Nvidia and Broadcom are relatively protected from a growing power shortage affecting US data centers, but other parts of the AI chip supply chain could face pressure if data-center projects are delayed, according to Morgan Stanley.
The investment bank said the growing demand for artificial intelligence is running into a major infrastructure problem: there may not be enough electricity available to power new data centers.
Morgan Stanley does not currently expect the power constraints to put Nvidia or Broadcom’s 2027 forecasts at risk. However, the bank warned that companies supplying memory, optical components, power-management chips and analog components could be more exposed if customers delay or cancel orders.
US Data Centers Face a Growing Power Gap as AI Demand Rises
The rapid growth of generative AI has led technology companies to invest billions of dollars in new data centers. These facilities require large amounts of electricity to run AI servers and cooling systems.
Morgan Stanley estimated last month that US data-center developers could face a 34% net power shortfall through 2028, equivalent to around 32 gigawatts (GW). The estimate takes into account additional measures such as behind-the-meter electricity generation and fuel cells.
Even after considering these alternatives, the bank expects a significant gap between the electricity available and what data-center developers may need. The power shortage is becoming an increasingly important issue for the AI industry because building more AI computing capacity depends on having enough electricity and grid connections available.
ALSO READ: Google’s Finland AI Expansion Includes $15 Billion Investment and Nuclear Power Deal
Nvidia and Broadcom Have More Visibility Into AI Chip Demand
Morgan Stanley believes Nvidia and Broadcom are in a stronger position than some other semiconductor companies because of their visibility into future chip deployments. The bank said both companies have greater insight into where their chips will be deployed and how data-center expansion plans are developing.
Their close coordination with data-center operators, semiconductor suppliers and companies involved in the power supply chain also gives them more visibility into potential delays.
As a result, Morgan Stanley does not currently believe the power bottlenecks will put Nvidia or Broadcom’s 2027 forecasts at risk.
Nvidia is a major supplier of GPUs used to train and run AI models, while Broadcom has expanded its position in AI infrastructure through networking products and custom AI chips.
This makes both companies important parts of the growing AI data-center market. Their ability to work with customers and plan chip deployments could help reduce the impact of power-related delays.
AI Power Delays Put Memory and Optical Chip Suppliers at Risk
While Nvidia and Broadcom may be relatively insulated, Morgan Stanley expects other parts of the semiconductor supply chain to face greater risks. If data-center projects cannot receive enough electricity, customers may delay the deployment of new computing systems.
That could lead them to push back chip deliveries or, in some cases, cancel orders. Memory suppliers could be particularly exposed because AI servers require large amounts of high-performance memory.
Optical components could also face pressure because they are used to move large amounts of data between servers and other parts of data-center networks. Morgan Stanley also identified power-management and analog chip suppliers as areas that could experience inventory disruptions if AI infrastructure projects are delayed.
The concern is not necessarily that demand for AI chips is disappearing. Instead, the issue is that companies may have the chips available but lack the electricity and infrastructure needed to deploy them.
AI Boom Faces New Challenges From Power and Infrastructure
The AI boom has created strong demand for computing infrastructure. Technology companies are building new data centers and expanding existing facilities to support increasingly large AI models and applications.
However, the speed of data-center construction is creating pressure on electricity networks. Morgan Stanley and Goldman Sachs have both highlighted constraints affecting the US data-center buildout.
Morgan Stanley has pointed to power, labor and political challenges, while Goldman Sachs has also examined the potential impact of resistance to new infrastructure projects. This means the semiconductor industry could increasingly depend on factors outside chip manufacturing.
Even if chipmakers can produce enough processors, memory and networking equipment, data centers still need land, construction capacity, electricity connections and other infrastructure before those components can be put to use.
ALSO READ: Anthropic Expands Australian AI Infrastructure With First Data Centre Deal
Power Shortages Could Slow the Next Phase of AI Growth
The latest Morgan Stanley assessment highlights how the AI infrastructure race is changing. For several years, the biggest concern was whether the semiconductor industry could produce enough advanced chips to meet demand. Companies have invested heavily to increase production and expand semiconductor capacity.
The challenge is now moving further down the infrastructure chain. AI data centers need enormous amounts of electricity, and connecting new facilities to the power grid can take time. In areas where electricity supply is already tight, new projects may have to wait for additional generation or grid upgrades.
That could create a situation where chip demand remains strong but actual deployments move more slowly. For Nvidia and Broadcom, Morgan Stanley’s view suggests their strong customer relationships and visibility into deployments could help protect their near-term outlook.
For other chip suppliers, however, delays could create a more difficult environment. Inventory could build up if customers postpone orders, while suppliers may have to adjust production plans to match slower deployments.
Electricity Access Becomes Critical to AI Data Center Expansion
The US data-center industry is expected to continue expanding as companies invest in AI infrastructure. However, electricity availability is becoming an important factor in determining how quickly those projects can move forward.
Morgan Stanley’s latest assessment suggests that Nvidia and Broadcom are better positioned to manage the problem, at least for now. The bank does not expect current power constraints to threaten their 2027 forecasts.
The wider semiconductor supply chain could face more uncertainty if power shortages cause AI data-center projects to be delayed.
As AI companies continue building larger systems, access to electricity could become just as important as access to advanced chips. The ability to secure power, expand data centers and deploy hardware on schedule may determine how quickly the next phase of the AI infrastructure boom can grow.
