Commentary: Sustainability Leaders Strategy

• U.S. stocks rose in the third quarter, powered by continued growth in the U.S. economy, low unemployment and expansion in the industrial sector, all of which were driven by continued investment in AI, data centers and related infrastructure.
• Portfolio positioning remains centered on high-quality compounders while selectively adding AI beneficiaries such as SiTime and Micron, power infrastructure exposure and software businesses with durable growth or lower disruption risk.
• The growth of AI and data centers increasingly depends on managing power, water and community impacts, creating opportunities for companies providing more sustainable infrastructure.

U.S. stocks rose in the third quarter, powered by continued growth in the U.S. economy, low unemployment and expansion in the industrial sector, much of which were driven by continued investment in AI, data centers and related infrastructure. Returns were also shaped by the petroleum flow disruption in the Middle East driving diesel prices higher and, partly a result of this, higher inflation expectations and bond yields. The Russell 3000 Index returned 1.4%. Sector performance in the index was mixed, with energy leading on higher oil prices, information technology (IT) and communication services lifted by hyperscaler strength as well as a bounce in software stocks, health care benefiting from defensive posturing and consumer and rate-sensitive sectors flat to down as a result of inflation and rate pressure.

The Strategy lagged the Russell 3000 Index in the quarter. This was in large part due to a rally in energy stocks and a strong quarter from Meta Platforms — stocks we do not own due to our sustainability focus. Additionally, our underweight exposure to software was a headwind as these companies bounced sharply after a steep selloff in the first half of the year when AI upended the sector. For the year, software has been a net neutral to performance.

Not owning Meta, which the Strategy has not historically owned due to social media controversies and governance concerns, was among the larger relative detractors, as the stock surged 29% driven by the early success of the launch of its personal AI agent, Muse.

While the Strategy is behind the benchmark year to date, we note that not owning traditional energy (oil and gas) has accounted for nearly half of the underperformance; this positioning is inherent to our sustainability mandate. Excluding the impact of energy, the Strategy is lagging the Russell 3000 Index by less than 100 basis points year to date, mostly due to the lower-beta nature of our IT holdings, which derives from our continued focus on quality.

In the third quarter, Nvidia, one of our largest overweights, was a standout performer as the market could no longer overlook its discounted valuation, especially as the company points to stronger than expected growth into 2027. Nvidia maintains leading share in AI chips and is expanding its wallet share opportunity in AI infrastructure. We also benefited from a strong quarter from Palo Alto Networks, which rallied as AI-driven security concerns boosted demand for enterprise cybersecurity. In addition, our IT underweight to the highest-beta AI infrastructure stocks such as Applied Materials and KLA, which we don’t own, helped as these sold off.

These positives in tech were outweighed, however, by profit taking in Flex, ASML and ASM International during the broader pullback in AI-exposed technology companies amid elevated valuations and questions about the trajectory of AI infrastructure spending. Our underweight to Microsoft also detracted as strong Azure growth and increasing Copilot adoption reinforced confidence in the company’s ability to monetize AI investment.

Both semis and software continue to be highly volatile as the market debates the winners and losers from AI disruption, the levels of capex spending, data center buildout constraints and elasticity of demand for AI compute.

Financials was the strongest-contributing sector, meanwhile, helped by insurers Travelers and MetLife. Travelers outperformed on a large second-quarter beat driven by favorable reserve development, light catastrophe losses and better underlying margins. It continues to distinguish itself through leading risk pricing management while leaning into AI-led business transformation. Life insurers like MetLife broadly outperformed as beneficiaries of rising interest rates, low valuations, and, in MetLife’s case, very strong returns on equity. MetLife also benefited from improvement in sentiment toward private credit, where it has some exposure.

Among held positions, global life sciences and diagnostics company Danaher was the biggest single contributor. It rose with a broad rerating of life sciences tools stocks amid mounting evidence the industry is emerging from a post-COVID slump. Additionally, the market began to appreciate the potential for life sciences tools providers like Danaher to benefit should AI-accelerated drug discovery drive more wet-lab work.

Not owning Tesla was a notable contributor as the stock underperformed the market amid a slower-than-anticipated robotaxi rollout.

