Morgan Stanley Re-Lists Nvidia (NVDA.US) as Top Semiconductor Pick: Industry Bottleneck Shifts to Data Center Infrastructure, SpaceX and Amazon Partnership Opens Growth Opportunities
Morgan Stanley emphasizes that Nvidia's price-to-earnings (P/E) ratio based on the fiscal year 2028 earnings forecast is only 15x, making its valuation highly attractive. Even if the P/E ratio does not expand, strong investment returns can still be achieved through earnings growth.
According to FinanceAPP, Morgan Stanley has released its latest research report, reinstating chip giant Nvidia (NVDA.US) as its top pick in the semiconductor sector. The firm maintained its "Overweight" rating and set a price target of $300. The firm believes that the industry's bottleneck is rapidly shifting from semiconductor production constraints to the speed of new data center construction and financing models. Leveraging its product architecture, global customer ecosystem, and financing capabilities, Nvidia is poised to capture industry dividends. Morgan Stanley is also optimistic about the incremental business opportunities brought by Muse intelligent agents and new customer collaborations, suggesting upside potential for the company's valuation. Morgan Stanley emphasized that Nvidia’s current price-to-earnings ratio based on 2028 fiscal year earnings estimates is only 15x, making the valuation particularly attractive. Even without multiple expansion, solid performance growth should deliver robust investment returns.
Core Industry Logic: Bottleneck Shift and High-Density Compute Architecture
Morgan Stanley points out that while the market has focused on semiconductor chip production constraints over the past year, the industry's core issue has shifted to the pace of data center construction and financing capacities—precisely where Nvidia's ecosystem excels. Land, electricity, and data center space have become the main limitations. Under these conditions, the higher the token output per gigawatt of compute power, the stronger the competitiveness. Presentation slides by Nvidia show that the next-generation Feynman (2028) architecture achieves up to 25x higher token output compared to the Blackwell architecture. The goal is to significantly raise revenue per gigawatt—from $4 billion to well over $5 billion—mitigating the pressure from rising per kilowatt costs.
The report also notes that total AI compute demand has exceeded the current infrastructure's carrying capacity, so there will not be a short-term chip supply shortage. The rise of intelligent agents (Agent) is expected to drive CPU demand, but the majority of incremental capital expenditure will still be allocated to GPUs. In Muse-like intelligent agent scenarios, each CPU core is only minimally engaged in framework processing—meaning a single server equipped with hundreds of CPU cores can serve massive numbers of users, while a single intelligent agent user can generate tokens tenfold compared to an ordinary user, giving rise to massive demand for GPU inference. Nvidia is confident in its CPU shipment growth plans, aiming to increase from $20 billion in 2026 to $40 billion in 2027, with a little over 50% of this growth coming from main node CPUs within AI racks, and the remainder from standalone CPU racks.
Customer & Business Increment: SpaceX and Amazon Deepen Collaboration, Overseas Cloud Ecosystem Expansion
On the customer front, SpaceX and Amazon continue to deepen their collaboration with Nvidia, serving as important growth drivers. SpaceX plans to reach about 2 GW of data center capacity by the end of 2026, and increase to 5-10 GW by the end of 2027. It is estimated that if SpaceX spends $132 billion on AI capital expenditure in 2027, with 79% allocated to actual deployment (in line with Q1 2026), every 1GW installed capacity will contribute about 5% of Nvidia’s revenue in FY2028. Amazon AWS is expanding cooperation, adding 2 million GPUs and Vera processors during 2027-2028, although the Vera processor is expected to be launched only by the end of 2027. Morgan Stanley’s internet research team predicts that Amazon’s spending on Nvidia is expected to reach $40 billion in 2026 and $78 billion in 2027.
Nvidia has a broad customer base, with its official website listing 80 cloud partners—55 of which are outside the U.S.—allowing the company to penetrate overseas markets that traditional large-scale cloud providers find difficult to reach. Nvidia benefits from local compute power infrastructure, sovereign AI businesses, revenue-sharing agreements, and expanded software service income. Nvidia is also facilitating $500 billion in non-Nvidia funding going into AI data centers, securing component and electricity resources in advance, building competitive barriers on its balance sheet, and delivering on its commitment to return 50% of cash flow to shareholders.
Performance, Gross Margin, and Multi-Scenario Valuation Forecasts
Morgan Stanley believes Nvidia’s FY2028 revenue guidance has room for upward revision—the previously forecasted 70% year-on-year growth was based on a supply-constrained assumption, but real demand growth is close to 100%. The firm estimates FY2028 gross margin will be a solid baseline at 72%-73%. Cost increases of raw materials like DRAM can be offset by raising product prices or lowering hardware specs. Based on a simplified gross margin model using the Rubin architecture, with a baseline gross margin of 75%, Nvidia could raise prices by 15% to absorb a 30% increase in cost of sales; reducing HBM specs could offset a 44% increase in cost of sales.
The firm provides three scenario forecasts:
Optimistic scenario: FY2027 EPS of $16, price target of $350. Drivers include ramp-up from CPUs, new Groq products, data center revenues exceeding expectations, high-margin AI software business, revenue sharing, and the expansion of AI PCs into new markets.
Base-case scenario: FY2027 EPS of $15.01, price target of $300, implying around a 20x price-to-earnings ratio; with 2026 and 2027 revenue expected to grow by 88% and 70%, respectively.
