Last updated: 2026-07-01
Supermicro designs AI server hardware spanning edge systems at 65W TDP up to 8-GPU data center nodes, generating $12.68 billion in quarterly revenue. NASDAQ: SMCI.
Supermicro is a NASDAQ-listed (SMCI) AI server hardware manufacturer with $12.68 billion in quarterly revenue, designing compute systems from 65W edge nodes to 8-GPU data center racks with direct liquid cooling. Servers support NVIDIA H100, H200, and B200 GPUs and are deployed by hyperscalers, cloud providers, and enterprise data centers worldwide for AI training and inference workloads.
Supermicro (Super Micro Computer, Inc.) is a San Jose-based server and storage systems manufacturer founded in 1993 by Charles Liang. The company went public on NASDAQ (SMCI) in 2007 and reported record quarterly revenue of $12.68 billion in Q4 2025, making it one of the largest suppliers of AI infrastructure hardware globally alongside Dell and HPE. Supermicro's AI hardware portfolio spans three deployment tiers. Data center systems support up to 8 NVIDIA RTX PRO Blackwell GPUs per 4U or 5U node, optimized for large-scale inference, training, and virtualization. The Super AI Station is a deskside AI supercomputer powered by the NVIDIA GB300 Grace Blackwell Ultra Superchip, delivering over 5x AI PFLOPS with 775GB coherent memory for local large-model development. Edge AI systems launched in April 2026 use AMD EPYC 4005 processors in mini-1U, short-depth 1U, and slim tower form factors, with TDPs as low as 65W for retail, manufacturing, and healthcare deployments. The company targets IT infrastructure teams at enterprises and cloud providers, AI research labs, edge computing operators in space-constrained environments (retail, manufacturing, healthcare), and government or defense organizations requiring on-premises AI. Supermicro's key advantage over competitors is build-to-order speed: its modular building block architecture lets customers configure and receive systems faster than Dell or HPE custom orders, important for AI teams under time pressure. Pricing is quote-based with no published list prices. Systems are purchased through Supermicro directly, authorized resellers, or partner channels including Thinkmate and Uvation. Gold Series pre-configured servers carry shorter lead times and are priced competitively for standard workloads. The company offers financing options through third-party leasing partners for large enterprise deployments. In 2026, Supermicro expanded its portfolio to include NVIDIA BlueField-4 STX storage servers for AI inference acceleration, Context Memory (CMX) servers for agentic long-context workloads, and NVIDIA RTX PRO Blackwell Server Edition GPU systems. The stock trades at a governance discount due to a 2024-2025 accounting investigation, despite record revenue growth.
Quote-based pricing through Supermicro direct sales and authorized resellers. No published list prices. Gold Series pre-configured systems available with shorter lead times. Enterprise financing available via third-party leasing.
Supermicro is the trading name of Super Micro Computer, Inc., a NASDAQ-listed (SMCI) AI server hardware manufacturer founded in 1993 by Charles Liang, who remains President, CEO, and Chairman as of 2026. The company is headquartered in San Jose, California, and is a component of the S&P 500. Supermicro designs and builds compute systems spanning 65-watt edge nodes up to 8-GPU data center racks with direct liquid cooling, supporting NVIDIA H100, H200, B200, and the newer Blackwell Ultra GB300 NVL72 and HGX B300 platforms. Fiscal 2025 revenue reached $21.97 billion, and the company is guiding for at least $33 billion in fiscal 2026 revenue, with Q2 FY2026 revenue alone hitting $12.7 billion driven by AI data center demand.
Supermicro does not publish list prices; all server, rack, and data center systems are quote-based and sold through direct sales or channel partners depending on configuration, GPU count, and cooling option. Pricing varies enormously by build: an edge AI node drawing 65 watts costs far less than an 8-GPU rack with direct liquid cooling built around NVIDIA's GB300 NVL72 or HGX B300 platforms, which are aimed at hyperscale data centers. Customers typically work with a Supermicro sales engineer or reseller to configure a system around CPU, GPU (NVIDIA H100/H200/B200/Blackwell Ultra or AMD MI350/355), memory, storage, and cooling, then receive a custom quote. There is no self-serve checkout or published per-unit pricing on Supermicro's site.
