Core Viewpoint - Morgan Stanley anticipates that NVIDIA's Q3 revenue will exceed market expectations of approximately $55 billion, with Q4 guidance reaching between $63 billion and $64 billion, significantly higher than the consensus estimate of $61.5 billion. The primary constraint on growth is supply chain capacity rather than demand, which is expected to persist until 2026 [1][3][5]. Supply Chain Capacity - The report emphasizes that supply chain capacity is the sole limiting factor for NVIDIA's growth, as demand for AI computing power continues to significantly outstrip supply. Major customers, including hyperscale cloud service providers and AI research labs, are still facing computing power shortages [5][8]. - NVIDIA's Blackwell/Blackwell Ultra rack shipments are projected to have increased by approximately 50% quarter-over-quarter, reaching around 10,000 racks in Q3. This growth momentum is expected to continue into Q4 [4][9]. - For the entire fiscal year 2026, NVIDIA's total rack shipments are expected to reach between 28,000 and 30,000 units, although supply chain capacity will remain a key limiting factor until 2026 [9][11]. Long-term Outlook - Morgan Stanley provides an optimistic long-term signal, suggesting that NVIDIA's supply chain has the capacity to support rack shipments doubling year-over-year by FY2027, reaching between 60,000 and 70,000 racks. The backlog of orders has already exceeded 70,000 racks, indicating strong future growth potential [11][12]. Key Focus Areas for Investors - Investors are advised to pay close attention to management's commentary on four key long-term variables during the upcoming earnings call: 1. The ramp-up trajectory of Blackwell/Blackwell Ultra capacity entering the first half of FY2027 [14] 2. The sustainability of AI spending, with expectations of ample funding in the AI sector until 2030 [15] 3. The impact of power limitations, as approximately 120 GW of data center power capacity is expected to come online globally over the next five years [16] 4. The influence of component cost inflation on gross margins, particularly regarding rising prices for memory and chips [17] Gross Margin Analysis - The report details that the rising prices of LPDDR memory pose a greater pressure point than HBM memory. Although HBM4 is expected to see a 30-40% increase in average selling price, NVIDIA can incorporate these costs into the pricing of its next-generation Rubin platform. However, for LPDDR memory, the company may have to accept variable market prices, limiting the potential for further gross margin improvement [18][19].
摩根大通:英伟达业绩超预期已是板上钉钉,市场最关心的是Blackwell产能爬坡速度