Power, Cooling Companies Rush to Meet AI Demand, With Strong Results
By: Madhulika Pathak
Mid-size power, cooling and infrastructure firms posted standout results in the third quarter, boosted by hyper-scalers racing for both the ability to deliver electricity to data centers for Artificial Intelligence, and also to keep those complexes at a mild temperature.
While GPUs, or chips used to train AI models, are essential to the massive scaling of the technology, it's power to run data centers that has become the forefront of the conversation. There was a clear indication of this demand in third-quarter earnings reports of mid-cap infrastructure firms.
A standout this reporting season was Vertiv, a power, cooling, and IT infrastructure solutions provider, which reported a 60% year-over-year surge in organic orders in the third quarter, raising its backlog to $9.5 billion and its net sales to $2.676 billion.
Deloitte projects global data center power demand could reach 96 gigawatts, or the equivilant of enough power for approximately 80 million U.S. homes, by 2026. Goldman Sachs pegs a 165% increase in data center power demand by 2030 over a 2022 baseline, estimating that AI could account for more than a quarter of that load.
Mid-size power, cooling and infrastructure firms posted standout results in the third quarter, boosted by hyper-scalers racing for both the ability to deliver electricity to data centers for Artificial Intelligence, and also to keep those complexes at a mild temperature.
While GPUs, or chips used to train AI models, are essential to the massive scaling of the technology, it's power to run data centers that has become the forefront of the conversation. There was a clear indication of this demand in third-quarter earnings reports of mid-cap infrastructure firms.
A standout this reporting season was Vertiv, a power, cooling, and IT infrastructure solutions provider, which reported a 60% year-over-year surge in organic orders in the third quarter, raising its backlog to $9.5 billion and its net sales to $2.676 billion.
Deloitte projects global data center power demand could reach 96 gigawatts, or the equivilant of enough power for approximately 80 million U.S. homes, by 2026. Goldman Sachs pegs a 165% increase in data center power demand by 2030 over a 2022 baseline, estimating that AI could account for more than a quarter of that load.
The numbers are clear: Without sufficient power AI operators can’t run their clusters effectively.
Optical network vendor Ciena saw a third quarter rise in revenue to $1.22 billion, and their interconnect revenue is expected to double in the fiscal year of 2026. Connector maker Amphenol also picked up AI related tailwinds, posting $3.46 billion in quarterly sales, which is a 53% increase from last year. It flagged this as “exceptional organic growth” in its data-center interconnect business.
“At a strategic level, for cloud providers to monetize their substantial AI investments in LLMs and GPUs and related data center infrastructure, they need to invest in the network infrastructure that interconnects data centers or they risk stranding their massive investments," said Gary Smith, President and CEO of Ciena during the company's earnings call on September 4.
Another major concern is the cooling systems in these data centers. Racks, or dozens of stacked servers and networking gear, are higher than ever. To handle the large amounts of heat generated, data centers are increasingly turning towards advanced cooling technologies, including liquid immersion and high efficiency thermal architecture.
Analysts tracking the data‑center cooling space say attention is turning to next‑generation cooling technologies like immersion cooling.
Optical network vendor Ciena saw a third quarter rise in revenue to $1.22 billion, and their interconnect revenue is expected to double in the fiscal year of 2026. Connector maker Amphenol also picked up AI related tailwinds, posting $3.46 billion in quarterly sales, which is a 53% increase from last year. It flagged this as “exceptional organic growth” in its data-center interconnect business.
“At a strategic level, for cloud providers to monetize their substantial AI investments in LLMs and GPUs and related data center infrastructure, they need to invest in the network infrastructure that interconnects data centers or they risk stranding their massive investments," said Gary Smith, President and CEO of Ciena during the company's earnings call on September 4.
Another major concern is the cooling systems in these data centers. Racks, or dozens of stacked servers and networking gear, are higher than ever. To handle the large amounts of heat generated, data centers are increasingly turning towards advanced cooling technologies, including liquid immersion and high efficiency thermal architecture.
Analysts tracking the data‑center cooling space say attention is turning to next‑generation cooling technologies like immersion cooling.
Comments
Post a Comment