Supermicro's main differentiator is being a first-wave integrator for the newest NVIDIA and AMD AI accelerators, announcing volume shipments of NVIDIA Blackwell Ultra GB300 NVL72 and HGX B300 systems alongside AMD MI350/355 platforms in 2025-2026, often ahead of larger rivals like Dell and HPE on time-to-market. The company also builds its own direct liquid cooling (DLC) systems in-house across its product range, rather than relying entirely on third-party cooling vendors, which lets it ship 8-GPU liquid-cooled racks as a more complete package. Supermicro's portfolio spans more than ten product types built around the latest NVIDIA technologies, from 65-watt edge systems to full rack-scale, plug-and-play data center solutions, giving it breadth from edge AI to hyperscale clusters under one brand.
Dell is a larger, more diversified hardware vendor with a broader enterprise IT portfolio beyond AI servers, and has been reported as a beneficiary when customers shifted large AI server orders away from Supermicro during 2024's accounting controversy, including a reported $6 billion order from xAI moving to Dell. Supermicro's advantage is speed: it has repeatedly been first to ship volume systems built on new NVIDIA platforms, including Blackwell Ultra GB300 NVL72 systems in 2025-2026, and its direct liquid cooling is built in-house. Dell tends to win on enterprise relationships, support infrastructure, and risk-averse procurement teams after Supermicro's 2024 delayed financial filings and auditor (Ernst & Young) resignation raised governance concerns. Pick Supermicro if you need the newest NVIDIA platform shipped fastest; pick Dell if vendor stability and existing enterprise contracts matter more.
No, Supermicro sells physical server hardware and data center systems, so there is no free tier, software trial, or subscription in the way SaaS tools are evaluated. Prospective buyers typically engage with Supermicro's sales team or an authorized reseller to discuss requirements (GPU type, rack density, cooling method, target workload) and receive reference architectures or proof-of-concept configurations before committing to a purchase order. Some configurations may be evaluated through partner data centers or cloud providers that have already deployed Supermicro systems, letting an organization benchmark performance on Supermicro hardware (for example, NVIDIA H100/H200/B200 nodes) before buying their own racks. There is no published 'free plan' because the product is capital equipment, not software.
Supermicro is best for hyperscalers, cloud providers, AI labs, and enterprise data center teams that need to deploy NVIDIA H100/H200/B200/Blackwell or AMD MI350/355 GPU systems quickly, especially when direct liquid cooling and rack-scale plug-and-play deployment matter; fiscal 2025 revenue of $21.97 billion and FY2026 guidance of at least $33 billion reflect strong demand from this segment. Organizations that prioritize long-term vendor stability and clean financial governance above time-to-market should weigh Supermicro's 2024 accounting delays and auditor resignation (Ernst & Young) before committing to large multi-year contracts, even though the company filed delinquent reports and resumed normal reporting in 2025. Small businesses or teams needing one or two general-purpose servers are generally better served by traditional enterprise vendors with smaller minimum order sizes and broader retail availability.
Yes, direct liquid cooling (DLC) is a core part of Supermicro's AI server lineup, built in-house rather than sourced entirely from third parties. Supermicro ships 8-GPU data center racks with DLC designed specifically for high-thermal-design-power chips like NVIDIA's H100, H200, B200, and the newer Blackwell Ultra GB300 NVL72 and HGX B300, which generate far more heat than air cooling can efficiently dissipate at rack scale. In 2025, Supermicro announced 1.1 ExaFLOPS of AI performance from its Blackwell Ultra DLC systems as part of its volume shipment announcements, positioning liquid cooling as a way to pack more GPU density per rack without thermal throttling. Air-cooled configurations remain available for lower-power edge systems (around 65 watts), but DLC is the default recommendation for high-density GPU deployments.
Yes, in 2024 Supermicro delayed filing its annual report for fiscal year 2024 by roughly 50 days after short-seller Hindenburg Research alleged accounting irregularities, undisclosed related-party transactions involving CEO Charles Liang's family, and quality issues with shipped products. The company's auditor, Ernst & Young, resigned during this period, citing governance and transparency concerns, and Supermicro was briefly at risk of Nasdaq delisting for failing to file on time. By early 2025, Supermicro filed its delinquent fiscal 2024 annual report along with overdue Q1 and Q2 fiscal 2025 quarterly filings, returning to regular reporting. Despite the controversy, Supermicro's reported fiscal 2025 revenue of $21.97 billion and FY2026 guidance above $33 billion suggest enterprise AI customers largely continued placing orders through the resolution.
Supermicro designs AI server hardware from edge systems (65W TDP) to 8-GPU data center nodes, with $12.68B quarterly revenue and NASDAQ: SMCI listing.
Super Micro Computer, Inc.