The growing demand for artificial intelligence computing has transformed data centres into one of the most closely watched investment opportunities in the technology and infrastructure sectors. However, institutional investors are beginning to show greater caution towards companies linked to the construction and operation of these facilities. The shift does not indicate that demand for artificial intelligence has disappeared. Instead, it suggests that investors are becoming more selective about where they expect the next phase of growth to generate sustainable financial returns.
Regulatory filings covering the third quarter of 2026 show that more institutional funds completely exited positions in major data centre companies than initiated new investments. An analysis of nearly 6,400 filings found that 138 funds sold their positions in data centre stocks, compared with 119 that entered the sector. The number of institutional holders across eight major publicly traded data centre companies also declined from 3,411 to 3,392.
The figures represent a relatively modest change in the overall number of investors, but they are significant because they point to a shift in sentiment towards a sector that had benefited from substantial enthusiasm surrounding artificial intelligence. Earlier investment in computing infrastructure had driven interest in data centre operators, equipment manufacturers and electricity providers expected to benefit from growing demand.
The retreat was particularly visible in utility and power-related stocks. At least 212 funds exited these positions, more than twice the 96 funds that initiated new investments. This suggests that investors are reassessing not only the companies operating data centres but also the wider infrastructure businesses expected to benefit from the expansion.
The important question is why enthusiasm is weakening at a time when technology companies continue to invest heavily in artificial intelligence. The answer lies partly in the difference between the growth of a technology market and the ability of individual infrastructure companies to generate adequate returns from that growth.
The AI Infrastructure Boom Is Becoming More Expensive
The expansion of artificial intelligence has created an unprecedented requirement for computing capacity. Large technology companies are investing in specialised processors, networking systems and data centres to support increasingly demanding applications. Training advanced artificial intelligence models and operating them at scale requires substantial computing resources, making infrastructure a central part of the industry's development.
This has created opportunities for companies supplying data centre equipment, cooling systems, electricity infrastructure and specialised computing services. Businesses such as Vertiv Holdings and Digital Realty Trust have attracted investor attention because their operations are connected to the growing demand for data centre capacity.
However, the cost of developing these facilities is also increasing. Modern data centres require significant investment in land, construction, electrical systems, cooling equipment and network connections. Their operating costs can remain substantial even after construction is completed, particularly when electricity prices rise or additional cooling capacity is required.
The commercial challenge is that infrastructure expenditure usually takes place before the full revenue potential of a facility becomes clear. Investors may be willing to finance construction when demand appears strong, but they still need evidence that customers will use the available computing capacity and pay enough to generate acceptable returns.
A facility that remains underutilised can become an expensive financial commitment. Equipment also requires maintenance and eventual replacement, while technological developments may change the type of computing systems customers need.
The rapid expansion of infrastructure therefore creates opportunities, but it also increases the possibility of overinvestment if construction moves faster than sustainable demand.
Investors Are Distinguishing Between AI Companies and Infrastructure Providers
One of the more important features of the third-quarter filings is that institutional investors did not abandon artificial intelligence and semiconductor stocks altogether. Their holdings in these categories remained broadly positive, although investment was concentrated among a smaller group of companies.
This distinction suggests that investors are becoming more selective rather than rejecting the wider artificial intelligence industry. Companies developing advanced processors and businesses with established positions in the artificial intelligence market may continue to attract capital even as investors reassess the infrastructure required to support them.
Semiconductor businesses can benefit from demand across several markets, including cloud computing, artificial intelligence, consumer electronics and industrial applications. Data centre operators, by comparison, are exposed to a different set of commercial risks involving physical assets, electricity costs, financing and customer utilisation.
The difference matters because the financial performance of an infrastructure company does not necessarily move in line with the success of the technology it supports. Demand for artificial intelligence may continue increasing while certain data centre investments generate disappointing returns.
Investors may also be concerned about the concentration of capital spending among a relatively small number of technology companies. If a limited group of major customers accounts for a large proportion of future computing demand, any reduction in their investment plans could affect several infrastructure providers simultaneously.
This creates a potential weakness in the sector's business model. Companies may have committed significant capital to facilities based on expectations of continuing expansion by major technology customers. If those customers slow their spending, the consequences could extend beyond individual data centre operators.
Electricity and Local Opposition Are Becoming Investment Risks
The sharp retreat from utility and power stocks deserves particular attention because electricity has become one of the most important constraints on data centre expansion. Artificial intelligence facilities require substantial and often continuous power supplies. Their growth can place additional pressure on local electricity networks, requiring investment in transmission systems, generation capacity and supporting infrastructure.
