Several AI-focused companies across thermal management, power and data center construction presented at Morgan Stanley’s Laguna conference this week. The takeaways and data points are compilations of the information from their presentations.
A few select themes and takeaways: consistent messaging that companies are not currently seeing delays to data center projects; thermal management for AI factories is moving towards integrated offerings and liquid cooling growth is above expectations; and power and electrical pricing and trends are intact. On this angle, we would point out that near-term behind-the-meter trends could compete with grid-scale orders if political and regulatory changes create further delay, which would ultimately be negative for GEV from a backlog build perspective, but for ETN it likely pulls or lengthens the overall cycle.
ECL: An positive read-through for liquid cooling and integrated thermal management benefits ECL with its emerging AI thermal management portfolio and upcoming November analyst day.
Management teams expressed confidence in the pace of data center projects, and while they are monitoring the situation, the primary tone was one of continued forward progress. HVAC companies had some comments around community focus and closed-loop systems, which validate an emerging theme around buy-in for data centers and reduced environmental impacts. There were no direct comments about new project approvals, the queue re-alignment in ERCOT or PJM, or similar moratorium and political debates. As a result, while positive for the current projects being built, the commentary did little to address “new” project approval pace.
Data center buildout pace
GE Vernova (GEV): Supportive of the continued strength in orders, they reiterated their backlog goals in 2027 and gave supportive evidence that the 12GW of extra capacity slots for the end of the decade would be fully spoken for by late this year. They also indicated pricing remains strong across the business and will give an update on 2026 backlog margins in early 2027, which will give a read-through to late-decade margins.
Eaton (ETN): They have seen the pipeline for announced projects increase by 11.4% since Q2 earnings, and do not see a large slowdown. If anything, the commentary indicates a longer cycle as current backlog is tied to 2027 and 2028 revenue and many announced projects have not come into backlog yet.
Cummins (CMI): Demand remains strong, and while they monitor the headlines, they confirmed the order book is now out to 2H 2028. Customers unable to secure the larger engines are stepping down to smaller sizes, signaling tightness in the market. Their prior target of $9B data-center-exposed revenue by 2030 was reaffirmed.
Gates Industrial (GTES): Seeing more opportunity today than a year ago within the pipeline, emphasizing their industrial pumps into data center cooling. Exiting 2026 at $25–30M of data center revenue, roughly 2.5x last year, against a $100–200M 2028 target. They plan to provide more data center exposure detail at their November analyst day.
Sterling Infrastructure (STRL): Management commented that they have at least five years of project demand and are not seeing any pauses or cancellations. Notably, they discuss that when a project starts for them, the permitting and power are complete. We would note this is precisely where the midterm and political pressure is focused, at the beginning of the process.
Liquid Cooling / Thermal Management
ETN, Trane (TT) and Honeywell (HON) gave data points that supported the liquid cooling demand trends, as well as our thesis about integrated solutions becoming a competitive advantage for larger AI factory buildouts.
ETN’s Boyd business, recently guided to generate $1.8B in revenue this year, is likely on pace to beat that number, supportive of the high growth rate for liquid cooling.
ETN’s content per MW has grown from $1.5M to $3.4M since buying Boyd, and while there remains a transition of the portfolio as they build towards 800V, the current content number holds for the near future.
TT discussed the positive trends around the CDU business that was part of LiquidStack, which they acquired earlier this year. In line with our view of single-phase liquid cooling being a dominant theme for AI factories, TT noted that immersion cooling is not ready for “primetime”.
Per HON, data center activity outside of the U.S. is seeing acceleration, notably Europe and early days in Korea and India.
Power Backlog and BTM / GEV
GEV spent time discussing their position on both grid power and behind-the-meter solutions. They did ultimately emphasize grid power as a desired route for most data centers, which is logical given the backlog mix. Their 30–35MW aeroderivative orders in Q2 equated to 1.8GW of the 12.1GW of orders, thus a smaller portion of the mix and capacity.
HON noted they are seeing an increasing trend of data centers building their own power due to interconnection challenges. They see an opportunity for their core automation business to accelerate if this trend continues and the company can offer an integrated power and data center solution to more buildouts.
AI Adoption Trends at Enterprises
3M (MMM): AI is helping the company more efficiently make the transition from a holding company to an operating company and optimize back-office activities. The higher value upside to AI will be across their supply chain and accelerating product innovation. The company is seeing the ability to compress the time to launch new products by at least 20% as a starting point. In addition, they note their “moat” is the company’s data and intellectual property where AI can be the powerful accelerant.
Honeywell (HON): The company is seeing customer demand for agentic AI within operating systems. The value-add is bringing domain expertise about system data, as well as having data that is in control systems (not in the public domain), which can be fed into LLMs to generate powerful returns for the customer.
Please reach out if you would like to discuss any of these topics further.
