. The competitive dynamic worth watching is the Arm-based share gain, both from merchant vendors and hyperscalers building silicon in-house, versus the incumbent x86 vendors — a share shift more than a category-wide volume story.

AI delta: AI servers still need a host CPU to orchestrate GPUs and run non-matrix workloads, so CPU volume rises with every GPU-dense server shipped even though the CPU itself does little of the AI math.

Appears at:Enterprise / small colo Large colo / hyperscale building AI campus GW-class campus· cooling:any — CPU count per server is roughly fixed regardless of facility size; total volume scales with server count across tiers, not with per-server intensity.

Companies on this part (5)

Show:

Company	Role	Ownership	Ticker	Price	1d	Confidence	Note
AMD	manufacturer	public	AMD	517 USD	+2.2%	disclosed	EPYC · source
Ampere · unit of SoftBank	manufacturer	subsidiary	9984.T*			inferred
Arm	manufacturer	public	ARM	210 USD	+1.0%	inferred	as licensor
Intel	manufacturer	public	INTC	86.96 USD	+2.6%	inferred	Xeon
Nvidia	manufacturer	public	NVDA	211 USD	-0.7%	disclosed	Grace — Arm-based · source

Hyperscaler Arm CPUs 1.05a

Company	Role	Ownership	Ticker	Price	1d	Confidence	Note
Alphabet (Google)	manufacturer	public	GOOGL	320 USD	-1.3%	inferred	Axion
Amazon (AWS)	manufacturer	public	AMZN	248 USD	-1.1%	inferred	Graviton
Microsoft	manufacturer	public	MSFT	525 USD	+0.5%	inferred	Cobalt

Who competes here

The same table read the other way: every company above competes for server CPUs sockets. Confidence tags matter — an inferred placement is a lead, not a confirmed rivalry.

Seen in real machines

server CPUs is named in 18 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
- Grace CPU and LPDDR5X — Micron, Samsung Electronics, SK Hynix
AMD Helios (MI455X) · rack teardown
- EPYC "Venice" host CPU (Zen 6, up to 256 cores, 1.6 TB/s memory bandwidth) — AMD
El Capitan · supercomputer
- AMD Instinct MI300A APU (CPU, GPU, cache and HBM3 in one package; 24 Zen 4 cores, 228 CDNA 3 compute units, 128 GB HBM3) — AMD
- Rabbit near-node storage (4U module: 16 SSDs plus one AMD EPYC storage processor, two per cabinet, each wired to 16 compute nodes) — HPE, AMD
Frontier · supercomputer
- Host CPU (one 64-core AMD Optimized 3rd Gen EPYC per node, coherent with the GPUs over Infinity Fabric) — AMD
Aurora · supercomputer
- Intel Xeon CPU Max "Sapphire Rapids" host CPU (52 cores, 4 tiles joined by EMIB, 64 GB on-package HBM2e, 350 W) — Intel
JUPITER · supercomputer
- GH200 Grace Hopper Superchip (Hopper GPU plus 72-core Grace CPU joined by NVLink-C2C) — Nvidia, Arm
- Cluster module CPUs, AMD EPYC Turin 9655 (582 dual-socket nodes, 96 cores per CPU) — AMD
- Cluster module CPUs, SiPearl Rhea1 (162 dual-socket nodes, 80 Arm Neoverse V1 cores and 64 GB HBM2e per CPU; first samples delivered to Bull 2026-09-22) — SiPearl, Arm, TSMC
Alps · supercomputer
- GH200 Grace Hopper Superchip (72-core Grace CPU plus Hopper GPU with 96 GB HBM3, joined by NVLink-C2C; 10,752 superchips) — Nvidia, Arm
- CPU-only partition (1,024 nodes of two AMD EPYC 7742 Rome CPUs, 64 cores each; installed from October 2020) — AMD
- AMD Instinct MI300A partition (128 nodes, 4 APUs per node, 512 in total; HPE Cray EX255A blades) — AMD, TSMC, HPE
LUMI · supercomputer
- LUMI-G host CPU (AMD EPYC 7A53 "Trento", 64 Zen 3 cores, wired to the GPUs by Infinity Fabric) — AMD
- LUMI-C CPU partition (2,048 nodes, each with two 64-core AMD EPYC 7763 "Milan"; 256 GiB to 1 TiB per node) — AMD
- LUMI-D visualization nodes (eight nodes with 2 x AMD EPYC 7742 and 8 x NVIDIA A40 each; 64 A40 GPUs in total) — Nvidia, AMD
Leonardo · supercomputer
- Host CPUs (Booster: one 32-core Xeon Platinum 8358 per node; Data Centric: two 56-core Xeon Platinum 8480+ per node) — Intel
- Frontend and service nodes (32 BullSequana X430-E6 and X450-E6 frontend servers on Xeon Platinum 8358 with 16 Nvidia Quadro RTX 8000 visualisation GPU pairs; 11 management nodes on AMD EPYC 7H12) — Eviden/Atos, Intel, AMD, Nvidia
Isambard-AI · supercomputer
- GH200 Grace Hopper Superchip (Hopper GPU plus 72-core Grace CPU joined by NVLink-C2C, 5,448 in the full system) — Nvidia, Arm
Venado · supercomputer
- GH200 Grace Hopper Superchip (Arm-based Grace CPU plus Hopper H100 GPU joined by NVLink-C2C; 2,560 superchips) — Nvidia, Arm
- Grace CPU Superchip partition (920 CPU-only superchips, each two 72-core Grace CPUs, 144 Arm cores, linked by NVLink-C2C; up to 960 GB LPDDR5X per superchip) — Nvidia, Arm
Perlmutter · supercomputer
- GPU-node host CPU (one 64-core AMD EPYC 7763 "Milan" per node, PCIe 4.0 to the GPUs and NICs) — AMD
- CPU partition (3,072 nodes, four per compute blade, each with two 64-core AMD EPYC 7763 and 512 GB DDR4) — HPE, AMD
- Scratch storage servers (single-socket AMD EPYC 7502 in each E1000 server canister, with two Slingshot NICs per server) — AMD
Eagle (Microsoft Azure) · supercomputer
- Host CPU (Intel Xeon Platinum 8480C, 4th Gen Xeon Scalable / Sapphire Rapids; 96 physical cores per VM) — Intel
Eos · supercomputer
- Host CPUs (two 56-core Xeon Platinum 8480C per DGX H100, 1,152 in total) — Intel
Fugaku · supercomputer
- A64FX CPU design (Fujitsu in-house core and memory system; 4 core-memory groups, 512-bit SVE, Tofu-D controller on die) — Fujitsu
- Arm instruction set (Armv8.2-A with the SVE vector extension, co-specified by Fujitsu) — Arm
MareNostrum 5 · supercomputer
- Host CPUs (Intel Sapphire Rapids: Xeon Platinum 8460Y+ in the accelerated partition, 8480+ in the general-purpose partition, Xeon Max 9480 in the 72 HBM nodes) — Intel
Horizon · supercomputer
- Grace Blackwell compute nodes (GB200 NVL4: 1 Grace CPU and 2 Blackwell GPUs per node; 2,000 nodes, 4,000 GPUs) — Nvidia
- Vera CPU nodes (4,750 planned, two 88-core Vera sockets each, 72 nodes per rack; due early 2027) — Nvidia
Doudna (NERSC-10) · supercomputer
- Compute platform (NVIDIA Vera Rubin CPU-GPU platform: Rubin GPUs and Vera CPUs) — Nvidia

