TipoftheIceberg:UnderstandingtheFullDepthofBigTech’sContributiontoUS

InnovationandCompetitiveness

HILALAKA|OCTOBER2025

Whilecriticsattack“bigtech”frommanyangles,thesefivecompaniesdevelopfrontier

technologiesthatrequirelarge-scaledevelopment,buildinfrastructurerangingfromdatacenterstosubseacables,andcreatespilloversfromhealthcaretonuclearenergy.

KEYTAKEAWAYS

BigtechcompaniesinvestmoreinR&Dthanbothmostcountriesandthefederal

government’snon-defenseresearchbudgetdo,providingthescalenecessarytopursuefrontiertechnologiesthatrequiremassiveup-frontinvestment.

Thesecompaniesdemonstrateauniquewillingnesstosustainbillionsofdollarsinlossesonhigh-risk,long-horizonprojectsthattraditionalinvestorsandevengovernmentscannotsupport.

Theiropenresearchmodelgenerateseconomy-widebenefitsthroughpublishedpapers,open-sourceframeworks,andfreelyavailabledatasets,enablinginnovationacross

academiaandindustry.

Technologyinvestmentsgeneratesubstantialspilloversintohealthcare,energy,

agriculture,transportation,anddefense—fromdrugdiscoverytocropdiseasedetectionsystems.

LargeU.S.techcompaniesprovidecriticalprivate-sectorcapacitytocompetewith

China’sstate-backedtechchampions,whichbenefitfrommassivegovernmentsubsidiesandcoordinatedindustrialstrategies.

Policydebatesthatfocusonlyonbigtechcompanies’harmswithoutrecognizingthesecontributionsriskunderminingthecapabilitiessupportingAmerica’stechnological

leadership,economy,andsecurity.

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATION|OCTOBER2025PAGE2

CONTENTS

KeyTakeaways 1

Introduction 3

SocietalBenefits 5

InnovationInvestment 5

UniqueFinancialCapacityforHigh-RiskR&D 5

ScaleCreatesFundamentallyDifferentInnovationPossibilities 8

Risk-TakingThatDrivesBreakthroughInnovation 10

CompetingWithState-BackedChineseCompetitors 11

OpenResearchandTalentConcentration 12

OpenResearchModel 12

TalentMultiplierEffects 15

StrategicInfrastructure 17

PrivateCapitalFillsCriticalInfrastructureGaps 17

InfrastructureOwnershipasGeopoliticalLeverage 17

StrategicCompetitionforGlobalDigitalInfrastructure 19

Dual-UseTechnologiesandNationalSecurity 20

TheCommercialTechnologyFoundationforDefenseInnovation 20

MilitaryAdaptationofCommercialPlatforms 22

SpilloversIntoOtherStrategicSectors 23

HealthCareandBiomedicalResearch 23

ManufacturingandMaterialsScience 24

AgricultureandFoodSecurity 24

EnvironmentalScienceandClimate 24

NuclearEnergy 25

TransportationandUrbanMobility 25

QuantumComputingBreakthroughs 26

NurturingtheStartupEcosystem 26

CreatingExitMarketsandProvidingCapital 27

RescuingandScalingPromisingTechnologies 28

AmericanSoftPower 29

CultureandNarrative 29

DevelopmentSubstitution 30

RecognizeandAccountforthe“FullIceberg” 30

Conclusion 31

Endnotes 32

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATION|OCTOBER2025PAGE3

INTRODUCTION

Aslateastheearly2010s,largeU.S.informationtechnologycompanieswerecelebratedby

politicians,media,andacademics.Nowtheyattackbigtechastoopowerfulandgreedy.Liberalscomplainthattheyaresurveillingsociety,spreadingharmfulcontent,andundermining

democracy.Conservativessaythattheyarecensoringfreespeech,influencingelections,and

underminingtraditionalAmericanvalues.Bothsidesbelievethattheyareharmingchildrenandbothrightandleftpopulistswanttobreakthemup.

