AI hardware isn't just NVIDIA—it's an 'AI factory'
With the sharp rise in AI-related capital expenditures, the market most readily notices chip suppliers like NVIDIA, while paying less attention to the fact that before chips reach data centers, they still need to be integrated into motherboards, memory, networking components, power supplies, cooling systems, and server racks. Super Micro Computer (NASDAQ: SMCI) is a key intermediary in this supply chain. The question is whether the rapid growth in AI server demand will enable the company to establish a technology-driven pricing premium or merely push it toward a high-revenue, low-margin hardware assembly business.
Super Micro’s first advantage is 'speed.' The company uses a modular Building Block architecture that allows it to quickly launch complete server systems tailored to new processors, storage, networking, and cooling specifications shortly after NVIDIA, AMD, or Intel release next-generation chips. For cloud service providers, AI chips are expensive; the earlier servers are deployed, the sooner they generate compute revenue—making delivery speed itself commercially valuable.
Liquid cooling is another potential moat. Power consumption and heat output from high-density AI racks have surged, pushing traditional air cooling close to its limits. Super Micro has integrated servers, racks, cooling distribution units, piping, software, and on-site services into a data center–level solution. The company claims its next-generation liquid cooling system can reduce power costs by up to 40% and total cost of ownership by up to 20% compared to air cooling. Even though actual savings vary by data center, the shift of liquid cooling from an optional feature to foundational infrastructure indeed helps the company move beyond simply 'selling machines' to 'delivering AI factories.'
Demand figures are also impressive. Fiscal year 2025 revenue grew 46.6% year-over-year to approximately $21.97 billion; the company subsequently projected fiscal year 2026 revenue of $38.9 billion to $40.4 billion. Third-quarter revenue for fiscal year 2026 reached $10.2 billion, more than doubling year-over-year, demonstrating that AI servers are not just a concept but already generating verifiable orders and deliveries.
However, revenue growth does not necessarily translate into improved profitability. Gross margin declined from 13.8% to 11.1% in fiscal year 2025, which the company attributed to competitive pricing to gain market share, intensifying competition, and shifts in customer and product mix. By the second quarter of fiscal year 2026, gross margin had fallen further to 6.3%, before rebounding to 9.9% in the third quarter. This highlights a key paradox of AI servers: although system prices are high, most of the value resides in GPUs, memory, and networking chips, leaving system integrators with limited profit margins.
The company expects its gross margin in the fourth quarter to rise to 15%–17%, but this improvement primarily stems from a favorable customer and product mix and should not be immediately viewed as structural. Large cloud customers place massive orders and wield stronger bargaining power. In fiscal year 2025, four customers each accounted for more than 10% of revenue; the more concentrated the order book, the more quarterly revenue, pricing, and payment terms are susceptible to influence by a few customers.
Another warning sign is cash flow. The company reported $6.6 billion in operating cash outflow in Q3 of fiscal year 2026, with total outflows of approximately $75.6 billion over the first nine months—reflecting the need to pre-purchase chips, build inventory, and extend credit to customers amid rapid expansion. If product transitions occur, orders are delayed, or customers cancel purchases, high-cost components could quickly become impaired. This growth model—where profits are recognized but cash has yet to be collected—requires more caution than simply looking at revenue growth alone.
Governance risks also cannot be ignored. Super Micro previously delayed filing its financial reports after its auditors raised concerns about governance, transparency, and internal controls. As of June 30, 2025, the company still acknowledged material weaknesses in internal control over financial reporting. Even if the financial statements are eventually audited, the valuation discount investors apply will not disappear solely due to a single quarter of earnings recovery.
Therefore, Super Micro is neither a low-value assembly shop nor has it yet proven capable of sustaining software-like high gross margins. Its investment value hinges on three factors: whether liquid cooling and rack-scale solutions can enhance value-added offerings, whether gross margins can remain stable across varying customer mixes, and whether operating cash flow can catch up with reported earnings. AI servers undoubtedly represent a high-growth market—but only when every dollar of revenue translates into healthier gross profit and cash generation will Super Micro truly evolve from a mere enabler of the AI wave into a long-term beneficiary.
Yee Hap Holdings Investor Relations Department
(Chips & Compute Power Series #81)
Risk Disclaimer: The above content only represents the author's view. It does not represent any position or investment advice of Futu. Futu makes no representation or warranty.Read more
Comments
to post a comment
1
1
