Morgan Stanley Analysis: Nvidia Has 42% Upside, Where Will Growth Come From?

TL;DR
· Morgan Stanley maintains an "overweight" rating on Nvidia with a target price of $288, approximately 42% higher than the $202.78 share price used in the report, and continues to list it as a semiconductor favorite.
· The optimistic outlook is driven by increased market share in AI labs, expansion by major cloud providers, and simultaneous growth in enterprise, sovereign AI, and new AI cloud customers.
· The $288 target primarily relies on earnings upgrades; if data center growth slows, self-developed chip replacements expand, or export restrictions intensify, the bear market scenario could drop to $160.

After completing a non-deal roadshow with Nvidia CEO Jensen Huang, CFO Colette Kress, and the investor relations team, Morgan Stanley maintains its "overweight" rating, $288 target price, and preferred position in the semiconductor industry. Based on the report's closing price of $202.78 on July 9, the target price corresponds to a potential upside of about 42%.

The core conclusion of this report is that even as Nvidia's quarterly revenue approaches $100 billion, the growth rate may continue to accelerate. This assessment is supported not by a single product cycle, but by the demand sources expanding from traditional major cloud providers to AI labs, new AI clouds, enterprise customers, and sovereign AI projects. The network equipment and CPU businesses are also expanding the market that the company can cover.

$288 Relies on Earnings Growth, Not Valuation Expansion

Morgan Stanley's baseline scenario uses an approximately 22 times expected price-to-earnings ratio for the natural year 2027, corresponding to earnings per share of $13.08 and a target price of $288. The report specifically notes that this valuation is roughly in line with the market and lower than peers in computing chips such as AMD, Broadcom, and Intel.

In other words, Morgan Stanley does not assume that Nvidia can continue to achieve a higher valuation multiple. With the company's market share, gross margin, and market capitalization already at high levels, there is limited room for stock price growth driven by valuation expansion in the short term. Whether the target price can be realized primarily depends on whether revenue and profits can continue to grow significantly.

The baseline model expects Nvidia's revenue to grow by 82% in 2026 and by another 52.4% in 2027. GAAP revenue is projected to rise from $215.9 billion in 2026 to $393 billion in 2027, reaching $598.8 billion in 2028 and further increasing to $783.9 billion in 2029. During the same period, earnings per share are expected to rise from $4.61 to $17.63.

The bull market scenario target price is $330, corresponding to about 23 times the expected earnings per share of $14 for the natural year 2027. This scenario requires data center revenue to maintain high growth, further scaling of the Vera Rubin system, network, and software business, while AI PCs and automotive businesses open new markets.

The bear market scenario is set at $160, corresponding to about 16 times earnings per share of $10. Trigger conditions include a significant slowdown in data center growth, rapid declines in AI development costs, competitors gaining market share, cloud providers increasing self-developed chips, and greater impacts from tariffs and export restrictions.

The risk-reward range is a bull market at $330, baseline at $288, and bear market at $160, with the baseline scenario approximately 42% higher than $202.78.

Three Growth Lines Activated Simultaneously

During the roadshow, management categorized growth sources into three types, which is also the main basis for Morgan Stanley's optimistic judgment.

The first category is AI labs, currently accounting for about 20% of Nvidia's overall demand. The report mentions that a major customer relying on ASICs to develop cutting-edge models had a very low usage rate of Nvidia in the past, which has now risen to nearly 50%. Other major cutting-edge models still predominantly use Nvidia's computing platform.

This information is used by Morgan Stanley to counter concerns that "ASICs will quickly replace GPUs." Cloud providers will continue to develop custom chips, but lower-cost chips do not necessarily lead to lower token costs. Software maturity, network efficiency, cluster utilization, and deployment speed will all affect the final costs. The report estimates that both Nvidia and Broadcom's AI businesses could grow by over 80% in the next year, indicating that GPUs and ASICs can grow simultaneously during a phase of rapid demand expansion.

The second category is traditional major cloud providers, which account for about half of the company's revenue according to the new grouping. This demand remains strong, with limiting factors gradually shifting to land, power, and data center space. Nvidia's growth within cloud providers is also beginning to extend from GPUs to networks and CPUs.