We remain consistent in focusing on owning a concentrated portfolio of high-quality names with strong fundamentals, sustainability drivers that reinforce the fundamental case, low debt and solid management teams that we believe can outperform over the business cycle. We are monitoring the portfolio for AI and momentum exposure, high/low volatility exposure, and P/E and EPS growth versus the index. We continue to focus the core of the portfolio on long-term compounders, in particular ones with physical assets facing limited AI risk like Republic Services, Linde, Ecolab, Johnson & Johnson, Danaher and TJX.

We also have a healthy allocation to AI-exposed names, including AI picks and shovels in semis and industrials where growth is extremely strong. This includes data center/reshoring exposure as well as contract manufacturing, semi cap equipment, memory, advanced logic and analog semiconductor stocks. We prefer AI exposure with volume growth and those taking share within the overall AI spend, not just pricing. Along these lines, in the quarter we bought SiTime, which supplies precision-timing semiconductors that synchronize data moving between GPUs, servers and networking equipment. Faster data center networks and increasingly complex computing architectures, such as those used in agentic use cases, require more precise and resilient timing. Inference systems may also require more timing content per deployment. SiTime has a dominant share of silicon-based timing but only a single-digit share of the much larger overall timing market, providing a long runway for penetration.

We also added a below-benchmark position in Micron Technology, a leading producer of memory and storage semiconductors. As AI-driven demand broadens beyond high-bandwidth memory to conventional DRAM and NAND, we believe this can extend the memory cycle; our smaller position reflects the industry’s cyclicality and the risk that new capacity eventually pressures pricing and margins.

We own renewables and power infrastructure, which benefit as demand for power rises. While returns on AI spend for hyperscalers are being debated, power is the building block of AI and to the extent that the market will demand more compute, there will be a need for more power and higher pricing.

We are picking our spots within software; we continue to prefer areas that have either AI-driven revenue acceleration like cybersecurity or chip design tools, or areas relatively less vulnerable to disruption risk, such as new position Tyler Technologies. This company provides integrated software solutions for the public sector, where products are stickier and the customer base is more risk averse. Tyler’s valuation after the software selloff earlier in 2026 provided a good margin of safety.

Outside AI, we initiated a position in Boston Scientific following significant share price weakness caused by a slowdown in two key cardiac devices, one that treats irregular heartbeats and one that lowers stroke risk. Slowing growth in these two products has dragged down the company’s overall growth rate, but we believe the market has mostly priced in near-term headwinds, while its broader medical device portfolio continues to grow attractively. We exited Flowserve as our profitability improvement thesis largely played out and the valuation rerated. We also sold Deere as we view an uncertain agricultural equipment recovery to be already in large part reflected in the shares. In addition, we exited Cooper Companies due to persistent operational challenges and American Water Works because its pending merger with Essential Utilities introduces integration, balance-sheet and regulatory risks while the shares are exposed to rising interest rates.

Continued expansion remains our base case for the U.S. economy, but elevated interest rates, stubborn inflation as well as lack of final resolution to the Strait of Hormuz crisis create uncertainty. At the same time, AI spending and its downstream effects continue to supercharge parts of the economy and represent a meaningful portion of GDP growth. While AI adoption seems to be picking up steam and compute demand is outstripping supply, we remain vigilant on returns from this spending as this will be a key determinant of the duration and magnitude of hyperscaler capex increases. A meaningful slowdown in absolute levels of hyperscaler spending or even a major deceleration in spending growth can have a major impact on the state of the economy and on specific companies’ outlooks.

The ClearBridge Sustainability Leaders Strategy underperformed its Russell 3000 benchmark during the third quarter. On an absolute basis, the Strategy had positive contributions from three of the 10 sectors in which it was invested (out of 11). The leading contributors were the IT and health care sectors, while the industrials and utilities sectors detracted the most.

On a relative basis, stock selection in financials contributed to performance. Conversely, stock selection in communication services, IT, materials and industrials detracted from performance. A lack of exposure to energy also weighed on results.

On an individual stock basis, the largest contributors to relative returns were Danaher, Palo Alto Networks, Nvidia as well as not owning Tesla and Applied Materials. The largest detractors were Flex, Regal Rexnord, UL Solutions, Centerpoint Energy and not owning Meta Platforms.

Held annually in New York alongside the United Nations General Assembly, Climate Week brings together investors, companies, policymakers, researchers and civil society organizations to discuss climate risks and solutions. ClearBridge attends these events to deepen our understanding of emerging sustainability issues, compare perspectives with experts across industries and identify developments that could be material to our portfolio companies.