1

This“voguish”critiqueisbasedonatbestasuperficialunderstandingofwhatthesecompaniesdo:ApplemakesiPhones;Amazonmakesanonlinestore;Alphabetmakesasearchengine;

MicrosoftmakesWord,Excel,andPowerPoint;andMetamakessocialnetworks.Especiallyforthosewhodismisssocialmediaasatbesttrivialentertainment,thisviewseesonlythetipoftheicebergofthemassivebenefitslargetechnologycompaniesprovidetotheworldandtheUnitedStates.

Atthesametimecriticsattackbigtech,theyalsoarguethatU.S.corporationsdontinvest

enoughinthelongterminresearchanddevelopment(R&D),incapitalinvestment,ineffortsthathavehighrisksandlong-termreturns.Ifonlythese“quarterlycapitalism”companieswouldactbetter,everyonewouldbebetteroff.Andyet,bigtechisdoingtheoppositeofquarterly

capitalism,andthoseeffortsshouldbepraisedratherthanexcoriated.

WhilemanymajorU.S.technologyfirmsincludingCisco,Qualcomm,andNvidiademonstratesimilarinnovationpatternsandpositivespillovers,spaceconstraintslimitthisreportto

examiningfive:Apple,Amazon,Alphabet,Meta,andMicrosoft.

R&D:In2024,thefivelargestU.S.technologycompaniesinvested$227billioninR&D,

exceedingthefederalgovernmentsnon-defenseR&DbudgetandoutpacingtheannualR&Dexpendituresofmostcountries.

2

Theseinvestmentswerenotlimitedtocorebusinesslines.

ResearchbytheInformationTechnologyandInnovationFoundation(ITIF)findsthat

approximately78percentofnon-manufacturingR&Dgrowthcamefromjustfourindustries:

electronicshippingandelectronicauctions;dataprocessing,hosting,andrelatedservices(e.g.,cloudproviders);softwarepublishers;andotherinformation(e.g.,Internetcompanies).

3

Bigtechcompaniesalsofundlong-horizonresearchintoartificialintelligence(AI),quantumcomputing,andsemiconductors;theyunderwriteexperimental“moonshot”projectsthatmayfail

commerciallybutadvancetechnicalcapabilities;andtheygeneratespilloversthatdiffuseacrossacademia,startups,andindustrymorebroadly.

4

CapitalInvestment:In2024,thesecompaniesinvestedover$250billionincapitalexpenditures,andcumulatively,theyhaveinvestedover$1trillioninthe21stcentury.

5

Trainingacompetitivelargelanguagemodelcancostmorethan$100millionincomputeresources.

6

Developing

quantumprocessorsrequiresbillion-dollarfabricationfacilitiesandprogramsextendingoveradecadeormore.

7

Buildingglobalsatelliteconstellationsorhyperscalecloudinfrastructure

involvesup-frontinvestmentsinthetensofbillionsofdollars.

8

Amazon,Alphabet,Microsoft,

Meta,andApplefinancedatacenters,subseacables,renewableenergyprojects,andcustom

semiconductorplatformsthatunderpinnotonlytheirownoperationsbutalsothebroaderdigitaleconomy.

9

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATION|OCTOBER2025PAGE4

PublicKnowledgeGeneration:Bigtechfirmspublishthousandsofresearchpapersannually,

releaseopensourceframeworks,andmakelargedatasetsandmodelsavailabletouniversitiesandstartups.

10

InitiativessuchasGoogleDeepMindsAlphaFold,whichmapped200millionproteinstructuresandreleasedthemfreely,acceleratebiomedicalresearchworldwide.

11

ThisopennessdistinguishesU.S.firmsfrommoretraditionalcorporateR&Dmodels,wherein

proprietaryknowledgerarelycirculatesoutsidecompanyboundaries.

STEM(Science,Technology,Engineering,andMath)SkillDevelopment:Thesecompaniesserveastraininggroundsforhighlyskilledresearchersandengineerswholatermoveintostartups,

universities,andgovernmentlaboratories.Alumshavefoundedthousandsofcompanies,raisingtensofbillionsofdollarsinventurefundinganddiffusingorganizationalknowledgeabout

buildingscalablesystems.