Nearby parts

Children: Hyperscaler Arm CPUs
Previous in layer: Startup accelerators
Next in layer: NICs / SuperNICs
Layer: Layer 1 — Compute silicon & packaging

Latest feed

FCC grants Intel a waiver to foreign robot ban to prevent chip disruptions — DigiTimes, 2026-10-23
The FCC granted Intel a waiver regarding foreign-made mobile robots to ensure continuity in chip production.
Analysis: Nvidia's Vera CPU takes on Intel and AMD with a platform play — DigiTimes, 2026-10-23
Nvidia secured a CPU supply order from Modal, challenging established CPU manufacturers like Intel and AMD by expanding its hardware offerings beyond accelerators.
HPE expands Private Cloud AI solution to support mixed Nvidia hardware clusters, launches Vera CPU rack — Data Center Dynamics, 2026-10-22
HPE launched a rack featuring Vera CPUs to expand its private cloud AI solution supporting mixed Nvidia hardware clusters.
Taiwan-France semiconductor alliance strengthens with CPU pact between SiPearl and Egis Group — DigiTimes, 2026-10-22
Egis Group and SiPearl signed a pact to jointly develop a Mobius 100 CPU utilizing UCIe chiplet technology for sovereign AI and HPC infrastructure.
[AI Infrastructure Insight] How AI infrastructure will be designed for the future — newsroom:sk-hynix, 2026-10-22
Effective 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, 2026-10-22
Amazon's planned $220B AI CAPEX for 2026 is hampered by power and construction delays, causing hardware like server racks to stockpile in warehouses.
AMD and Intel will work together to create a single PyTorch code path for AI inference on CPUs — Tom's Hardware, 2026-10-21
AMD and Intel are collaborating to provide unified CPU inference optimizations for PyTorch via the oneDNN library.
AMD: Physical AI demands more than TOPS — DigiTimes, 2026-10-21
AMD suggested that measuring performance solely by TOPS is insufficient for physical AI systems, which require evaluating end-to-end, real-time responsiveness.
Nvidia's Vera CPU signals a broader shift in AI infrastructure — DigiTimes, 2026-10-21
Nvidia is emphasizing the growing importance of CPU resources for complex AI workloads, suggesting a potential architectural shift in AI compute design.
Modal taps Nvidia Vera CPUs after firm's internal tests show 2x throughput over Intel and AMD chips — Tom's Hardware, 2026-10-20
Modal claims that Nvidia's Vera CPU achieves double the throughput of chips from Intel and AMD when running AI agent workloads.

Questions about server CPUs

What are server CPUs?: The host CPU in an AI server — from AMD, Intel, or Arm-based designs including Nvidia's own and hyperscaler in-house parts — that orchestrates the GPUs and handles work the accelerators don't do.
Which companies make server CPUs for AI data centers?: This catalog lists 5 companies on Server CPUs, 5 of them with listed equity: AMD, Ampere, Arm, Intel, Nvidia.
How does AI change demand for server CPUs?: AI servers still need a host CPU to orchestrate GPUs and run non-matrix workloads, so CPU volume rises with every GPU-dense server shipped even though the CPU itself does little of the AI math.
How reliable are these server CPUs supplier placements?: The 5 placements on this part break down as 2 disclosed, 3 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