Super Micro Computer (SMCI)

AI server systems (Supermicro) — rack-scale GPU servers + liquid cooling

AI exposure 75% — derived from reported segment revenue, not asserted.

Where it sits

Physical Infrastructure → Storage & Server Systems · figures rebased to FQ4 FY2026, reported 2026-08-11

AI/GPU servers + rack-scale solutions (NVIDIA/AMD, liquid-cooled DLC) — AI compute systems leads 2 businesses in this sub-tier that together are 100% of revenue, so Storage & Server Systems is effectively the whole company rather than one leg of it.

Reported segments

• Server & storage systems (by type) — 100% of revenue

Super Micro ($39.1B FY2026, +78% from $22.0B) — server/storage systems, AI-server leader. Single reported segment. AI platforms were ~60% of Q4 FY2026 revenue (against >80% in Q3, on GPU deal timing) with more than $60B of new orders booked in the quarter, roughly 70% pure GPU-based AI and 30% CPU or CPU-based AI — management expects >80% AI again on backlog. Mix is now two-sided: enterprise/channel $5.6B (50% of Q4 sales, +172% YoY) vs OEM/large data centre $5.5B (50%, -26% QoQ). Nine customers each exceeded $1B of FY2026 revenue, up from four. Primary B2 Storage & Server Systems.…

How the 75% is built

• Server & storage systems (by type) — 100% of revenue × 75% AI-driven (estimate). Disclosed AI-platform mix swung between ~60% (Q4 FY2026) and >80% (Q3 FY2026) on deal timing rather than demand; the backlog composition and management's >80% expectation put the through-cycle share around three-quarters

Investment read

Moat. Speed-to-market and configurability on each new NVIDIA generation, plus direct-liquid-cooling incumbency and manufacturing capacity above 6,000 racks a month — a timing advantage rather than a structural one. Dell and HPE run comparable rack programmes with balance sheets SMCI cannot match, and the deliberate push into enterprise (now 50% of sales) is an attempt to build a margin floor the hyperscaler business has never provided.

Bottleneck / pricing power. No pricing power, and Q4 FY2026 is the cleanest proof yet: non-GAAP gross margin printed 17.6% against 8.2-8.4% guided — and management immediately guided Q1 back to 10.4-10.8%. The swing across four quarters is 6.3% to 17.6% to a guided ~10.5%, driven entirely by which customer and product mix lands in which quarter. Roughly 75% of the Q4 beat was mix and contract deferrals, not price. Component cost and NVIDIA allocation, the two largest levers, sit outside the company.

Role in the AI stack. The AI-server-integration layer — high-beta, low-moat.

Connected companies

• NVIDIA (NVDA) — SMCI is supplied by: GPUs for Supermicro AI server/rack systems
• Advanced Micro Devices (AMD) — SMCI is supplied by: Instinct GPUs + EPYC for AI servers
• Microsoft (MSFT) — SMCI supplies: Azure-certified edge/HCI hardware ecosystem — no public evidence Microsoft buys SMCI AI racks directly
• Dell Technologies (DELL) — SMCI competes with: AI servers — Supermicro vs Dell/HPE
• Hewlett Packard Enterprise (HPE) — SMCI competes with: AI rack/server systems head-to-head
• CoreWeave (CRWV) — SMCI supplies: liquid-cooled rack-scale GPU systems; CoreWeave was reportedly SMCI's largest customer (~22% of revenue FY2024)
• Meta Platforms (META) — SMCI supplies: AI servers for Meta
• Oracle (ORCL) — SMCI supplies: GPU rack systems for OCI (GB300 NVL72 order placed then reportedly cut in 2026)
• Lattice Semiconductor (LSCC) — SMCI is supplied by: FPGAs for AI-server rack management
• Alcoa (AA) — SMCI is supplied by: primary aluminum feedstock that fabricators convert into server chassis, rails and heat sinks

Others in Storage & Server Systems

• Seagate Technology (STX)
• Western Digital (WDC)
• NetApp (NTAP)
• Everpure (fmr Pure Storage) (P)
• Dell Technologies (DELL)
• Hewlett Packard Enterprise (HPE)

All 174 companies · What moved this quarter · How the map is built