Morgan Stanley re-adds NVIDIA Corporation (NVDA.US) as top semiconductor pick; industry bottleneck shifts to data center infrastructure; SpaceX and Amazon.com, Inc. partnership opens up growth space.

GMTEight APP has learned that Morgan Stanley has released its latest research report, re-adding chip giant NVIDIA Corporation (NVDA.US) as its top semiconductor pick, maintaining an "Overweight" rating with a price target of $300. The institution believes the industry bottleneck is rapidly shifting from semiconductor capacity constraints to the pace of new data center construction and financing models, and NVIDIA Corporation is fully capturing industry tailwinds through its product architecture, global customer ecosystem, and financing support capabilities. It is also bullish on incremental growth from Muse intelligent agents and new customer partnerships, with upside potential in the company's valuation. Morgan Stanley emphasizes that NVIDIA Corporation currently trades at only 15x forward P/E on FY2028 earnings estimates, making the valuation highly attractive. Even without multiple expansion, earnings growth alone can deliver solid investment returns. Core Industry Logic: Bottleneck Shift, High-Density Compute Architecture Breaks Through Morgan Stanley points out that over the past year the market focused on semiconductor chip capacity constraints, but the current core industry contradiction has shifted to data center construction speed and financing capabilityexactly where NVIDIA Corporation's ecosystem advantages lie. Land, power, and machine room space have become the biggest development constraints. Under this premise, the higher the number of tokens produced per gigawatt of compute, the stronger the competitiveness. NVIDIA Corporation's demonstration slides show that its next-generation Feynman (2028) architecture delivers up to 25x higher token output compared to the Blackwell architecture, targeting a substantial increase in revenue per gigawattraising revenue per gigawatt from $40 billion to well above $50 billionalleviating upward pressure on cost per kilowatt. The institution also notes that total AI compute demand exceeds the carrying capacity of existing infrastructure, and there will be no chip supply shortage in the short term. The rise of intelligent agents (Agent) will drive CPU demand, but the vast majority of incremental capital expenditure will still flow to GPUs. In Muse-type intelligent agent scenarios, a single CPU core is only used for framework processing a small fraction of the time, and a single several-hundred-core CPU can serve a massive number of users; meanwhile, a single intelligent agent user can generate ten times the daily tokens of an ordinary user, which will spur massive GPU inference demand. NVIDIA Corporation is confident in its own CPU shipment growth plan, targeting an increase from $20 billion in 2026 to $40 billion in 2027, and believes that slightly more than half of that growth will come from main node CPUs inside AI racks, with the remaining growth from standalone CPU racks. Customers and Business Increment: SpaceX, Amazon.com, Inc. Deepen Cooperation, Overseas Cloud Ecosystem Expands On the customer front, SpaceX and Amazon.com, Inc. continue to deepen cooperation with NVIDIA Corporation, becoming important growth engines. SpaceX plans to build approximately 2GW of data center capacity by the end of 2026, rising to 5-10GW by the end of 2027; calculations assume that of SpaceX's $132 billion AI capital expenditure in 2027, 79% will be used for actual deployment, consistent with the ratio in Q1 2026. For every 1GW of installed capacity SpaceX brings online, it will contribute approximately 5% of NVIDIA Corporation's revenue in FY2028. Amazon.com, Inc. AWS is expanding cooperation, adding 2 million GPUs and Vera processors in 2027-2028, with Vera processors expected to launch only by the end of 2027. Morgan Stanley's internet research team forecasts that Amazon.com, Inc.'s spending on NVIDIA Corporation could reach $40 billion and $78 billion in 2026 and 2027, respectively. NVIDIA Corporation has a broad customer base, with 80 cloud partners on its official website, 55 of which are located outside the United States, enabling it to penetrate overseas markets that traditional large cloud vendors find difficult to cover, benefiting from national local compute infrastructure and sovereign AI businesses, implementing revenue-sharing agreements, and expanding software service revenue. NVIDIA Corporation has also mobilized $500 billion in non-NVIDIA Corporation capital investment into AI data centers, locking in components and power resources in advance, building competitive barriers through its balance sheet, and fulfilling its commitment to return 50% of cash to shareholders. Performance, Gross Margin, and Multi-Scenario Valuation Analysis Morgan Stanley believes there is room for upward revision to NVIDIA Corporation's FY2028 revenue guidance. The previous 70% year-over-year growth guidance was based on supply-constrained assumptions, while true demand growth is close to 100%. The institution judges that a 72%-73% gross margin for FY2028 is a solid floor. Upward cost pressure from raw materials such as DRAM can be hedged through two methods: product price increases and downgrading hardware specifications. Building a simple gross margin model based on the Rubin architecture transition, using a 75% gross margin baseline, NVIDIA Corporation can absorb a 30% increase in cost of sales with a 15% price increase; downgrading HBM specifications can absorb a 44% increase in cost of sales. The institution provides three scenario forecasts: Bull case: FY2027 EPS of $16, price target of $350. Driven by volume ramp of new CPU and Groq products, data center revenue exceeding expectations, high-margin AI software business and revenue sharing taking effect, and AI PC expanding market space. Base case: FY2027 EPS of $15.01, price target of $300, corresponding to approximately 20x P/E; revenue growth of 88% and 70% expected in 2026 and 2027, respectively. Bear case: FY2027 EPS of $13, price target of $160. Risks include downward revision of growth expectations, supply releasing too quickly, weakening inference demand, competitors grabbing share, and negative impacts from export controls and tariffs. Investment View Morgan Stanley re-adds NVIDIA Corporation as its top pick. NVIDIA Corporation is currently in the early stage of a new product cycle, compute demand is highly matched with the company's competitive advantages, and Vera products are expected to accelerate to market. FY2028 P/E is only 15x, which is low; a recovery in AI investment enthusiasm could bring valuation expansion, and even without valuation re-rating, the stock is still expected to deliver good performance on sustained earnings growth. Although recently, amid the Muse hype, a large number of CPU-related reports have appeared in the market and NVIDIA Corporation's near-term performance has been relatively weak, Morgan Stanley believes consumer devices will continue to provide long-term momentum for AI applications and token demand, and the stock is expected to outperform the broader market going forward.