Who makes NICs / SuperNICs for AI data centers

Who competes here

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

Seen in real machines

NICs / SuperNICs is named in 16 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
  • ConnectX-8 SuperNICs, BlueField-3 DPUs — Nvidia
• AMD Helios (MI455X) · rack teardown
  • Pensando "Vulcano" AI NIC (800 Gbps scale-out, PCIe Gen 6, OCP form factor, UALink to CPU/GPU) — AMD
• El Capitan · supercomputer
  • Slingshot-11 network interfaces (4 x 200 Gb/s per node, one per APU, on PCIe 5.0 x16) — HPE
• Frontier · supercomputer
  • Network interface (four HPE Slingshot 200 Gb/s NICs per node, 100 GB/s injection bandwidth) — HPE
• Aurora · supercomputer
  • HPE Cassini NIC (200 Gb/s, 2 ports per NIC, 8 NICs per node, PCIe Gen4 host interface) — HPE
• JUPITER · supercomputer
  • InfiniBand adapters (ConnectX-7, NDR200: 4 per Booster node, 1 per Cluster node) — Nvidia
• Alps · supercomputer
  • Network interface cards (HPE Cassini, Slingshot-11 at 200 Gb/s; 4 per GH200 node, 2 per Rome node) — HPE
• LUMI · supercomputer
  • Network interface (200 Gb/s HPE Slingshot NIC; four per LUMI-G node, one per MI250X module, 800 Gb/s injection) — HPE
• Leonardo · supercomputer
  • Network adapters (ConnectX-6 HDR100: two dual-port cards per Booster node, one single-port card per Data Centric node) — Nvidia
• Isambard-AI · supercomputer
  • Slingshot network interface (HPE Cassini NIC, 200 Gb/s, one per superchip, 4 per node) — HPE
• Venado · supercomputer
  • Network interface cards (HPE Cassini NIC in the Slingshot 11 interconnect) — HPE
• Perlmutter · supercomputer
  • Network interface (200 Gb/s HPE Slingshot "Cassini" NIC; four per GPU node, one per CPU node) — HPE
• Eagle (Microsoft Azure) · supercomputer
  • InfiniBand adapters (ConnectX-7, 400 Gb/s NDR, one per GPU, 8 per VM) — Nvidia
• Eos · supercomputer
  • Network adapters (eight single-port ConnectX-7 InfiniBand cards for the compute fabric, two dual-port ConnectX-7 cards for storage and in-band management, per DGX H100) — Nvidia
• Fugaku · supercomputer
  • Tofu Interconnect D (on-die network controller and router, 6D mesh/torus, 6 network interfaces at 6.8 GB/s each, about 216,864 cables) — Fujitsu
• MareNostrum 5 · supercomputer
  • InfiniBand adapters (ConnectX-7 NDR200, one per GPU in the accelerated nodes; one per two nodes in the general-purpose partition) — Nvidia

Latest feed

• [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.
• Taiwan raising $4 billion to build industrial park in Texas to support US chipmaking supply chain — Tom's Hardware, 2026-10-16
  Taiwan plans to build a $4 billion industrial park in Texas to help smaller Taiwanese companies expand their presence supporting major chipmakers.
• Intel says AI 'gold rush' is over, pushes prompt routing across open hardware — DigiTimes, 2026-10-15
  Intel announced at the OCP Global Summit that the era of massive, centralized AI models is shifting toward distributed prompt routing across open hardware.
• Qotom Q30952UE Review the New Black Box for 10G Networking — ServeTheHome, 2026-09-21
  The Qotom Q30952UE is a new networking device that features two SFP+ 10G ports and six 2.5GbE ports.
• Compal takes Compal Broadband Networks private to fold networking into its edge AI stack — DigiTimes, 2026-09-18
  Compal Electronics is buying its networking arm, Compal Broadband Networks, to integrate the router and gateway business fully into its edge AI hardware stack.
• AI server exports to US slow; electronics parts jump 1.5 times in August — DigiTimes, 2026-09-13
  Taiwan reported that electronic parts exports reached a record high of $32.22 billion in August 2026, with integrated circuits and printed circuit boards seeing significant growth.
• Mature-node chip scramble may persist through 2028 as supply chain management becomes bargaining leverage — DigiTimes, 2026-09-09
  Supply chain constraints for advanced semiconductor processes, packaging, and mature-node chips are expected to persist for several years.
• Strategy Paper Urges Canada to Add Semiconductors to AI Strategy — EE Times, 2026-09-08
  A strategy paper recommended that Canada must integrate semiconductors into its national AI strategy to support its domestic hardware capabilities.
• Renewable power shortage puts Taiwan's chip cluster under scrutiny — DigiTimes, 2026-09-08
  Taiwan's semiconductor and AI infrastructure expansion is intensifying pressure on chip manufacturers to adopt renewables, due to prohibitive green power costs and severe power supply constraints.
• China interconnect chipmaker Montage enters Samsung, SK Hynix CXL 3.2 supply chains and begins DDR6 development — DigiTimes, 2026-09-04
  Montage Technology is expanding its role in AI data center infrastructure by developing products for CXL 3.2, DDR6, and PCIe, moving beyond traditional DDR interfaces.