The availability of electricity is not always guaranteed simply because a company has secured land and financing for a new facility. Connecting large installations to the grid can involve lengthy planning procedures, infrastructure upgrades and negotiations with electricity providers. Delays can increase construction costs and postpone the point at which a facility begins generating revenue.
Water consumption is another issue, particularly for facilities using water-based cooling systems. The increasing size of data centres has raised concerns about pressure on local water resources, especially in areas where supplies are already constrained.
These concerns have become more visible in communities where residents fear that large computing facilities could increase electricity prices, place pressure on water supplies and create environmental costs without providing sufficient local economic benefits.
Political opposition is therefore becoming a commercial consideration rather than simply a public relations issue. Changes in planning regulations, restrictions on electricity connections and local resistance can affect the timing and financial viability of projects.
The third-quarter filings also need to be understood in the context of political uncertainty in the United States. Concerns about possible restrictions on data centre construction and the changing political environment have contributed to a reassessment of the sector. However, the filings alone cannot establish why individual funds sold their positions.
The Investment Cycle May Be Entering a More Selective Phase
Another important consideration is the difference between the timing of investment decisions and the timing of market developments. The third quarter ended on September 30, before the November market reaction following the US midterm elections. The filings therefore provide a snapshot of institutional positions at the end of September rather than a complete picture of subsequent trading activity.
The data also comes with limitations. Quarterly institutional filings do not disclose the reasons behind individual investment decisions and do not capture every transaction in real time. A fund exiting a particular stock may be responding to portfolio adjustments, valuation concerns or company-specific developments rather than a broader loss of confidence in artificial intelligence.
Nevertheless, the overall pattern suggests that investors are becoming more cautious about the companies expected to benefit indirectly from the artificial intelligence boom. The strongest opportunities may increasingly be concentrated among businesses with established customers, reliable power access, efficient operations and the financial capacity to manage large infrastructure commitments.
For technology companies, this could mean greater scrutiny of capital expenditure and a stronger focus on demonstrating how new computing facilities will generate revenue. For infrastructure providers, it may become more difficult to secure financing without clear evidence of customer demand and long-term commercial agreements.
The retreat also challenges the assumption that every company connected to artificial intelligence will automatically benefit from its expansion. The sector's growth creates demand for physical infrastructure, but that demand must translate into profitable operations for investors to justify the valuations attached to these businesses.
The next phase of artificial intelligence investment may therefore be less about building as much computing capacity as possible and more about determining which projects can deliver sustainable returns. The technology industry still requires infrastructure, but investors appear increasingly unwilling to treat every data centre, power supplier or related company as an automatic winner.
(Source:www.reuters.com)
Regulatory filings covering the third quarter of 2026 show that more institutional funds completely exited positions in major data centre companies than initiated new investments. An analysis of nearly 6,400 filings found that 138 funds sold their positions in data centre stocks, compared with 119 that entered the sector. The number of institutional holders across eight major publicly traded data centre companies also declined from 3,411 to 3,392.
The figures represent a relatively modest change in the overall number of investors, but they are significant because they point to a shift in sentiment towards a sector that had benefited from substantial enthusiasm surrounding artificial intelligence. Earlier investment in computing infrastructure had driven interest in data centre operators, equipment manufacturers and electricity providers expected to benefit from growing demand.
The retreat was particularly visible in utility and power-related stocks. At least 212 funds exited these positions, more than twice the 96 funds that initiated new investments. This suggests that investors are reassessing not only the companies operating data centres but also the wider infrastructure businesses expected to benefit from the expansion.
The important question is why enthusiasm is weakening at a time when technology companies continue to invest heavily in artificial intelligence. The answer lies partly in the difference between the growth of a technology market and the ability of individual infrastructure companies to generate adequate returns from that growth.
The AI Infrastructure Boom Is Becoming More Expensive
The expansion of artificial intelligence has created an unprecedented requirement for computing capacity. Large technology companies are investing in specialised processors, networking systems and data centres to support increasingly demanding applications. Training advanced artificial intelligence models and operating them at scale requires substantial computing resources, making infrastructure a central part of the industry's development.
This has created opportunities for companies supplying data centre equipment, cooling systems, electricity infrastructure and specialised computing services. Businesses such as Vertiv Holdings and Digital Realty Trust have attracted investor attention because their operations are connected to the growing demand for data centre capacity.
However, the cost of developing these facilities is also increasing. Modern data centres require significant investment in land, construction, electrical systems, cooling equipment and network connections. Their operating costs can remain substantial even after construction is completed, particularly when electricity prices rise or additional cooling capacity is required.
The commercial challenge is that infrastructure expenditure usually takes place before the full revenue potential of a facility becomes clear. Investors may be willing to finance construction when demand appears strong, but they still need evidence that customers will use the available computing capacity and pay enough to generate acceptable returns.