12

InsectorssuchasAI,whereglobalcompetitionfortalentisintense,thisclusteringeffectstrengthenstheU.S.ecosystemasawhole.

13

FundingforStartups:Acquisitionsandpartnershipslinklargefirmstothestartupecosystem,

providingcapital,technicalinfrastructure,andglobalreach.Whilecriticsdescribethisas

anticompetitive,empiricalevidenceshowsthatacquisitionactivityoftenvalidatesmarketsandattractsfurtherventureinvestment.

14

High-profilecases,suchasthatofYouTubeandAndroid,demonstratehowintegrationwithlargerplatformscanrescuepromisingtechnologiesfrom

collapseandscalethemtohaveaglobalimpact.

15

Blockingacquisitions,bycontrast,hasat

timesproducedtheoppositeoftheintendedresult,leadingtofirmfailureandstrategiclossestoforeigncompetitors.

16

InnovationsBeyondtheInformationSector:BigTechisprimarilyaboutinformationgeneration,storage,analysis,andtransmission.Butitisaboutmuchmorethanthat.Thesecompaniesareinnovatinginarangeofimportantareas,includingnationaldefense,transportation,agriculture,education,healthcare,spacetravel,andmore.

TheUnitedStatesfindsitselfinanexistentialtechno-economicwarwithChina,wherethe

ChineseCommunistParty(CCP)isseekingtodominatealladvancedindustries,including

manufacturingandinformation.

17

Andithascoreadvantages,includingacloseddomestic

market,massivegovernmentsubsidies,andfavorableantitrustrulesthatencouragesizeandscale.GiventhesevereU.S.governmentfiscalconstraints,itisnotrealistictoexpecttheU.S.

governmenttomatchthefinanciallargesstheCCPbestowsonitsbigtechchampions.IftheUnitedStatesdoesnotwanttolosetoChina,itneedstostrengthen,notweaken,itslarge

technologycompanies.

ToomanyhavedefaultedtoadoptingtheEuropeanmodelofhobbling,Gulliver-like,Americas

largetechnologycompaniesthroughtaxes,regulations,fines,antitrustbreakups,andother

restrictions.

18

TheideathatdoingsowillboostU.S.innovationandcompetitorsissofarcicalthatitsamazingthatanyonecanbelieveit.IfwecontinuedownthatBrusselspath,theresultwillbeclear:anAmericawhere,inadecadeortwostime,todaysbigtechfirmsarelookedat

nostalgicallythewaywelookatpasttechleaderssuchasXerox,RCA,Lucent,andKodak.

Thatiswhy,attheveryleast,whattheU.S.policycandowithregardtoitsownbigtech

companiesisnoharm.DoingsowillcertainlystrengthenAmericasglobalcompetitivenessand

thejobsbigtechcreatesathome.ItwillalsosupportU.Snationalsecurity.

19

AmericanadvancesinAI,semiconductors,andquantumcomputingareoverwhelminglydrivenbyprivateR&D.The

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATIONIOCTOBER2025PAGE5

PentagonadaptsthesecapabiIitiesfordefenseappIications,benefitingfrombiIIionsin

commerciaIinvestmentthatgovernmentcontractsaIonewouIdnotreproduceatasimiIarscaIeorspeed.

20

AnditwiIIheIpenabIeinnovationacrossarangeoftraditionaIsectorswheretraditionaIfirmsareunderinvestinginhigh-riskactivities.

OveraII,theprevaiIingdiscoursereducesthesefirmstoconsumerpIatformsandmarketpower,

overIookingthebroadersystemofinnovation,infrastructure,andsecuritythatdependsontheirscaIe.PoIiciesthattreatbigtechprimariIyasaprobIemriskweakeningassetsthatunderpin

U.S.competitiveness.AmorecompIeteassessmentrequiresrecognizingthepositiveexternaIitiestheygenerate.

