Global electricity demand is forecast to expand at a 3.6% average annual rate across 2026-2030, according to the Electricity 2026 report [S1].
For pump specifiers, the implication is direct: the demand curve is decoupling from GDP. Utilities, hyperscale data centres, district cooling and desalination now set the marginal kW, while EU industrial production indicators in spring 2026 remain soft, damping baseline chemical and refining pump orders [S2].
Electricity 3.6% CAGR sets the macro pull on pump intake
For a 500 MW combined-cycle block, the auxiliary pump population typically spans 8-12 large units (boiler-feed, HP/LP condensate, cooling-water, circulating-water and service-water), each rated between 50 kW and 6 MW. Multiply that across the implied ~2,400-3,000 GW of net new and replacement capacity the IEA scenario implies, and auxiliary-pump volume tracks generation capex with a multiplier between 0.4-0.8% of plant cost, a level corroborated by Franklin Templeton's reading of the same IEA dataset [S3].
Data centre load: 415 TWh in 2024 rising to ~945 TWh by 2030
Data centre electricity consumption reached ~415 TWh in 2024 and is projected to climb to ~945 TWh by 2030, equivalent to roughly 3% of total global electricity demand, per IEA figures cited by Franklin Templeton [S3]. Cooling and process-water pumps for hyperscale campuses are the most direct beneficiaries.
A typical 100 MW liquid-cooled AI campus can require 30-60 MW of chilled-water and condenser-water pumping, with redundancy configured as N+1 or N+2. That translates into 25-60 medium-voltage centrifugal pumps per site, with flow rates of 200-1,500 m3/h and heads of 20-60 m. Chillers and CDUs (coolant distribution units) used in two-phase or single-phase direct-to-chip loops rely on a different sub-class: smaller, high-head end-suction or multistage units at 5-50 m3/h and 40-80 m head, often in stainless 304/316 or duplex builds to handle glycol-water mixtures.
EU industrial production: spring 2026 still soft, chemical/refining pumps lag
The European Commission's Spring 2026 forecast frames industrial production and confidence indicators in a subdued band, with ESI, PMI and production expectations all sitting below the 2018-2019 average and pricing expectations still volatile [S2]. For chemical, refining and general-process pump OEMs, this confirms a 2026-2027 replacement-driven order book, not a volume expansion.
Process-industry buyers in the EU should plan around API 610 (centrifugal) and API 674/675 (positive-displacement) overhauls on a 5-7 year cycle rather than net-new unit additions, and the EC's indicators support that cadence: new-capacity FID announcements for EU bulk chemicals and refining are not in the spring 2026 document, while maintenance, decarbonisation and electrification retrofits dominate listed capex themes [S2]. For spec detail on the rotating-equipment side, see the hydraulic system manufacturing cost breakdown 2026 spec-first guide for the parallel motion-control capex profile.
Where the demand actually lands: end-use split 2026-2030
Pulling the IEA and Franklin Temploten datasets together, the 2026-2030 incremental pump demand splits into four buckets: (1) power-generation auxiliaries, (2) data-centre cooling and water, (3) water-stress-driven desalination and reuse, and (4) general industrial (chemical, refining, mining, pulp and paper). Buckets 1-3 grow above the IEA's 3.6% power CAGR; bucket 4 tracks EU industrial production at low-single-digit growth and is exposed to the soft confidence backdrop [S1][S2][S3].
A useful side-by-side for procurement teams:
Bucket 1, Power-gen auxiliaries: flow 50-3,000 m3/h, head 20-400 m, materials ASTM A216 WCB / 12% Cr / SS316, standards API 610 and ASME B73.1, typical lead 16-26 weeks. Bucket 2, Data-centre cooling: flow 5-1,500 m3/h, head 10-80 m, SS304/316 and duplex, N+1/N+2 redundancy, lead 8-16 weeks on stocked frames. Bucket 3, Desalination and reuse: flow 100-5,000 m3/h, head 40-80 m (RO high-pressure 600-800 m), super-duplex and titanium, ASME B73.1 and API 610, lead 20-40 weeks. Bucket 4, General process: flow 1-500 m3/h, head 5-200 m, broad material menu, API 610 / API 674/675 / ISO 5199, lead 12-24 weeks.
McKinsey projects global compute workloads, AI and non-AI combined, will nearly triple by 2030, reinforcing that bucket 2 is the highest-velocity line item even if bucket 1 carries the largest absolute volume [S3]. For buyers also sizing compressed-air and pneumatic systems into the same capex envelope, the pneumatic system 2026 three-tier supplier map and the air compressor market share 2026 ranking sit adjacent to pump selection on the utility side.
Spec boundaries that move with the demand mix
Two technical boundaries are tightening under the new demand mix. First, minimum efficiency thresholds: IE5 / premium-efficiency motors paired with VFDs are now the default for any pump >7.5 kW in EU industrial builds, per the EC's efficiency-policy direction signalled in the spring 2026 package [S2]. Second, materials: super-duplex (UNS S32750 / S32760) and titanium grade 2 are now standard for seawater and high-chloride cooling loops, replacing 904L in many hyperscale specifications.
For data-centre secondary loops, the spec floor is rising towards 70/30 or 60/40 propylene-glycol-water with stainless 316L wetted parts, and pump seal plans increasingly default to API Plan 52/53A rather than Plan 11, because glycol contamination of gland water is no longer acceptable. Buyers who have not yet refreshed their industrial-pump material-and-seal matrix against these two trends are likely to see shortlists shrink at the bid stage.
Risks that can bend the curve down
Three downside vectors are visible in the source set. EU industrial sentiment is below long-run averages, so bucket 4 can stay flat through 2027 if chemical and refining margins do not recover [S2]. Hyperscaler power-procurement speed is the binding constraint on bucket 2, with renewable PPAs hitting multi-year price highs as grid integration lags, and that cost pressure feeds back into cooling-system capex pacing [S3]. A third, smaller vector: permitting and lead-time for large desalination builds, where 20-40 week pump deliveries can sit behind 18-30 month EPC schedules, so orders cluster in waves rather than a smooth run.
None of these vectors invalidates the 2026-2030 base case, but each argues against a single-line "pump demand up X%" model. For a more complete adjacent read on the materials side, the rare-earth sourcing from China 2026 procurement guide covers the magnet and motor input chain that IE5/VFD packages depend on.
Trackable signals for the next quarter: IEA mid-year electricity revisions, EC Autumn 2026 industrial-production update, and hyperscaler quarterly capex disclosures for cooling and water infrastructure spend, all three of which will narrow or widen the band above and below the 3.6% power CAGR.
Component reference pages worth checking: industrial adhesive, and industrial borescope.