The Zhitong Finance App learned that after attending the GTC Berlin conference held by Nvidia (NVDA.US) this week, Morgan Stanley released a research report stating that although artificial intelligence (AI) infrastructure construction is facing multiple challenges such as chip supply, data center power, and financing, Nvidia is further consolidating its competitive advantage by expanding supply chain capacity, optimizing AI data center deployment efficiency, and promoting the next-generation Vera Rubin platform. The bank believes that Nvidia is expected to maintain strong growth in 2027, reaffirm its position as the preferred stock in the semiconductor industry, and maintain its “plus” rating and target price of $300. Based on the closing price of $221.97 on October 21, Morgan Stanley's target price means that Nvidia's stock price still has potential room to rise by about 35%.
Morgan Stanley analyst Joseph Moore pointed out that although the GTC Berlin conference did not reveal many new details about the next Rubin Ultra products, Nvidia's progress in supply chain expansion, AI data center deployment, and software ecosystem has further strengthened the bank's confidence in the company's 2027 growth prospects.
Morgan Stanley expects Nvidia's revenue to grow by around 88% and 70% in 2026 and 2027, respectively. The bank pointed out that the demand growth rate previously described by Nvidia management was about 100%, which is significantly higher than the 2027 revenue growth forecast of about 70%, reflecting that supply capacity is still an important factor limiting further revenue growth.
Therefore, how to close the gap between market demand and actual delivery capacity is the key to Nvidia's future growth. The bank believes that Nvidia is gradually easing supply chain bottlenecks by expanding production capacity for chip substrates, advanced packaging, and testing. Among them, TSMC's 3nm wafer production capacity is still the most important supply limit in the industry's eyes, but the related pressure is expected to gradually ease over time.
According to research reports, Nvidia is expanding cooperation with substrate suppliers such as IBIDEN (IBIDEN), Samsung Electric (SEMCO), and Unimicron (Unimicron), and it is expected that related production capacity will increase 2.5 times the current level in the next three years.
In the field of advanced packaging, AMKOR plans to increase CoWos-R wafer packaging production capacity by more than 300% by the end of 2028 to support the growth of CPUs and other processor products.
In terms of testing, Nvidia has promised to invest 1.1 billion US dollars to reserve production capacity for new facilities and equipment built by KYEC in the US. It is expected that related facilities will enter large-scale production in April 2027. Nvidia also expects its overall semiconductor testing capacity to double in 2027 compared to 2026.
Furthermore, in the field of co-packaged optics (CPO), Silicon Precision (SPIL), a subsidiary of Sun Moon Light, plans to expand related production capacity 15 times the current level by mid-2027, and is expected to further double this by the end of 2028.
Morgan Stanley notes that these supply chain investments will not only help improve Nvidia's product delivery capabilities, but may also enable the company to further expand its competitive advantage in the context of continued rapid growth in AI infrastructure demand.
In addition to supply chain expansion, the DSX POD reference design launched by Nvidia at the GTC Berlin conference has also become a key technological advance for Morgan Stanley. DSX POD is a modular AI data center design solution for the Vera Rubin system. Relevant equipment can be integrated and verified before delivery, and customers can put the equipment into operation faster after receiving it.
Morgan Stanley pointed out that traditional AI data centers often take months from equipment delivery to formal generation of computational revenue, but after adopting DSX POD, this process is expected to be shortened to a few days.
For new cloud computing service providers that rely on massive debt financing to build AI computing power infrastructure, shortening the equipment deployment cycle is of great significance. The bank believes that AI servers and GPUs usually cannot generate revenue before they are officially put into operation, yet they have taken up a large amount of capital. If Nvidia can significantly shorten this phase, customers will be able to obtain cash flow earlier, thereby improving working capital turnover and relieving financing pressure.
Morgan Stanley emphasized that the market usually pays more attention to GPU performance, computing power scale, and product launch, but the time required for AI infrastructure from delivery to revenue generation may also be an important factor affecting customers' return on investment.
Nvidia also introduced the DSX MaxLPS solution. According to information disclosed at the conference, the solution can reduce electricity consumption by about 30% without reducing the AI model's token output.
Morgan Stanley believes that in the context of global AI data centers facing power supply and construction land restrictions, increasing the computing power generated per unit of electricity will be an important way to bridge the gap between Nvidia and its competitors. Nvidia anticipates that more than 10 European customers will use the DSX platform to build Vera Rubin AI factories.
In addition to hardware and infrastructure, Nvidia is also accelerating the expansion of AI software and other application fields. In the field of intelligent AI, the company launched the OpenShell reference architecture. This is a Linux-based open source framework that can isolate different AI agents in a separate secure environment and limit their system access rights.
At the same time, Nvidia Sentry running on the BlueField-4 platform can monitor AI agent activity from outside the system, further enhancing security.
Morgan Stanley pointed out that these products will help Nvidia expand its business scope in the field of AI smart devices from GPUs to Vera CPUs and Bluefield network processors, and increase the variety of Nvidia products used in individual customer deployments.
In the field of communications, Nvidia collaborated with Nokia to complete commercial phone calls based on CUDA technology on the T-mobile network for the first time. Nvidia CEO Hwang In-hoon also proposed that in the future, about 100,000 telecom base stations and data center sites, with a total infrastructure of about 100 gigawatts, are expected to gradually transform into distributed AI computing networks.
In terms of autonomous driving, Nvidia announced that four more car brands will be compatible with the Hyperion autonomous driving platform, further promoting it as an industry-standardized architecture.
In the field of humanoid robots, Nvidia and Agile Robots have partnered to launch the Agile ONE SN reference robot, which is expected to begin delivering products with pre-assembled and equipped computing and software systems in the first quarter of 2027.
Morgan Stanley believes that these developments indicate that Nvidia is expanding its competitive advantage from traditional GPUs to various fields such as AI infrastructure, software, communications, autonomous driving, and robotics, creating more opportunities for long-term growth.
In terms of valuation, Morgan Stanley maintains Nvidia's basic scenario target price of $300, which is about 20 times the price-earnings ratio of $15.01 per share forecast for the 2027 natural year. The bank believes that although Nvidia already has a high market share and gross profit margin, there may be limited room for further expansion of the short-term valuation multiplier. Strong AI demand and the continued advancement of the Vera Rubin platform are expected to push profit forecasts to continue to improve.
In an optimistic scenario, Morgan Stanley gave Nvidia a $350 valuation, which mainly assumes that data center revenue will further exceed expectations as supply improves, while CPU, AI software, and other new businesses will accelerate growth.
Under a pessimistic scenario, the bank gave Nvidia a valuation of 160 dollars. The main risks include a marked slowdown in AI data center demand growth, competitors' launch of more competitive products, and large cloud computing customers speeding up adoption of self-developed chips.
Overall, Morgan Stanley believes that Nvidia's core advantage at the GTC Berlin conference is not only the improved performance of next-generation GPU products, but also that it is establishing a complete system covering chip supply, advanced packaging, data center deployment, and software ecosystem.
As the Vera Rubin product cycle gradually begins and supply chain capacity and AI data center deployment efficiency continue to improve, the bank believes that Nvidia is expected to further narrow the gap between demand and supply and continue its strong growth momentum in 2027.