SOCIETALBENEFITS

InnovationInvestment

R&DdetermineswhichnationsIeadintechnoIogicaIadvancementandeconomic

competitiveness.

21

ThesourceandscaIeofR&DfundingshapewhattechnoIogiesgetdeveIoped,howquickIytheyreachmarket,andwhichcountriescapturetheresuItingeconomicbenefits.

Today,privatecorporationshaveemergedasthedominantfundersofR&D,withfiveU.S.

technoIogycompaniesaIoneinvestingmorethantheentirefederaIgovernmentandmostothercountries.

Thisconcentrationofresearchcapacitycreatesdistinctiveadvantages,nameIytheabiIityto

sustaindecade-IongIossesonexperimentaItechnoIogies,pursueinnovationsrequiringbiIIionsinup-frontinvestment,maintainportfoIioapproachestohigh-riskresearch,andcompetewithstate-backedentitieswithoutdirectgovernmentsupport.

TheseinvestmentsarefundamentaIIyprofitdriven.ButwhiIecompaniespursueR&Dtomaintainacompetitiveadvantageandcapturefuturemarkets,theyaIsogeneratesubstantiaIpositive

externaIities.ThetechnoIogies,knowIedge,andcapabiIitiesdeveIopedthroughcorporateR&DfrequentIydiffusebeyondtheinvestingfirms,creatingbenefitsforcompetitors,academia,andsocietyatIargethatexceedtheprivatereturnscapturedbythecompaniesthemseIves.

UniqueFinanciaICapacityforHigh-RiskR&D

TransformativetechnoIogiestypicaIIyrequireextendedperiodsofinvestment,withhighIy

uncertainreturns.Thetransistortook23yearsfromitsinventionatBeIILabstoreach

widespreadcommerciaIadoption.

22

TheInternetevoIvedoverdecadeswithgovernmentsupportbeforeprivatecommerciaIization.

23

TheseextendedtimeIines,combinedwithhighfaiIurerates,don’tmakeitverydesirabIetofundresearchthatmaytakedecadestopayoff,ifever.

TraditionaIcapitaImarketsdemandreturnswithin5-10years.GovernmentgrantsoperateonevenshortercycIes.ThismismatchbetweeninnovationtimeIinesandfundinghorizons

constrainsbreakthroughresearch.

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATION|OCTOBER2025PAGE6

Figure1:Bigtechcompanies’R&Dspending,2015–2024(billions)

24

$250

$200

$150

$100

$50

$0

2015201620172018201920202021202220232024

AmazonMicrosoftApple

Meta

Alphabet

TheleadingfiveU.S.technologycompanies—Amazon,Apple,Alphabet,Meta,andMicrosoft—invested$227billioninR&Din2024.

25

Thisexceedsthe$172billionU.S.federalR&DbudgetandthetotalannualR&Dexpenditureofmostcountriesin2023.

26

Thescalerepresentsnotjustlargenumbersbutalsoaqualitativedifferenceininnovationcapacity.Whenindividual

companiesdeploytensofbillionsofdollarsannuallyonresearch,theycansustainprojectsthatwouldexhausttraditionalfundingsources.

Figure2:U.S.privatesectorR&Dspending,2015–2024(billions)

$700

$600

Allothers

$500

$400

$300

AllothertechBigtech

$200

$100

$0

2015201620172018201920202021202220232024

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATION|OCTOBER2025PAGE7

Google’sparentcompany,Alphabet,demonstratesthiscapacitythroughitsexperimentalX

division.In2016,thedivisionreportedquarterlylossesof$859millionpursuing“moonshot”projects,withestimatedannuallossesbetween$3billionand$4billion.

27

Theseprojects

includedInternet-deliveryballoons,glucose-monitoringcontactlenses,andautonomous

vehicles—technologieswithuncertaincommercialprospectsbutpotentiallytransformative

impacts.Despitethelosses,Alphabet’sstockrose6percentaftertheearningsannouncement,reflectinginvestoracceptanceoflong-termresearchinvestmentssupportedbyprofitablecorebusinesses.

