manufacturing capacity & why the power bottleneck will ease fordatacentersandbeyond.ForExcelaccess,contactyourBernsteinsalesperson.Themid-scalepower generation industry ismoving from shortage to surplus. chad.dillard@bernsteinsg.comAlasdairLeslie alasdair.leslie@bernsteinsg.comSunaina Ocalan Assuming all announced capacity expansionsgoes ahead,webelieve supplywill exceeddemand by~15-45Gw or 10-35%.Breaking it down byapplication,prime power willenteroversupplyafter2029-30,totaling 5-10 GwW,while standbypower will enter oversupplyafter2027-28,totaling~10-30GWby2030.OEMs areaggressively rampingmanufacturing capacity.We have aggregated every Madison Rezaei +1 917 344 8622madison.rezaei@bernsteinsg.comAdrien Rabier capacityexpansionannouncementsbyindustrial/aeroderivativeturbine (<12oMw/unit)andreciprocating engine(2-10MW/unit)0EMs.Between2024and2030,capacitywillquarterof thepowerconsumed in theUS,and 14xand 8xtheannual data center capacityadded in the US and worldwide,respectively.Reciprocating engines are thepreferredtechnology,giventheyaccountfor90%ofnew capacity.Returns are attractiveandbarrierstoentryarelow;thepayback on brownfield expansion can be<6months.As a result,newparticipants (Boom Supersonic, FTAl, etc.)are entering,and incumbents like CAT and CMIareaggressively expanding. +442076766820adrien.rabier@bernsteinsg.com +1917344 8430douglas.harned@bernsteinsg.com +1 917 344 8432miguel.marques@bernsteinsg.com and non-data center.Our US data center demand scenarios are5-25% CAGR on new adds,and we further split this demand into prime vs.standby,while we assume a 7% CAGR forinternational. As for non-data center demand, we assume a 5% CAGR, in line with the LTaverage and nominal GDP.Our 2030 total mid-scale power gen demand scenarios rangefrom~125-150GWoran incremental35-60GWfrom2024to2030fortheUS.Ineachscenario, supply outweighs demand; either US data center demand must grow at a +30%CAGRornon-datacenterdemandgrowthmustaccelerate to+10%/year. +44 20 7550 2192om.kela@bernsteinsg.com +44 20 7762 1411nicholas.witting@bernsteinsg.com Mid-scale power gen pricing will likely mean-revert. The industry currently enjoysHSD pricing vs.the historical LSD average. This premium vs. history is likely compress withoversupply,starting with standby in 2027-28.In contrast, to put a new dollar into CAT andCM, the market needs tounderwrite sustained 10% and 5%growth into 2030. Specialist Sales Steve Song+1 917 344 8401steve.song@bernsteinsg.com Keycompany conclusions.cMlismostexposedtothestandbymarketin ourcoverage(>95% of salesfromthis market through endof decade until their 4Mw primepowerengine is ready), so it may be the first to see pricing slow. CAT is better insulated from thestandby risk, given the greater flex manufacturing (prime growing as a % of sales). INio isbest positioned in this space,given itis narrowlyfocused onprime,notstandby JamesBrady+44 20 7762 5272james.brady@bernsteinsg.com INVESTMENTIMPLICATIONS US Machinery: we maintain our Market-Perform ratings on Caterpillar (TP, $879)and Cummins (TP, $700).Our workreinforces our view of unattractive risk/reward profiles given what investors need to believe to put a new dollar of capitalbusiness needs to growat a 28% CAGR (10% coming from price given volumes are limited by capacity).Similarlyfor CMl,theirPower Systems business needsto justifya47xmultiple,implyingthebusiness needstogrowata 17%CAGR(5%comingfromprice). Though the decision to increase capacity is the right economic decision for CAT and CMI (payback periods of<6 months),in a world of more abundant capacity (come 2028/29), volume growth will be increasingly competitive and pricing is likely tomean-revert to low-single digits (vs. the required high-singe digits). EuropeanCapitalGoods:WerateSiemens EnergyOutperform,witha pricetargetof 210. EuropeanAerospace&Defense:WerateRolls-RoyceMarket-Perform,withapricetargetof11.5. US Aerospace&Defense: Howmetis thekey supplier of turbineblades forindustrial gas turbines.IGT demand for primepower remains strong,as questions over demand levels should not come into question until the 2030 timeframe.The risksdescribedhere in the 2027-28 timeframe aremorefor standbypowerapplications,which arelesssignificant.GEAerospacesupplies aeroderivative enginesforpowerapplications.Thisdemand remainsstrong inthenear-term,as hasbeenreflected inrecent pricing trends. For GE, however, this is a small portion of its overall propulsion business. U.S.CommunicationsInfrastructure: We value DLR ona Price to AdjustedFunds from Operations (AFFO)per sharemultiple;our $232 price target is based on 27x our 2027EAFFO per share of $8.60.We value EQIX on a Price to Adjusted Funds FromOperations (AFFO)persharemultiple.Our $1,222price target isbased on25xour2027EAFFOper share of$48.63. GE Vernova:(GEV)-Outperform, Target Price S1,206: Right place at the right time,with one eye on the short term,andanotheronthelongterm GEVis poisedto becomean endto end powerand electrificatio