2025全球人工智能（AI）的未来研究报告

The Future of Global A.I. Siddharth Yadav Abstract As the global economy moves towardsubiquitous digitisation, the demandforand generation of data isexperiencing exponential growth,as are computational requirementsand user adoption of AI products and services.Thisgrowth is forcing big-tech incumbents toexpandcapital expenditure.Undergirding thisdynamic is the confluence of increasing frontiermodel training costs and plummeting inferencecosts.Revenue streams of AI companies arefurther being strained due to the rise of open- source systems offering competitive performance.Toexplore this techno-economic matrix,thisreport analyses the findings of BOND’s ‘Trendsin Artificial Intelligence’ report released in May2025. The analysis is followed by recommendationsforaccelerating adoption through inclusivepractices, supporting infrastructure developmentand strategic capacity-building in the Middle Eastand India. Introduction: The Present andFuture of Digital Intelligence ArtificialIntelligence(AI)—atechnology that was once conjuredinthe accelerationist dreams ofscience-fiction writers to challengeconventionalnotions of humanidentity through a technological ‘other’—has nowfound its way into national and global discourseson the future of geopolitical and geo-economicformations. Nations have begun racing to leavetheir footprint on the digital foundations of thefuture. BOND’s ‘Trends in Artificial Intelligence’reportof May 2025 attempts to quantify theundulationsoftheongoingtechnologicalshift.1 The report charts the evolution of AI in recentdecades and attributes unprecedenteddevelopments in the field to breakthroughs inlarge language models (LLMs). This is exemplifiedby the launch of OpenAI’s ChatGPT in November2022, which was facilitated by an existing andaccessible global internet infrastructure and theproliferation of digital datasets over the last threedecades. The unfolding AI revolution is furtheracting as a “compounder” of growth in digitaluser engagement, developer ecosystem growth,and capital investment.2These factors primed theglobal economy for innovation, investment and theadoption of AI in the financial, social, geopolitical,and technical spheres in a much shorter time scalethan previous technological cycles. A crucial observation made in the report is thatglobal connectivity has been a key determinantof the pace of AI scaling. While previous digitalplatformshad to build their foundationalinfrastructure from the ground up, AI developersanddeployers have been able to leverageapproximately5.5 billion digitally connectedpeople.It also underscores the importance ofenablinginfrastructure as,“[the]document isfilled with user, usage and revenue charts thatgoup-and-to-the-right…oftensupportedbyspending charts that also go up-and-to-the-right.”3 The accelerating development and adoptionof AI products, services and platforms presentbothchallenges and opportunities for regionslike the Middle East and North Africa (MENA)and India that have ambitions of integrating AIinto their economies. Data presented in the reportsuggests that the mobile user bases in India andMENA are primed for AI products and serviceson mobile platforms. For the Middle East, AI is acrucial enabler of economic diversification beyondits hydrocarbon industries, whereas for India, AIcan be transformative for its world-leading digitalpublic infrastructure, public service delivery, anddigital payments platforms. Whilethe trends highlighted may seemunidirectional, the trajectory of growth metricsalso reveals an intricate and fast-evolving economiccontext. AI-focused entities, including both legacycompanies and startups, are facing a higher degreeofcash burn relative to their revenue.Risingpre-trainingcosts,falling inference costs,andperformance convergence across frontier modelsare favouring application programming interface(API) developers and deployers over foundationmodel developers. The global competitive arenais also intensifying with challenges to Big Tech byagile AI-native startups, the achievements of Chinain AI foundation models and AI-powered roboticsdespite years of United States (US) sanctions, andnoteworthy innovations in open-source models. Although the BOND report provides extensivedatasets to support an optimistic investmentmessagefor the AI sector,it pays insufficientattentionto longitudinal risk factors such asdisproportionately high valuations of AI companiesand the impact of AI on the labour market. Forinstance,the report highlights the discrepancybetween the high valuations of AI startups andtheir low revenues, but it does not expand on thecorrelations between the ongoing AI boom andthe dot-com bubble of the 1990s, revealing anoptimistic bias.4The report notes that AI adoptionis leading to a global “cognitive automation”, whichis particularly threatening to white-collar jobs.5 Whetherworkers can adapt to and worksymbiotically with AI will decide the future ofworkand displacement risk,according to thereport. It projects tha