Figure3:U.S.bigtechcompanies’R&Dspendingvs.selectcountries’totals,2015–2024(billions)

28

$250

U.S.bigtech

Japan

$200

Germany

$150

Korea

$100

France

$50

$0

201520162017201820192020202120222023

Alphabet’sexperimentalXdivisionreportedquarterlylossesof$859millionpursuingmoonshot

projects.Despitethelosses,Alphabet’sstockrose6percent,reflectinginvestoracceptanceoflong-termresearchinvestments.

Amazon’sapproachtohardwaredevelopmentillustratessustainedcommitmentdespite

commercialfailure.Thecompany’sdevicesdivision,centeredonAlexa-enabledproducts,lost

over$25billionbetween2017and2021.

29

Injustthefirstquarterof2022,thedivisionlost

another$3billion.

30

ThecompanysoldEchodevicesatorbelowcost,hopingtogenerate

revenuethroughvoicecommercethatnevermaterialized.Yet,Amazonsustainedtheselossesfornearlyadecade,advancingvoicerecognitionandnaturallanguageprocessingtechnologiesintheprocess.

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATION|OCTOBER2025PAGE8

Customsemiconductordevelopmentrepresentsanotherdomainwhereonlythelargest

technologycompaniescancompete.Apple’stransitiontoproprietarysiliconrequiredyearsofdevelopmentandbillionsininvestment,culminatinginchipsthatnowpoweritsentireproductline.Thisverticalintegration,designingprocessorsspecificallyoptimizedfortheirsoftwareandAIworkloads,requiressustainedinvestmentthatfewcompaniescanmatch.

Figure4:Subsetofbigtechcompanies’lossesonhigh-risk,high-rewardmoonshotprojects

31

$5B

MicrosoftHoloLens

$10B

AppleProjectTitan

$20B

AlphabetOtherBets

$25B

AmazonAlexa/Echo&Devices

$60B

MetaRealityLabs

Thisfinancialcapacityderivesfromprofitablecorebusinessesthatgeneratemassivecashflows.AmazonWebServices(AWS)produced$62billioninrevenuein2021withoperatingmarginsexceeding30percent.

32

Alphabet’sadvertisingbusinessgeneratesover$200billionannually.

33

Bigtechcompaniesfunnelsignificantsharesofprofitstoresearchportfoliosthatnoother

entitiescaneasilymatchinscaleorduration.Theprivatesectorcross-subsidizationmodelbypassescapitalmarketconstraintsthatwouldterminateunprofitableresearchintraditionalsettings.

ScaleCreatesFundamentallyDifferentInnovationPossibilities

Frontiertechnologiesexhibitescalatingcapitalrequirementsthatcreatenaturalbarriersto

participation.BreakthroughsinAI,quantumcomputing,andautonomoussystemsdemand

massiveup-frontinvestments.Theserequirementsstemfromfundamentaltechnicalconstraintsratherthanartificialbarriers.

Traininglargelanguagemodelsexemplifiesthescale.OpenAI’sGPT-3requiredover$4.6millionincomputecostsaloneforasingletrainingrun.

34

GPT-4pushedcostsabove$100million.

35

Industryestimatessuggestthatnext-generationmodelswillrequirebillionsofdollarsin

infrastructureinvestment.Theexponentialrelationshipbetweenmodelcapabilityand

computationalrequirements,whereineachorderofmagnitudeimprovementdemandsroughlytwoordersofmagnitudemorecompute,createsmountingbarrierstoentry.