This work complements the company-level research and engagement performed throughout the year by ClearBridge analysts and portfolio managers. We seek to understand how environmental and social factors affect costs, competitive positioning, access to resources, regulatory exposure and long-term value creation.

Our overarching observation from this year’s Climate Week was that investors appear reinvigorated about climate. Conversations focused on the mechanisms through which climate and sustainability issues are already affecting businesses: energy affordability, resource efficiency, infrastructure resilience, insurance availability, community acceptance and economic competitiveness.

Not surprisingly, AI was central to almost all of these discussions. Participants debated not only the environmental footprint associated with data-center growth, but also AI’s potential “handprint”: how the technology might enable scientific advances, optimize power systems, improve industrial processes and accelerate the development of lower-carbon technologies. Energy efficiency was a key pillar of AI’s potential positive effects on climate, although we acknowledge projections far out into the future remain highly theoretical. All in all, while stakeholders debate the pros and cons of AI center development — and headlines over the summer have overwhelmingly highlighted the cons — AI infrastructure is on track to be built: the question for investors is how it can be built responsibly.

Food Security and Responsible Mining Remain Material

Although AI and data centers dominated the week, sustainable food and human rights in mining were also important areas of focus for ClearBridge.

For sustainable food, the central question is what happens when geopolitical disruption and climate variability affect agricultural inputs at the same time. Recent disruptions to fertilizer production and trade illustrate the issue. Nitrogen-based fertilizer must be available during relatively inflexible planting windows; a delayed shipment cannot necessarily be recovered later in the growing season. If fertilizer shortages or higher prices coincide with an El Niño-related disruption to rainfall and temperature patterns, the effects could extend from farm input costs to crop yields, commodity prices and food inflation.

This is an example of climate acting as a threat multiplier. It also highlights the importance of supply-chain diversification, soil health, efficient fertilizer use and agricultural technologies that help farmers adapt to more volatile conditions. For investors, historical weather and production patterns may be less and less useful for assessing agricultural risk. At Climate Week we discussed how JPMorgan’s food-security analysis emphasizes how fertilizer disruptions, planting windows and climate variability can combine to create longer-lasting food-system pressures.¹ AI is a potent source of innovation in this regard: ClearBridge holding US Foods, for example, uses AI-driven inventory forecasting to cut food waste.

For mining, the key question is how investors can connect community-level impacts to the companies in their portfolios and turn that information into actionable engagement. This is particularly important in industrial parks and complex mineral supply chains, where responsibility may be distributed among mine owners, operators, tenants, suppliers, customers, lenders and public-sector partners. Investors need information that sometimes embodies a complex balance: comparable across portfolios but sufficiently detailed to identify specific harms, responsible parties and potential remedies. Discussions focused on the Indonesian nickel supply chain in light of the EU Battery Regulation, which requires greater supply chain traceability, carbon footprint disclosure and environmental and human rights due diligence for battery materials sold in the EU and is relevant for EV and battery companies that source nickel such as China’s CATL and Tesla.

Sustainability Is a Key Gating Factor for Data Centers and AI

The scale of anticipated AI infrastructure development is intensifying pressure on power systems and local resources. BloombergNEF estimates U.S. data-center power demand could reach 118 gigawatts by 2030 and that data centers could account for roughly one-fifth of U.S. electricity consumption by 2035 (Exhibit 1).

Exhibit 1: Data Centers’ Growing Electricity Consumption

Meeting that demand will require substantial new generation and grid infrastructure. Yet long interconnection timelines, transmission constraints and equipment shortages are already affecting where and how quickly projects can proceed. Developers are exploring nuclear power, on-site generation and other behind-the-meter solutions, but each option introduces its own cost, emissions and community considerations.

Power is not the only constraint. Communities are asking whether data centers will increase electricity bills, consume scarce water, create persistent noise or require infrastructure whose costs are ultimately borne by residents. They also want to understand the local benefits, including employment, tax revenue and investments in public infrastructure.

These concerns can delay or prevent development. Sustainability is therefore becoming a practical requirement for translating demand for AI into operating data-center capacity.

Microsoft Raises the Standard for Development

Microsoft, a holding in several ClearBridge portfolios, offers a good example of how hyperscalers are responding. Its community-first infrastructure framework addresses several of the issues that can determine whether a project earns local support. The company has committed to paying the electricity and grid-infrastructure costs associated with its facilities rather than passing them through to residential ratepayers. It is working with utilities to forecast demand, contract for new generation and fund necessary transmission and substation improvements. The company is also targeting a 40% improvement in data-center water-use intensity by 2030 and has begun deploying closed-loop cooling designs that do not require potable water for cooling.

