3 Key Debates Post H1

Research evidence excerpt

3 Key Debates Post H1

ex. Our bottom-up work EV/EBIT** 41.1 32.4 26.9 22.8

shows how total hyperscaler compute capacity is set to reach ~120 GW by 2028, up DPS (SFr) 10.00 11.21 16.83 20.75

Div yld (%) 1.3 1.4 2.2 2.7

4x from ~30 GW in 2025. The forecasts imply higher capex spend and GW additions FCF yld ratio (%)** 1.0 1.4 2.2 2.7

in 2027 and 2028. So why the debate? Belimo removed quantitative top line Net debt (SFr mn)* (54) (74) (144) (192)

Net debt/EBITDA** NM NM NM NM

guidance for FY26. Qualitatively management expects continued strong demand RNOA (%)** 34.8 37.9 38.1 40.1

ROE (%)** 31.3 39.2 39.5 38.3

across traditional HVAC and data centres in H2, with strong dynamics across all

regions. The decision to remove the quantitative guide was put down to uncertainty Unlessframeworkotherwise noted, all metrics are based on Morgan Stanley ModelWare

on DC deployment. The market is conflating build timing and potential slippage with **§ ==ConsensusBased on consensusdata is providedmethodologyby Refinitiv Estimates

weaker end market demand, whilst reading management's decision not to quantify *e == GAAPMorganorStanleyapproximatedResearchbasedestimateson GAAP

H2 growth as a warning.

2) Is liquid cooling already saturated, and is the mix benefit fully realised?

Certainly not globally; the opportunity can shift from US penetration to regional

catch-up, more deployed GW and richer mix through higher rack densities and

renovated liquid cooled data centers. Our data shows that 36% of respondents have

Morgan Stanley does and seeks to do business with

liquid cooling deployed as of 2026, with forecasts of an 80% liquid cooling adoption

companies covered in Morgan Stanley Research. As a result,

rate for new facilities by 2030.