A facility that remains underutilised can become an expensive financial commitment. Equipment also requires maintenance and eventual replacement, while technological developments may change the type of computing systems customers need.
The rapid expansion of infrastructure therefore creates opportunities, but it also increases the possibility of overinvestment if construction moves faster than sustainable demand.
Investors Are Distinguishing Between AI Companies and Infrastructure Providers
One of the more important features of the third-quarter filings is that institutional investors did not abandon artificial intelligence and semiconductor stocks altogether. Their holdings in these categories remained broadly positive, although investment was concentrated among a smaller group of companies.
This distinction suggests that investors are becoming more selective rather than rejecting the wider artificial intelligence industry. Companies developing advanced processors and businesses with established positions in the artificial intelligence market may continue to attract capital even as investors reassess the infrastructure required to support them.
Semiconductor businesses can benefit from demand across several markets, including cloud computing, artificial intelligence, consumer electronics and industrial applications. Data centre operators, by comparison, are exposed to a different set of commercial risks involving physical assets, electricity costs, financing and customer utilisation.
The difference matters because the financial performance of an infrastructure company does not necessarily move in line with the success of the technology it supports. Demand for artificial intelligence may continue increasing while certain data centre investments generate disappointing returns.
Investors may also be concerned about the concentration of capital spending among a relatively small number of technology companies. If a limited group of major customers accounts for a large proportion of future computing demand, any reduction in their investment plans could affect several infrastructure providers simultaneously.
This creates a potential weakness in the sector's business model. Companies may have committed significant capital to facilities based on expectations of continuing expansion by major technology customers. If those customers slow their spending, the consequences could extend beyond individual data centre operators.
Electricity and Local Opposition Are Becoming Investment Risks
The sharp retreat from utility and power stocks deserves particular attention because electricity has become one of the most important constraints on data centre expansion. Artificial intelligence facilities require substantial and often continuous power supplies. Their growth can place additional pressure on local electricity networks, requiring investment in transmission systems, generation capacity and supporting infrastructure.
The availability of electricity is not always guaranteed simply because a company has secured land and financing for a new facility. Connecting large installations to the grid can involve lengthy planning procedures, infrastructure upgrades and negotiations with electricity providers. Delays can increase construction costs and postpone the point at which a facility begins generating revenue.
Water consumption is another issue, particularly for facilities using water-based cooling systems. The increasing size of data centres has raised concerns about pressure on local water resources, especially in areas where supplies are already constrained.
These concerns have become more visible in communities where residents fear that large computing facilities could increase electricity prices, place pressure on water supplies and create environmental costs without providing sufficient local economic benefits.
Political opposition is therefore becoming a commercial consideration rather than simply a public relations issue. Changes in planning regulations, restrictions on electricity connections and local resistance can affect the timing and financial viability of projects.
The third-quarter filings also need to be understood in the context of political uncertainty in the United States. Concerns about possible restrictions on data centre construction and the changing political environment have contributed to a reassessment of the sector. However, the filings alone cannot establish why individual funds sold their positions.
The Investment Cycle May Be Entering a More Selective Phase
Another important consideration is the difference between the timing of investment decisions and the timing of market developments. The third quarter ended on September 30, before the November market reaction following the US midterm elections. The filings therefore provide a snapshot of institutional positions at the end of September rather than a complete picture of subsequent trading activity.
The data also comes with limitations. Quarterly institutional filings do not disclose the reasons behind individual investment decisions and do not capture every transaction in real time. A fund exiting a particular stock may be responding to portfolio adjustments, valuation concerns or company-specific developments rather than a broader loss of confidence in artificial intelligence.
Nevertheless, the overall pattern suggests that investors are becoming more cautious about the companies expected to benefit indirectly from the artificial intelligence boom. The strongest opportunities may increasingly be concentrated among businesses with established customers, reliable power access, efficient operations and the financial capacity to manage large infrastructure commitments.
For technology companies, this could mean greater scrutiny of capital expenditure and a stronger focus on demonstrating how new computing facilities will generate revenue. For infrastructure providers, it may become more difficult to secure financing without clear evidence of customer demand and long-term commercial agreements.
The retreat also challenges the assumption that every company connected to artificial intelligence will automatically benefit from its expansion. The sector's growth creates demand for physical infrastructure, but that demand must translate into profitable operations for investors to justify the valuations attached to these businesses.
The next phase of artificial intelligence investment may therefore be less about building as much computing capacity as possible and more about determining which projects can deliver sustainable returns. The technology industry still requires infrastructure, but investors appear increasingly unwilling to treat every data centre, power supplier or related company as an automatic winner.
(Source:www.reuters.com)
