AI Data Center Power Shortage Threatens Chip Supply Chains, Nvidia and Broadcom Less Exposed
The rapid expansion of artificial intelligence infrastructure is creating new pressure on the US electricity network, raising concerns about potential disruptions across the semiconductor supply chain. While Nvidia and Broadcom appear relatively protected, other chip manufacturers could face growing uncertainty as data center power shortages threaten AI deployment schedules.
According to a Morgan Stanley assessment reported by Reuters on October 5, 2026, US data center developers could face a significant electricity supply deficit through 2028. The shortage could delay the installation and operation of AI computing systems, creating challenges for semiconductor manufacturers and technology infrastructure providers.
The findings highlight a growing issue in the global AI industry: manufacturing advanced processors is only one part of expanding computing capacity. Companies must also secure sufficient electricity, cooling infrastructure, physical space and supporting equipment to operate increasingly powerful data centers.
US Data Centers Could Face a 32 GW Power Shortfall
Morgan Stanley estimates that US data center developers could experience a net electricity supply shortfall of approximately 32 gigawatts through 2028, representing a 34% gap between projected requirements and available power.
The estimate already accounts for alternative energy arrangements, including on-site electricity generation and fuel-cell technologies.
This suggests that measures intended to reduce dependence on traditional electricity networks may not be sufficient to support the pace of planned AI infrastructure expansion.
Generative AI applications require substantial computing resources for training models, processing information and delivering real-time services. These systems depend on large clusters of advanced processors operating continuously inside power-intensive facilities.
As businesses invest heavily in AI capabilities, demand for data center electricity is increasing alongside the need for specialized hardware.
The challenge is becoming more complex because expanding electricity supply requires infrastructure development, grid connections, equipment installation and regulatory approvals. These processes may take longer than manufacturing and delivering semiconductor components.
Why Nvidia and Broadcom May Remain Protected
Despite growing infrastructure constraints, Morgan Stanley believes Nvidia and Broadcom are relatively insulated from immediate disruption.
The brokerage does not currently expect power-related bottlenecks to put the companies' 2027 forecasts at risk.
Several factors support this assessment.
Both companies have significant visibility into where their semiconductor products are expected to be deployed. This allows them to coordinate production, deliveries and capacity planning more effectively with their customers.
Geographic diversification also helps reduce exposure to individual power-constrained markets. AI infrastructure investments are expanding internationally, potentially allowing some deployments to proceed in regions with more readily available electricity.
In addition, coordination among semiconductor suppliers, data center operators and energy infrastructure providers could help prioritize installations where sufficient power is available.
However, being relatively protected does not mean these companies are completely immune to infrastructure delays. Extended shortages could still affect the broader pace of AI computing expansion.
Secondary Chip Manufacturers Face Greater Risks
The situation could be more challenging for companies supplying components that support the operation of AI servers and data centers.
Morgan Stanley identified several semiconductor categories that may be particularly exposed to deployment delays:
-
Memory chips: Components responsible for storing and rapidly accessing data during AI computing operations.
-
Optical components: Technologies used to transmit large volumes of data through high-speed connections between computing systems.
-
Power-management chips: Components that regulate and distribute electricity across electronic equipment.
-
Analog semiconductors: Chips used in monitoring, signal processing and supporting electronic systems.
These components are essential to modern computing infrastructure, but their manufacturers may face greater inventory uncertainty if customers postpone data center installations.
For example, a company may order semiconductor components based on an expected data center completion date. If electricity connections are delayed, the customer may not be ready to install or operate the equipment.
This can result in postponed deliveries, excess inventory or changes to purchasing schedules.
In more severe cases, customers could reduce or cancel orders.
The risk is therefore not necessarily a decline in long-term demand for AI technology, but a mismatch between the timing of component production and the availability of operational computing capacity.
Electricity Availability Is Becoming a Critical AI Infrastructure Challenge
The AI industry's expansion has traditionally been associated with semiconductor manufacturing capacity, advanced chip technology and access to high-performance computing hardware.
Electricity availability is now emerging as another important constraint.
New AI data centers require dependable power connections, substantial cooling capacity and supporting electrical infrastructure.
Even when the necessary computing equipment has been purchased, a facility cannot operate at its intended capacity without adequate energy supply.
Developers may need to consider alternative electricity sources, phased facility launches or different geographic locations to address these limitations.
These adjustments can introduce new costs and scheduling challenges.
The impact also extends beyond semiconductor manufacturers. Engineering contractors, electrical equipment suppliers, construction companies and infrastructure service providers must coordinate their activities around changing project schedules.
As a result, delays in one part of the infrastructure development process can create knock-on effects across multiple industries.
Labor and Regulatory Challenges Add Further Pressure
Electricity supply is not the only issue confronting US data center developers.
Morgan Stanley has also highlighted labor availability and political challenges as potential obstacles to infrastructure expansion.
Building large computing facilities requires specialized construction teams, electrical engineers and technical personnel.
At the same time, developers may face regulatory requirements, local concerns and lengthy approval processes.
Goldman Sachs has also identified increasing constraints on US data center development, although its assessment suggests that political opposition may have a more limited effect in the near term.
Together, these challenges indicate that scaling AI infrastructure will require more than additional capital investment.
Successful expansion will depend on the ability of technology companies, utilities, contractors and government authorities to coordinate infrastructure development.
Potential Impact on Global Technology Supply Chains
Power-related delays at US data centers could have implications for international semiconductor and electronics supply chains.
AI infrastructure depends on globally connected networks of manufacturers, component suppliers, equipment providers and logistics companies.
A change in deployment schedules in the United States could influence production planning, component deliveries and inventory management in other markets.
For semiconductor suppliers, maintaining flexibility may become increasingly important as customers adjust their delivery requirements.
Manufacturers may need to reassess procurement schedules, production volumes and inventory positions to avoid holding excess components.
Logistics providers supporting the technology sector could also experience changes in shipment timing and cargo demand if planned installations are postponed.
These are potential downstream consequences rather than disruptions that Morgan Stanley has confirmed across the entire logistics industry.
The broader issue highlights how infrastructure constraints in one market can affect planning across interconnected global industries.
AI Industry Growth Increasingly Depends on Infrastructure Readiness
The findings point to an important development in the AI sector.
Strong demand for artificial intelligence does not automatically translate into equally rapid deployment of computing equipment.
Companies must align semiconductor supply with electricity availability, facility construction and operational readiness.
Morgan Stanley's assessment suggests that established AI semiconductor leaders such as Nvidia and Broadcom may be better positioned to manage near-term constraints, while suppliers of supporting components face greater exposure to scheduling changes.
This distinction could become increasingly relevant as data center developers reassess project timelines and infrastructure requirements.
Outlook: Power Supply Could Shape the Next Phase of AI Expansion
The rapid growth of artificial intelligence is transforming investment priorities across the technology industry. However, the next phase of expansion may depend as much on energy infrastructure as on advances in computing hardware.
Morgan Stanley's projected 32 GW US data center power shortfall through 2028 illustrates the scale of the challenge.
Although Nvidia and Broadcom are expected to remain comparatively resilient, the wider semiconductor supply chain may face uncertainty if new computing facilities cannot become operational as planned.
For technology companies, component manufacturers and infrastructure providers, aligning production and delivery schedules with actual power availability will become increasingly important.
The central challenge is no longer just producing enough AI chips. It is ensuring that the infrastructure needed to operate them is ready when those chips arrive.