The company reiterates that its CPU business target for 2026 is about $20 billion, with nearly half potentially coming from standalone CPU cabinets rather than supporting nodes in GPU systems. The Vera CPU is planned to enter standalone cabinets in the second half of the year, targeting AI scheduling, single-threaded computing, and memory optimization tasks.

The third category includes AI cloud, industrial, and enterprise customers, currently accounting for about half of data center revenue. Morgan Stanley believes that limitations in power and land, localized construction demands, and countries developing their own AI capabilities may drive the growth rate of new AI clouds and sovereign AI projects to exceed that of traditional cloud providers.

Retail, finance, and biotech enterprise customers are also building more specialized AI systems. The pace of project initiation may be slow, but once individual projects enter the construction phase, procurement scales may rapidly expand.

Rubin Progressing as Planned, Memory Shortages Have Dual Effects

In terms of product cycles, Morgan Stanley expects Vera Rubin to drive growth over the next 12 months.

In response to market rumors that Rubin Ultra may be delayed until 2028, Nvidia's management stated that the product will still be delivered next year. The company is indeed adjusting cabinet forms, with the original Kyber cabinets being replaced by new solutions to support larger single-cabinet expansion scales, but the 800V power supply and cross-cabinet optical interconnects are still progressing as planned.

This distinction is important. The report confirms that the product timeline remains unchanged, which does not mean that there are no adjustments in cabinet design and mass production processes. Whether Vera Rubin can smoothly ramp up is still influenced by power supply, optical interconnects, data center construction, and customer deployment progress.

Memory supply may remain tight for several years. Management suggested that computing, networking, and memory can compensate for each other through system design. The report cites that Nvidia once halved the LPDDR5 usage in a single cabinet to ensure more cabinets could be delivered, and also mentions that the SRAM technology used by Groq may provide another architectural direction.

This is not solely beneficial for memory manufacturers like Micron, Samsung, and SK Hynix. Years of shortages benefit prices and profits, but Nvidia may control usage by reducing memory in single cabinets, changing architectures, or optimizing networks. Morgan Stanley's judgment is that architectural adjustments will still depend on sustained years of supply tightness, which is temporarily insufficient to change the memory market trend.

Why Nvidia is Back at the Forefront of the Supply Chain

Morgan Stanley had previously shifted its semiconductor preference from Nvidia to SanDisk, and then to Micron, citing that storage and supply chain targets have higher profit elasticity during an upcycle. Now, Morgan Stanley has returned to Nvidia, focusing on relative valuation and earnings certainty.

Nvidia's stock price has risen since being re-listed as a favorite, but its performance has lagged behind some supply chain peers. The report believes that other parts of the supply chain may still have greater elasticity, but Nvidia has become a more prominent value target in the sector. Its expected price-to-earnings ratio for 2027 is about 22 times, lower than some computing chip peers, while the visibility of growth over the next two years remains high.

Management's roadshow also actively engages more value-oriented investors. The report states that the company plans to allocate at least 50% of its cash flow for cash returns from now on. For Nvidia, which is widely held by growth funds and where some funds are nearing their position limits, expanding the pool of value-oriented investors may become a new source of funds.

However, holding positions is not light. Page 5 of the report shows that the proportion of active institutional holdings is about 50.9%, with a long-short ratio of related hedge funds at 2.1 times, and the net exposure of the sector reaching 29.5%. Market consensus is also highly bullish, with 94% of analysts giving an "overweight" rating.

Revenue is expected to rise from $215.9 billion in 2026 to $783.9 billion in 2029, with gross margins maintained at around 72% in the long term.

This means that Nvidia's subsequent stock price increase still requires continuous upward revisions of earnings forecasts. Simply maintaining strong demand may not be sufficient to drive a stock with a market capitalization close to $5 trillion and already high institutional holdings to continue to outperform significantly.

Morgan Stanley's $288 target price ultimately bets on three things: AI labs and new customers continue to expand demand, Vera Rubin maintains product leadership, and CPU and network businesses increase Nvidia's revenue in individual AI systems. Any shortfall in these areas could compress profit growth. Only if all three are realized can Nvidia have the opportunity to approach $288 relying on profit growth without further increases in valuation multiples.