Carrier Global entered 2026 with a roughly USD 1 billion data center chiller backlog and a publicly stated USD 1.5 billion revenue target for the year, after Q4 2025 data center orders rose approximately 400% year-over-year [S2].
Global data center chiller revenue reached USD 5.23 billion in 2026, up from USD 4.80 billion in 2025, and is forecast to reach USD 11.26 billion by 2035 at an 8.90% CAGR, driven primarily by AI-driven heat loads and hyperscale buildouts [S4].
Where the 2026 Capacity Pipeline Actually Stands
Across roughly 140 projects, at least 16 GW of data center capacity is slated to come online in 2026, yet only about 5 GW shows visible construction activity, leaving roughly 11 GW in the announced stage with no confirmed build progress against typical 12–18 month construction timelines [S3].
Historical slippage supports the cautious read: 26% of expected 2025 capacity slipped, and another 10% of projects quietly pushed their commercial operation dates without public restatement [S3]. Applying that track record, 30–50% of the 2026 pipeline is unlikely to energize before year-end, which is a far tighter range than the "half is canceled" framing that circulated after an April 1, 2026 Bloomberg piece on Chinese electrical parts [S1][S3].
Independent tracking by SemiAnalysis shows the 2026 North American hyperscaler self-build forecast has moved by only about 1% over the prior six months, with North American colocation forecasts shifting by less than 5%, evidence that the "50% canceled" narrative is built on a flawed denominator and inflated by vibe-coded forecast models [S1].
Carrier's Backlog, Capacity Expansion, and Product Mix
Carrier expanded its Charlotte facility by 50% and converted a Monterrey, Mexico facility entirely to chiller production specifically to convert the order surge into shipped units [S2]. The product slate is split between massive 10–18 MW water-cooled chillers with frictionless magnetic-bearing (maglev) compressors, actively winning hyperscaler bids, and 2–5 MW air-cooled units scaling for smaller sites and edge builds [S2].
Coolant Distribution Units (CDUs), the mini-chillers feeding direct-to-chip liquid loops, are expected to contribute less than USD 100 million to 2026 revenue, but Carrier Ventures has already taken a stake in ZutaCore to secure two-phase, waterless direct-to-chip technology for the post-2026 AI rack roadmap [S2]. CEO David Gitlin, speaking at the JPMorgan Industrials Conference on March 18, 2026, framed the product portfolio as "night and day" relative to the post-spin baseline [S2].
Chiller Market Segmentation: What Specs Are Actually Selling
Air-cooled chillers led the 2025 segment with 45% share, while water-cooled chillers are projected to grow faster at a 9.5% CAGR through 2035, reflecting hyperscaler migration toward higher-density liquid loops [S4].
By cooling capacity, units above 2,000 kW held 45% of 2025 share, with the 500–2,000 kW band as the second-largest segment growing at a 7.5% CAGR, a useful two-tier split for spec-in decisions between hyperscale halls and edge or colocation rooms [S4].
Hyperscale data centers commanded 40% of 2025 chiller demand, colocation was the second-largest segment, and the cloud service provider application bucket is forecast to grow at 11% CAGR through 2035 as AI training capacity continues to scale [S4]. North America held 40% of 2025 chiller revenue, while Asia Pacific is the fastest-growing regional segment at 11.5% CAGR, an important geographic split for sourcing and lead-time planning [S4].
Why the Power Bottleneck Reshapes the Chiller Order Book
Hyperscalers are giving up on grid-only builds for AI training capacity: on-site and hybrid powering models account for less than 10% of project count but represent nearly half of announced gigawatt capacity, anchored by gigascale, grid-independent campuses such as New Era Energy & Digital's 7 GW New Mexico project [S3].
Google's acquisition of Intersect Power's 10.8 GW pipeline and Amazon's direct investments in solar and storage signal a shift from standard PPA structures toward project-level power control, a change that compresses chiller delivery windows because power and cooling must arrive on the same construction schedule [S3]. Nearly half of tracked projects have not disclosed a power strategy at all, which is a leading indicator that backlog-to-revenue conversion will be lumpy through 2026 [S3].
Selection Criteria: Matching Chiller Type to Site Profile
For greenfield hyperscale AI halls above 50 MW with high-density GPU racks, water-cooled chillers in the 10–18 MW class with maglev compressors are the dominant bid target, since they reject higher heat fluxes at favorable part-load efficiency [S2][S4].
For colocation retrofits and edge sites between 500 kW and 2,000 kW, air-cooled chillers remain the default selection, offering simpler siting, no condenser-water plumbing, and faster install, which matters when 30–50% of the 2026 pipeline is at risk of slipping into 2027 [S3][S4].
For liquid-cooled AI racks using direct-to-chip cold plates, CDUs effectively act as mini-chillers at the rack row, with two-phase waterless designs like ZutaCore positioned for the post-2026 generation; specifying a primary loop chiller sized to absorb the full CDU return enthalpy is the binding engineering constraint, not the rack-level CDU itself [S2].
Constraints, Failure Modes, and Sourcing Risks
The April 1, 2026 Bloomberg piece on Chinese electrical parts framed a fragile, China-dependent equipment supply chain as the main risk to 2026 deliveries, a narrative that the chiller OEM community has pushed back on but that nonetheless affects chiller electronics, variable-frequency drives, and maglev compressor controller boards [S1].
Permit and gas-pipeline risk remains material: STACK Infrastructure's Oracle site has been delayed to 2029 due to gas pipeline and regulatory burden, a cautionary example for any chiller spec tied to a gas-peaker-backed power package [S1].
AI-driven rack density is pushing chiller plant designs past historical norms, with 10–18 MW single-unit frames now being bid at hyperscale, and contractors should verify that building floor loading, structural penetrations, and condenser water chemistry programs are rated for these larger frames before locking a data logger-based commissioning plan to the chiller plant acceptance test [S2][S4].
Trackable Signals for the Next Quarter
Watch Carrier's quarterly disclosure of data center backlog conversion against the USD 1.5 billion 2026 revenue guide, since any shortfall there is the cleanest read on whether the 400% Q4 2025 order surge is shipping on schedule or rolling into 2027 [S2].
Track the delta between announced 2026 GW (roughly 16 GW) and under-construction GW (roughly 5 GW), plus any update to the 30–50% slippage band, as the leading indicator for whether chiller order intake re-accelerates or pushes construction machinery and equipment demand into the following year [S3].
Related analysis: Counter-Capability Procurement Signals: Space, Air, Maritime, and Information Domains.