Is the next wave of AI catalysts coming? Morgan Stanley elaborates on the three major highlights of NVIDIA's GTC: European investment catching up, quantum commercialization, and accelerated industrial AI

Eddie Wu delivered a speech at the GTC conference yesterday, and Morgan Stanley strongly supports NVIDIA, stating that three major engines will provide additional growth momentum, maintaining an overweight rating and a target price of $170, which implies about 20% growth potential, and listing it as a top pick in the semiconductor sector.

According to news from the Wind Trading Desk, Morgan Stanley released its latest research report after attending NVIDIA's GTC developer conference held in Paris, indicating that in addition to the strong acceleration expected in the second half of the year, three new growth drivers have emerged: Europe's catch-up in AI data center investment, opportunities in quantum computing, and industrial/physical AI applications will provide incremental momentum for NVIDIA.

Morgan Stanley emphasized that their enthusiasm for the stock is primarily based on the significant acceleration of inference workloads. The company stated at the GTC conference in March that the top four cloud customers are deploying 1.3 million Hopper chips in 2024, with an expected deployment of 3.6 million Blackwell chips (1.8 million chips) in 2025.

Analysts believe that since March, these numbers have accelerated for all four customers. The 5GW of power capacity corresponds to a demand for about 2 million Blackwell chips, slightly higher than the quantity that NVIDIA indicated all four major U.S. cloud service providers ordered at the March GTC conference. This translates to about $70 billion in GPU investment and a total investment of approximately $100 billion.

In Morgan Stanley's view, everything is accelerating: rack supply is in place, transitioning to GB300, increasing inference demand, physical AI training, and sovereign AI demand are all growing simultaneously. Morgan Stanley believes that management is highly confident about the outlook for the next few quarters, and this growth is expected to manifest progressively on a quarterly basis rather than providing a 12-month forecast.

European AI Investment Accelerates Significantly, GPU Demand Soars

The European Union launched a €200 billion AI investment plan earlier this year, which includes €20 billion to fund the construction of five AI super factories (each equipped with over 100,000 GPUs). NVIDIA's announced plans will require "over 3000 exaflops of NVIDIA Blackwell computing power."

Morgan Stanley stated that specific deployments include:

• 18,000 GB200 first-phase deployments in collaboration with France's Mistral, expanding to multiple locations next year
• 14,000 Blackwell development projects in collaboration with Nebius and Nscale in the UK
• Collaborations with partners like Domyn and telecom companies such as Swisscom, Telefónica Spain, Telenor, Fastweb, and Orange
• Building the "world's first industrial AI cloud" for European manufacturers, equipped with 10,000 Blackwell GPUs

Additionally, NVIDIA provided financial forecast data at the conference: the installation of GPU chips is expected to triple by 2025 compared to 2024, and grow tenfold by 2026 compared to 2024 Morgan Stanley analysts assume that in 2024, the Hopper chip will be used throughout the year at a cost of $22,000 per chip, and by 2026, the Blackwell chip will be fully adopted at approximately $18,000 per chip. This means that revenue in the European market could grow eightfold or more from 2024 to 2026, with a growth rate exceeding 150% in 2026.

The Path to Quantum Computing Becomes Clearer

CEO Jensen Huang stated in a speech that quantum computing is reaching a critical turning point and will become powerful enough in the coming years to "solve some interesting global problems."

Morgan Stanley noted that Huang has shown a more positive attitude towards quantum computing, particularly regarding NVIDIA's opportunities to sell classical computing technology that simulates and ultimately complements quantum computing.

He reiterated previous comments about the industry nearing a turning point and clarified that while he believes the timeline beyond the Noisy Intermediate-Scale Quantum (NISQ) era will not accelerate, he is more optimistic about the timeline for commercial applications of GPU/QPU hybrid systems.

Physical AI and Industrial Applications Become New Focus

Additionally, the conference emphasized the next wave of activities in physical AI and industrial AI. Even though humanoid robots will take time to develop, related investments are actively underway. The focus is on creating digital twins for manufacturing and warehousing planning using Omniverse, driving new technologies, and training robots physically