36

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATION|OCTOBER2025PAGE9

Figure5:R&Dandinfrastructurecostsforfrontiertechnologies

37

$200M

$400M

$1B

ComputingandAI

Globalconnectivity

Next-genenergyandmobilityAdvancedmanufacturing

$3B

FrontierAImodeltrainingTransoceanicunderseacableQuantumcomputingprogramHyperscalecloudregion

$4B

Next-genAItrainingcluster(100KGPUs)

$5B

EVGigafactory

$10B

$10B

Autonomousvehicledevelopment

$20B

$50B

LEOsatelliteconstellationCommercialfusionpowerdevelopmentLeading-edgesemiconductorfab(3nm)

OnlyorganizationswithexceptionalresourcescandevelopAImodelsthatrelyondedicated

accesstohundredsofthousandsofspecializedprocessors.WhenOpenAIneededtoscaleGPT-4training,itsecuredresourcesthroughanexpandedpartnershipwithMicrosoft,whichprovided

bothfundingandexclusiveaccesstoAzureinfrastructure.

38

Smallerorganizations,regardlessoftechnicaltalent,cannotaccesscomparablecomputationalresourcesbecausethecostissimplytoogreat.

Autonomousvehicledevelopmentpresentssimilarscalerequirements.Alphabet’sWaymohasinvestedover$5billionsinceitsinception,fundingnotjustresearchbutalsoextensive

infrastructure.

39

Theprogramrequirescustomsensordevelopment,high-definitionmappingofentiremetropolitanareas,sophisticatedsimulationenvironmentsprocessingmillionsofvirtualmiles,andreal-worldtestingfleets.Eachcomponentdemandsspecializedexpertiseand

sustainedfundingthatfeworganizationscanprovide.

Quantumcomputingpushesscalerequirementsevenfurther.Alphabet’s2019achievementofquantumsupremacymarkedtheculminationof13yearsofsustainedinvestment.

40

Theprogramrequiredcustomfabricationfacilitiesforquantumprocessors,cryogenicsystemsmaintaining

temperaturesnearabsolutezero,andinterdisciplinaryteamsspanningphysics,computer

science,andengineering.

41

Amazon,IBM,andMicrosoftarepursuingsimilarprograms,eachinvestingbillionswithoutnear-termrevenueprospects.

42

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATION|OCTOBER2025PAGE10

Theseexamplesillustratehowtechnologicalcomplexitydrivescapitalconcentration.As

innovationsrequireintegrationacrossmultipledisciplines,longerdevelopmentcycles,andmoresophisticatedinfrastructure,theminimumviableinvestmentrisescorrespondingly.Thiscreatesanaturalfilterwhereonlyorganizationswithexceptionalfinancialcapacitycanpursuefrontier

technologies.

Risk-TakingThatDrivesBreakthroughInnovation

Innovationinherentlyinvolvesexploringmultipletechnicalpathswithuncertainoutcomes.

Historicalanalysisofmajortechnologiesrevealsconsistentpatterns:numerousfailedapproachesprecededeachbreakthrough,successoftenemergedfromunexpecteddirections,andthe

ultimateapplicationsdifferedfrominitialvisions.

43

Effectiveinnovationsystemsmustthereforesupportportfolioapproaches,pursuingmultiplecompetingsolutionssimultaneouslywhile

maintainingtoleranceforfailure.

BusinessscholarClaytonChristensenandotherslongcriticizedU.S.corporationsforfocusingonwhathetermedincremental,ratherthanbreakthrough,innovation.

44

Theformerismakingsmall,gradualimprovementstoexistingproductsforcurrentcustomersratherthanpursuing

transformativetechnologies.Andhewasright.Toomanycompaniesbecametoorisk-averse,losingmarketsharetoforeigncompaniesoftenbackedbythedeeppocketsofthestate.

Largetechnologycompanieshavefocusedsignificantlyonbreakthroughinnovationand

incorporatedsignificantrisk-takingpractices,whicharedifficultfortraditionalorganizationstomatch.Manyoftheseprojectsfailandnevermakeittofullcommercialization,butatleastthesecompaniestooktherisk.Otherprojectsdosucceed.Microsofthaspursuedtopologicalquantumcomputingsince2005,investingbillionsofdollarsinanunprovenapproachthatcould

revolutionizethefieldorproveimpossible.