"Data center sustainability and economics are becoming closely aligned." 

The framework extends beyond environmental performance to include social dimensions. Microsoft has committed to paying full local property taxes, supporting workforce training and investing in schools, libraries and community organizations. It also plans to provide greater local transparency about water use and replenishment.

These measures don’t eliminate every impact associated with data-center development, but they illustrate the higher standard increasingly required of developers: mitigate costs, provide measurable local benefits and engage communities early. Microsoft’s community-first framework reflects the growing recognition that technical capacity alone is insufficient to secure a project’s social license to operate.²

Equinix Focuses on Long-Term Community Trust

A recent engagement with Equinix, another ClearBridge portfolio holding, provides the perspective of a data-center owner and operator. In September 2026 we asked whether faster anticipated capacity growth threatened Equinix’s goal of covering 100% of its power consumption with renewable energy by 2030. The CEO told us the commitment had not changed, noting both Equinix’s own priorities and demand from customers seeking to meet their sustainability commitments.

We also discussed community opposition to new data centers in general. Equinix views itself as a long-term neighbor rather than a temporary developer because it expects to own and operate its facilities for decades. This creates an incentive to establish durable relationships with local stakeholders.

Equinix also distinguishes its average approximately 65-megawatt retail colocation facilities from the hundreds-of-megawatts — and sometimes gigawatt-scale — campuses attracting the greatest scrutiny. While that difference does not remove the need for responsible development, it shows why investors should consider a facility’s scale, design and operating model when evaluating community risk.

Johnson Controls Improves Cooling Efficiency

ClearBridge’s recent engagement with Johnson Controls (JCI) highlighted how equipment suppliers can help address the physical constraints on data centers. JCI’s thermal-management offering includes cold plates, coolant distribution units, chillers and air-cooled technologies. Because it owns several major cooling subsystems, the company can optimize the entire cooling architecture rather than focusing only on the efficiency of an individual component. This system-level approach becomes more important as computing density rises and cooling represents a larger share of facility power requirements.

JCI’s technologies can also respond directly to local environmental concerns. Its cooling solutions can generally be configured for zero-water operation. Magnetic-bearing technology can reduce mechanical noise as well as energy consumption, while major product lines are available with low- and ultra-low-global-warming-potential refrigerants. In certain applications, absorption chillers can use waste heat in place of electricity as a primary energy source.

The engagement reinforced that data center sustainability and economics are becoming closely aligned. Reducing the power, water and noise associated with cooling can lower operating costs while making projects more acceptable to utilities, regulators and neighboring communities.

Vertiv Helps Make Power-Dense AI Feasible

As AI chips increase rack-level power density, data centers require tightly integrated power and thermal-management systems. Vertiv’s closed-loop cooling technologies are well suited to higher-density computing and can reduce dependence on continual water consumption. Its broad offering also allows customers to integrate cooling and power systems more easily — an increasingly important consideration as infrastructure becomes more complex.

Modular construction is another part of the solution from Vertiv. Its prefabricated OneCore platform brings multiple systems together before they reach the construction site, reducing on-site complexity and helping customers bring computing capacity online faster. Its global service network, real-time monitoring and predictive maintenance capabilities support reliability and efficient operation over the facility’s life.

Vertiv’s example illustrates that the sustainability of AI infrastructure is not limited to how electricity is generated. It also depends on how efficiently power is delivered, heat is removed, water is managed and complex facilities are constructed and maintained.

Conclusion

AI’s rapid expansion is creating investment opportunities across technology, utilities, construction and industrial infrastructure. As it is also exposing physical and social constraints, the need to deploy more efficiency could be a tailwind for AI Enablers (Exhibit 2).

Exhibit 2: AI Efficiency Enabler Group Returns by Year

ClearBridge’s research seeks to understand these constraints across the data-center value chain. With developers and operators, we ask whether energy and water commitments can keep pace with growth and how local communities will share in the benefits. With equipment suppliers, we examine total system efficiency, water requirements, refrigerants, noise, reliability and speed to deployment.

A common thread: sustainability is increasingly the mechanism connecting AI demand with practical development. Companies able to lower resource requirements, protect ratepayers, address community concerns and deliver reliable capacity may be better positioned to support — and benefit from — the next phase of AI infrastructure growth