Super Micro Computer(US:SMCI)2025-10-14 14:44Summary of AI Server Industry Conference Call Industry Overview - The AI server market is approaching $300 billion, accounting for 72% of the total server TAM, with a growth rate of 46% driven primarily by generative AI [1][2][17] - By 2028, it is expected that over 80% of data center computing power will be used for inference rather than training [2][20] Key Challenges - AI servers face significant challenges in power consumption and heat dissipation, with single card power nearing 1,000 watts [1][4][3] - Liquid cooling technology is rapidly gaining traction to address these challenges, helping new data centers achieve a PUE below 1.2 [4][7] Market Dynamics - The AI server industry is experiencing a shift in competitive landscape, with OEM manufacturers like Dell and HPE having a gross margin of about 20%, while ODM manufacturers like Quanta and Foxconn hold nearly half of the market share [1][5][10] - Super Micro, as a quasi-ODM, offers deep customization and has a unique business model that allows for rapid product launches [11][12] Liquid Cooling Technology - Liquid cooling technology is becoming essential for AI servers due to increasing power density and the need for efficient heat management [6][7] - This technology is expected to significantly improve overall energy efficiency in new data centers [7] Competitive Characteristics - Different types of AI server manufacturers have distinct characteristics: - OEMs like Dell and HPE focus on traditional channels and support services [9] - ODMs like Quanta and Foxconn customize products for large clients, achieving market share through volume [9] - Quasi-ODMs like Super Micro provide flexible customization to meet client needs [9] Regional Dynamics - Taiwanese manufacturers are deeply integrated with North American cloud giants, which limits their gross margins to below 10% due to low brand premiums [10][19] - Domestic Chinese manufacturers, such as Inspur, leverage local Capex and policies to customize products for local internet giants [14][15] Future Trends - Edge computing is emerging as a new trend for AI inference, with domestic manufacturers having advantages in deployment capabilities [15] - The market for inference servers is expected to grow significantly, with a shift from training servers, which are more capital-intensive [20][21] Investment Considerations - Concerns about potential bubbles in computing power servers stem from over-reliance on large enterprise CAPEX, with many actual demands not being captured [22] - The AI wave has significantly impacted the valuation of server hardware companies, with some experiencing stock price increases of nearly 10 times [17][18] Performance Metrics - North American server manufacturers have seen a decline in performance, with SMCI's gross margin dropping from nearly 20% to around 9% due to increased competition and rising supply chain costs [19]