45

MetafundsneuralinterfaceresearchthroughitsRealityLabs,exploringelectromyography(EMG)wristbandtechnologyforhuman-computer

interactiondespitelongdevelopmenthorizons.

46

Microsofthascommittedtopurchasing

electricityfromHelionPower’sfusionreactorsstartingin2028,althoughnocompanyhasyetsuccessfullyproducedcommercialfusionpower.

47

Thistoleranceforextremetechnicalrisk

enablesthepursuitoftransformativetechnologieswhereintheoreticalfeasibilityremainsunproven.

48

Theportfoliostrategyextendsbeyondsimplefailuretolerance.Thesecompaniespursuemultiplecompetingtechnicalapproachessimultaneously,hedgingagainsttechnicaluncertainty.InAI,

majorfirmsmaintaininvestmentsacrossdifferentmodelarchitectures,trainingapproaches,andapplicationdomainsratherthancommittosinglepaths.MicrosoftinvestsinbothitsownAI

researchandtheOpenAIpartnership.AlphabetbothdevelopsTensorProcessingUnits(TPUs)andmaintainsGPUinfrastructure.

49

Thisparallelexplorationincreasestheprobabilitythatatleastoneapproachsucceeds.

Nvidiaexemplifiesthisstrategythroughitsmulti-trackchipdevelopment,whichrunsthree

designteamssimultaneouslyonstaggered18-to24-monthcycles,allowingforcontinuous

innovationwhilehedgingagainstthefailureofanysinglearchitecture.

50

Thecompanyinvestednearly$10billiondevelopingitsB200Blackwellchipalone,whilemaintainingparallel

investmentsinitssoftwareecosystem:over900CUDA(ComputeUnifiedDeviceArchitecture)librariesthatensureitshardwareadvantagestranslateintopracticalapplications.

51

INFORMATIONTECHNOLOGY&INNOVATIONFOUNDATION|OCTOBER2025PAGE11

FinancialresilienceenablescountercyclicalR&Dinvestment,increasingresearchspending

duringeconomicdownturnswhencompetitorsretreat.Academicresearchdemonstratesthat

companiesthatmaintainR&Dinvestmentthroughrecessionsemergewithsignificantadvantagesduringrecovery.

52

Whilesmallerfirmsmustcutresearchspendinginordertosurvivedownturns,andlarger,moretraditionalfirmscuttokeeptheirsharepriceup,largetechnologycompanies

canaccelerateinvestmentwhenresources(talent,equipment,facilities)becomemoreavailableandaffordable.

Theriskcapacitymanifestsinthewillingnesstopursuetechnologieswithextremeuncertainty.GoogleBrainbeganasaspeculativeprojectwithunclearcommercialapplications.

53

Theteamexploredwhetherdeeplearningtechniques,largelyabandonedbytheAIcommunity,could

achievebreakthroughresultswithsufficientscale.Theprojectconsumedmassivecomputationalresourceswithnoguaranteeofsuccess.Today,thetechnologiesdevelopedgenerateover$10

billionannuallythroughsearchimprovementsandcloudservices.

CompetingWithState-BackedChineseCompetitors

Chinasinnovationmodelposesasystematicchallengetomarket-basedsystemsthroughthe

coordinateduseofstateresources.TheChinesegovernmentallocatedover$450billiontoR&Din2023,representingan8.1percentincreaseaimedatsustaininggrowthabove7percent

annually.

54

Thisspendingoperateswithinanintegratedsysteminwhichstate-ownedenterprises,researchinstitutes,andnominallyprivatecompaniesalignwithnationalobjectives.Recent

initiativesincludea$138billionfundforquantumcomputing,AI,andsemiconductors.

55

Thesestatesubsidiesconstitutemorethanaggressivecompetition;theyviolatefundamental

principlesofinternationaltrade.AsITIFhasdocumentedextensively,Chinassystematicuseofbelow-marketfinancing,productionsubsidies,andstate-directedovercapacityrepresents

predatoryinnovationmercantilismdesignedtocaptureentireindustriesrather