SLB agrees to acquire Kelvion for $4.1B to boost data center cooling capabilities

SLB, the company formerly known as Schlumberger, is spending $4.1 billion to buy Kelvion, a thermal management and heat exchange specialist. The deal signals just how seriously the oil services giant is pivoting toward the infrastructure backbone powering the AI boom.

The acquisition, announced on August 31, values Kelvion at approximately $3.4 billion in cash, with SLB also assuming roughly $700 million in debt. That total enterprise value works out to about 11 times Kelvion’s estimated 2026 EBITDA.

The numbers behind the pivot

Kelvion is projected to pull in between $2.3 billion and $2.4 billion in revenue for 2026. The interesting part: more than half of that, roughly $1.2 billion to $1.3 billion, comes from data center customers alone.

SLB’s own Data Center Solutions business has been on a tear, posting a compound annual growth rate north of 90% between 2024 and 2026. Combine the two operations, and the merged entity expects pro forma data center revenues exceeding $2 billion and adjusted EBITDA of around $300 million by year-end 2026.

The longer-term targets are even more ambitious. SLB is projecting combined revenues of $4.5 billion to $5 billion and EBITDA of $700 million to $800 million by 2028. The company also expects to unlock roughly $120 million in annual EBITDA synergies within three years of closing the deal.

Kelvion’s estimated adjusted EBITDA for 2026 falls between $350 million and $400 million.

Why an oil company is buying a cooling business

SLB CEO Olivier Le Peuch has framed the acquisition as central to meeting the surging infrastructure demands created by artificial intelligence. Modern AI training clusters and inference operations generate enormous amounts of heat, and traditional air cooling is hitting its physical limits. Liquid cooling and advanced heat exchange systems, exactly what Kelvion specializes in, are becoming essential rather than optional.

Traditional air-cooled data centers struggle to handle power densities above 30-40 kilowatts per rack, while cutting-edge AI clusters can push well beyond that.

The transaction is expected to close in the first half of 2027, subject to regulatory approvals.