SKN | Morgan Stanley Sees AI Selloff as Buying Opportunity Amid $3.4 Trillion Hyperscaler Spending Outlook

Key Takeaways:

• Morgan Stanley maintains its hyperscaler capital expenditure forecasts of $1.6 trillion in 2027 and $1.8 trillion in 2028, despite potential delays affecting US data center construction.
• Around 20% of US data center projects planned for 2028 could be postponed until 2029, representing approximately 5 gigawatts of capacity, while a shortage of powered capacity is expected to persist through 2028.
• The bank sees large hyperscalers as relatively well positioned to navigate infrastructure constraints but warns that some equipment and data center suppliers could face slower orders, inventory accumulation and margin pressure in 2028.

Morgan Stanley Maintains Confidence in the AI Investment Cycle

Morgan Stanley views the recent decline in artificial intelligence-related equities as a potential buying opportunity, arguing that markets may have overreacted to the US midterm election results and their implications for data center construction.

The bank’s assessment, presented in its Buy the Drop: AI Investment Opportunities Post the US Midterms report, maintains that the broader AI investment cycle remains intact. Although political opposition and regulatory requirements could delay some infrastructure projects, Morgan Stanley has left its hyperscaler capital expenditure forecasts unchanged at $1.6 trillion in 2027 and $1.8 trillion in 2028, while acknowledging some downside risk to the latter estimate.

The distinction is between a delay in infrastructure development and a deterioration in underlying demand for AI computing capacity. Morgan Stanley does not expect projects already under construction to be disrupted. Instead, it anticipates that permitting and regulatory obstacles could shift some planned 2028 developments into 2029.

This suggests that the investment cycle may become more extended rather than coming to an end. For financial markets, the timing of capital expenditure remains important because delays can affect near-term supplier revenues, construction schedules and financing requirements without necessarily eliminating the long-term demand for computing infrastructure.

Data Center Delays Reshape Capacity Expectations

The most immediate construction concerns are concentrated in Texas and Ohio, where newly elected governors could introduce additional requirements for data center development. Morgan Stanley expects approximately 20% of US projects scheduled for 2028 to be delayed until 2029, representing around 5 gigawatts of capacity.

The bank forecasts that US data center capacity additions will decline by 4% in 2028 before increasing by 38% in 2029. Globally, capacity additions are expected to grow by 9% in 2028 and 20% in 2029.

The potential delays could also encourage hyperscalers to reconsider the geographical distribution of their infrastructure investments. Morgan Stanley identifies Tennessee and Louisiana as possible alternative locations because of their comparatively accommodating development environments. It also sees the potential for up to 20 gigawatts of additional data center development across Asia and Europe, the Middle East and Africa.

Electricity availability remains a central constraint. The bank expects a shortage of more than 30% in powered data center capacity to persist through 2028. This indicates that the challenge extends beyond construction approvals and equipment procurement. Access to sufficient electricity and operationally ready facilities will remain important determinants of how quickly new computing capacity can become available.

Morgan Stanley argues that a slower construction schedule could ultimately support the broader investment cycle by moderating the supply-demand imbalance. Delayed capacity additions may help sustain lease rates for existing computing infrastructure and extend the period during which additional investment remains necessary.

Hyperscalers and Infrastructure Suppliers Face Different Prospects

A central element of Morgan Stanley’s analysis is the distinction between the largest technology companies and the businesses supplying the data center construction cycle.

The bank is particularly constructive on major hyperscalers, which have the financial resources, existing infrastructure and technical capabilities to navigate electricity constraints and regulatory delays. Its preferred companies include Amazon, Meta, Alphabet and Microsoft, all rated Overweight.

CoreWeave and Nvidia are also rated Overweight. CoreWeave represents exposure to AI computing infrastructure, while Nvidia remains important to improving the amount of AI output generated per watt of electricity.

The largest technology companies may be better positioned to absorb construction delays because of their access to capital and established operating ecosystems. Their ability to finance infrastructure internally and adjust development schedules could provide advantages over smaller operators with greater dependence on external financing.

Morgan Stanley also highlights the potential for significant improvements in AI computing efficiency. Meta’s target of achieving a tenfold increase in compute efficiency in 2027 illustrates the scale of technological progress being pursued.

Advances in model architecture, distillation, training techniques and specialized chips could allow AI applications to expand without requiring computing resources to increase at the same rate. Greater efficiency could therefore help hyperscalers manage power constraints while improving the economics of AI infrastructure.

Equipment Suppliers Could Face a 2028 Downcycle

The outlook is less straightforward for companies supplying equipment and services to the data center construction market.

If the primary bottleneck shifts from equipment availability to access to powered buildings, suppliers could experience reduced pricing power, slower orders and rising inventories. Morgan Stanley warns that parts of the sector could enter a traditional downcycle in 2028, involving margin normalization and earnings downgrades even if underlying AI demand remains strong.

The bank identifies Equinix and powered-shell providers including WULF, RIOT, HUT and APLD among the infrastructure-related opportunities it is monitoring. These companies are viewed as less exposed to the specific Texas-related risks identified in the assessment.

By contrast, CIFR, GLXY and MARA are considered potentially more vulnerable to construction delays or moratoriums. This differentiation highlights the importance of geographical exposure and project execution in evaluating companies connected to the data center investment cycle.

For investors, the implications extend beyond whether a company participates in AI infrastructure development. The timing of construction, access to electricity, inventory management and the ability to maintain margins could increasingly determine financial performance.

Financing Conditions Favor Established Companies

The financing outlook also points toward a more selective environment. Morgan Stanley expects AI-related debt issuance to increase in 2027 as companies continue funding infrastructure requirements.

However, new data center developments and neocloud operators are expected to face greater scrutiny from lenders. Projects with uncertain construction schedules, regulatory exposure or unresolved power constraints may encounter more demanding financing conditions.

Established operators could benefit from stronger balance sheets, existing infrastructure and more developed operating models. As capital requirements increase, access to financing and the ability to demonstrate project viability may become increasingly important competitive advantages.

This could reinforce the divide between large technology companies capable of funding substantial infrastructure commitments and smaller businesses that depend more heavily on external capital to execute their development plans.

Closing Insights

Morgan Stanley’s assessment suggests that the AI infrastructure cycle is undergoing a change in timing and competitive structure rather than a fundamental reversal. Its combined hyperscaler capital expenditure forecasts of $3.4 trillion for 2027 and 2028 remain intact, although regulatory delays and persistent power shortages could shift construction activity into 2029.

The distinction between hyperscalers and their suppliers is particularly important. Large technology companies may be able to use their financial resources, infrastructure and technological capabilities to navigate capacity constraints, while some equipment suppliers and smaller operators could face weaker orders, inventory pressure and more difficult financing conditions.

For investors and institutions assessing the technology sector, the next stage of the AI buildout may depend less on aggregate spending commitments and more on which companies can convert those commitments into operational, powered computing capacity. Morgan Stanley’s positive view of the recent market weakness is therefore selective, reflecting differences in business models, infrastructure exposure and financial resilience rather than an indiscriminate endorsement of all AI-related equities.

For a confidential discussion regarding retail banking strategy, insurance distribution models, customer loyalty ecosystems, digital financial services, or cross-border financial innovation opportunities, contact our senior advisory team.