铜的长期展望：围绕数据中心和铜的三个误解

copper Weorganizein foursectionsasfollows:wesummarize ourcallinafewcharts.We thenreviewwhat'snew&improvedinthis currentoutlook.Finally,weprovidethe completedetails.We wait fora better entry point intoafuture coppermarketwe expectto tighten.On demand, our updated data center (DC) work highlights three areas where +1917344 8422bob.brackett@bernsteinsg.comAndrianto Guntoro, CFA andrianto.guntoro@bernsteinsg.com approaching~200 GW by 2030, copper intensity is expected to fall meaningfully (Exhibit10).The shift toward 800 VDC architecture, particularly in Al training facilities, reducescopper requirements by as much as 45%, with a more gradual impact in inferencing.Second,theread-throughtoT&D capex and copperdemand withingrids isless direct than headline growth implies.Interconnection delays across the U.S.and Europeare pushing more operators toward behind-the-meter (BTM) solutions, with up to~33-44% of demand potentially met on-site (Exhibit 21).This dampens near-term grid build-outintensity. However, given that BTM is stll viewed as a bridge rather than a substitute, we stillby~$50-80bnperannumover2030-2050(Exhibit22). copperdemandthandatacenters.Capacityis settoscalefrom~550GWh in2025to~1,500 GWh by2030 (Exhibit 26),driven by rising renewable penetration and theneedfor system flexibility. Unlike data centers,we do not see ameaningful near-term decline inion, likely to scale in the 2030s. On the supplyside,werevisitkeyprojectsand adjust ourconvictionlevels (Exhibit23).We upgrade BHP's Escondida OGP2 and First Quantum's Cobre Panama to"GoingAheadreflecting improved visibility on execution.Conversely,wetakea more cautiousview on Barrick's Reko Diq,downgrading it to"Probable"We also refresh the status ofseveral other projects, including Elang, Coroccohuayco, Copperwood, Nifty Restart,Cascabel, and Gunnison SX-EW. In terms of investment ideas: Weseecopperpricecurrentlyelevatedwithreasonably skewed risktothedownside.We see copper-focused equitiesmostly reflecting optimism on copperasthey trade at meaningful premium to their 5-year average multiples (Exhibit 46).For example,Lundin and Southern Copper aretrading at more than+2 standarddeviations above theirrespective 5-year averages.In comparison,ANTO (Market-Perform)and FCX (Market-Perform)are trading atapproximately+1.7 and+0.6 standard deviations above their 5-yearaverages.With2027consensus atc.$12,500/t,valuations implyexpectationsfor>$13,500/t copperprice.WereiterateourMarket-Performratingon our copperpureplays ANTO and FCX. INVESTMENTIMPLICATIONS We have Outperform ratings on ABX, NEM and RIO. We have Market-Perform ratings on AAL, ANTO, BHP, BOL, FCX, GLEN andVALE. ANTO (Market-Perform) We raise our pricetarget from GBP26.00 to GBP 28.00 to reflect ourlatest commodity price deck and exchange rateestimates.Weuseanunchanged25/75combinationof DCFand an EV/EBITDAmultipleof 8.Oxagainst ourforward2027EBITDAestimate. FCX (Market-Perform) WeraiseourpricetargetfromUSD53.50toUSD58.50toreflectourlatestcommoditypricedeck,andexchangerateestimates.We continue touseEV/EBITDAmultipleof6.50xagainstourforward2027EBITDAestimate. What'sNewWhat's New-Integrating 8oovDC Architecture intoOurData Center Copper Demand Model..9What's New - Updating Our Supply/Project Database...18What's New- Updating Demand from energy storage Systems (ESS)..21Full details...25SupplyIntroduction25SupplyForecast...27Demand....28Demand Forecast...29Market Balance....30Pricing Copper...30Copper Thesis - How Much is Already Priced-ln?...34Appendix...36.Copper Long-term Charts..36CopperDemand by SectorFactors Causing Tight(er) Copper Supply..58Copper Wallof Supply...77Copper Supply Calculation Methodology....84TheTopCopper CompaniesandRecentPerformance.85Definitions..86 DETAILS becomemorepositiveon BHPs'Escondida OGP2 (organicgrowthprogram)and First Quantum's CobrePanama projectandof Barrick's Reko Diqto progress, and assign"Probable"status (previously Going Ahead).We also updatethe status of otherprojects like Elang,Coroccohuayco,Copperwood, Nifty Restart, Cascabel and Gunnison RestartSxEw in Exhibit 23. demandfrom data center capacitygrowthtobe offset bydecliningcopper intensity.As a result,we see demandto stay requires more flexibility to manage intermittency,curtailment, peak shifting and ancillary services. More importantly,we don'tforesee declining copper intensity in ESS battery in the short-term. The shift from LFP to Sodium-ion technology, which hasdeficit. Overall,we still see an overalldeficit in2026 (responsiblefora portion but not the entirety ofcopper price runup),anda finelybalanced marketbetween2027and 2029,followed bywideningdeficitfrom2030 onward (Exhibit 1). likely that a significant supply deficit begins to open up and growfrom 2030 onwards as supply growthfails tokeep up with demand from electrification. Werevise our H22026price estimates higherto$11,500/t(Exhibit 2)as we continuetoexpectnotariffsonrefined copperimport to the US will be introduced this month (Glo