Bearish scenario: FY2027 EPS of $13, price target of $160. Risks include downward revisions in growth expectations, chip supply coming online too fast, weakening of inference demand, competitors taking market share, as well as negative impacts from export controls and tariffs, etc.
Investment View
Morgan Stanley reinstates Nvidia as its top pick. Nvidia is in the early phase of a new product cycle, with compute demand strongly aligned to the company’s competitive advantages, and the Vera product expected to ramp up soon. At only 15x FY2028 P/E, the stock remains undervalued; a resurgence of enthusiasm for AI investment could drive valuation expansion. Even with no multiple expansion, sustained growth should result in strong performance. Although the market has recently been flooded with coverage related to CPUs amid the Muse boom and Nvidia’s performance has been relatively subdued in the short term, Morgan Stanley believes consumer devices will continue to provide long-term demand for AI applications and tokens, and that Nvidia is likely to outperform the market going forward.
Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.
You may also like
NEAR Loses Critical $5 Level After Brutal $20M Long Squeeze: More Pain Ahead?
If US Treasury yields continue to rise, what will Washington do next?
The Treasury has maintained liquidity by increasing the issuance of short-term Treasury bills and conducting small-scale buybacks. Some advocate for reducing expenditures to address the debt burden. Political constraints tilt the risk toward inflation, which harms bondholders' interests. Karen Brettell, Reuters, October 5 - The cost of borrowing for the U.S. government is rising, while it has almost exhausted straightforward ways to control those costs. Long-term Treasury yields are now near their highest levels in two decades, and the causes don't appear to be temporary. Washington is issuing large amounts of government debt to cover a fiscal deficit that shows no signs of shrinking. Inflation is cooling only slowly. Moreover, while the real estate and automotive sectors are struggling, the artificial intelligence investment boom is keeping the economy robust enough to prevent interest rates from falling. As a result, with over $40 trillion in debt, annual interest payments alone amount to around $1 trillion. Washington has options, from relying more on short-term borrowing to, in the most extreme case, having the Federal Reserve cap long-term yields. The more policymakers resort to such measures, the higher the risk of fueling inflation, potentially causing more pain for bondholders in the future. Torsten Slok, Chief Economist at Apollo Global Management, noted that for every $5 the government collects in taxes, $1 goes to service the debt. "That's a very, very high number, and it's only going to grow." U.S. President Donald Trump said in a September 28 interview with Time magazine that debt can be repaid through economic growth or inflation. But if these methods fail, the Treasury has other options ranging from moderate to radical. At present, the Treasury is increasingly relying on issuing short-term bills and conducting small-scale buybacks of old debt to help boost market liquidity. In a worse scenario, the next step would require Fed intervention. One method is large-scale purchases of long-term bonds, akin to 1961's "Operation Twist", another is directly capping long-term yields—a measure not used by the U.S. since World War II. The more aggressive the measures, the more they can suppress rates, but also the greater the risk of spurring inflation. “We are getting to a point where it's clear the government is uncomfortable with current rate levels," said Jeffrey Gundlach, CEO of DoubleLine Capital, at a recent investment event. Operation Twist Historically, the next escalation would likely be a full-scale reactivation of "Operation Twist." Launched in 1961, this strategy involved selling short-term Treasuries and purchasing long-term ones to flatten the yield curve. Implementing a substantial twist would require the Fed's assistance, but the Fed may stand pat unless there is an obvious financial emergency. Slok said that without the Fed's balance sheet, the Treasury has very limited tools for lowering rates. However, Fed Chair Kevin Warsh has criticized holding large amounts of government debt and other securities, arguing that massive bond buying blurs the line between monetary policy and government debt management. He has called for a new agreement between the Treasury and the Fed, under which the Fed Chair and Treasury Secretary would communicate publicly about the Fed's balance sheet and the Treasury’s debt issuance plans. Yield Curve Control If Operation Twist–style purchases don't work, the next move would be explicit yield curve control. In this scenario, the central bank commits to buying an unlimited amount of government debt to keep long-term rates under a set cap. From 1942 until the 1951 Treasury-Fed Accord, the Fed capped long-term Treasury yields at 2.5% to help fund WWII and the postwar recovery. The Bank of Japan implemented a version of this policy from 2016 to 2024. By artificially lowering rates, yield curve control can ease the political pressure of fiscal deficits. But it only works as long as investors aren't worried about being repaid with dollars devalued by inflation. Once that confidence is shaken, bond-buying meant to suppress rates only fuels the inflation it's designed to conceal. Veronique de Rugy, Senior Research Fellow at the Mercatus Center at George Mason University, said that ultimately, the only way to solve the debt problem is by cutting expenditures. “Congress needs to implement fiscal consolidation—in other words, austerity. The Fed cannot do this alone.” Divergent Paths John Higgins, Chief Economic Advisor at Capital Economics, notes that since World War II, the U.S. has only significantly reduced its debt-to-GDP ratio twice, but bondholders' experiences differed substantially each time. After the war, the debt-to-GDP ratio fell from about 106% in 1946 to 23% in 1974, while the 10-year Treasury yield climbed from 2.2% to 7.5%. In the 1990s, the ratio declined from 48% to 32%, and yields fell as well. What made the difference? After WWII, restr
MercadoLibre stock closes up 0.68% at $1,696.56, still below every key moving average
