[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86441-en":3,"doc-seo-86441-105":30,"detail-sidebar-cat-0-en-105":84},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86441,4398048950312,"Violet","https://ap-avatar.wpscdn.com/avatar/400002538284de19e3c?_k=1778320343897328908",8,"Research & Report","Quota Marketplace Dynamic Pricing for Efficient Allocation of ML Training Resources","Rising demand for machine learning (ML) training resources has widened the gap between scarce, expensive accelerators and available supply, making ROI-maximizing allocation essential. Existing approaches can ensure Pareto efficiency and max-min fairness under time-varying demands, yet they break down when workloads carry heterogeneous values. Quota Marketplace introduces a market-based mechanism that prices resources dynamically and lets teams express workload value, enabling efficient allocation aligned with organizational priorities. The work reports Google deployment, metrics, and theoretical guarantees.","Quota Marketplace: Dynamic Pricing for Efficient Allocation of ML Training  \nResources  \nBalasubramanian Sivan∗, 1 Renato Paes Leme∗, 1 Mihai Tiuca∗, 1 Ian McFarlane 1 Vasilis Gkatzelis 1 ,2 Nehal Mehta 1 Soheil Hassas Yeganeh 1 Vahab Mirrokni 1 Amin Vahdat 1 ( 1 Google, 2Drexel University)  \narXiv :2607 .09802v 1 [ cs .LG] 9 Jul 2026  \nAbstract  \nThe escalating demand for Machine Learning (ML) training resources in recent years has resulted in a substantial gap between the high demand and the available supply. Efficient allocation of these scarce and expensive resources is crucial for organizations to maximize their return on investment. Existing resource allocation mechanisms, like Karma [26], are designed to guarantee Pareto efficiency and max-min fairness in settings with dynamic (time-varying) user demands, but fail to preserve these key properties in the presence of demands with heterogeneous values. Given the ubiquity and inevitability of heterogeneity in organizational values of different workloads, effective resource allocation policies must accommodate these variations.  \nIn this paper, we describe the design, implementation, deployment, and theoretical analysis of Quota Marketplace, a market-based mechanism to efficiently allocate ML training chips (like GPUs), explicitly addressing scenarios with demands of heterogeneous value. We detail the implementation of this mechanism within Google and present metrics that demonstrate its impact. We also discuss many businesscritical requirements that the Quota Marketplace handles quite effectively, and document the gains and opportunities it has unlocked. We establish theoretically how this market-based approach achieves the essential properties of Pareto efficiency and max-min fairness by allowing the users to express the value of their workloads and enabling dynamic resource pricing based on supply and demand fluctuations. Ultimately, the market facilitates resource allocation that aligns with organizational priorities.  \n1 Introduction  \nThe demand for Artificial Intelligence (AI) compute has experienced unprecedented growth, increasing by over tenfold annually for eight consecutive years, resulting in a 100 million  \nfold increase over that 8-year period [1] . Major technology *Equal contribution.  \ncorporations have publicly disclosed substantial capital expenditure (CapEx) budgets, tens of billions of dollars each for 2025, to address this growing demand. This exponential growth in demand creates a widening gap between the demand and supply of AI compute resources, specifically Machine Learning (ML) accelerators like GPUs. Consequently, due to the high cost and limited availability of these resources, organizations must implement return-on-investment (ROI) maximizing, efficient allocation strategies for ML accelerators among individual teams and products. In this paper, we describe Quota Marketplace (QM)– a market-based system designed, implemented, theoretically analyzed, and fully deployed at Google, to allocate ML chips to teams across business units. The deployed system allocates several hundreds of thousands of ML accelerators for ML training workloads across all the business units in Google. The footprint of the Quota Marketplace is a double-digit percentage of the entire ML fleet of Google.  \nWe begin with an overview of QM’s position in the ML scheduling stack and the notion of pools. A detailed description of QM concepts and implementation follow in Section 2.  \nQM and the ML Training Scheduling Stack in a Nutshell.  \nML compute resources (ML accelerators or ML chips), much like non-ML compute resources (CPUs), have traditionally been partitioned into static pools. Each pool corresponds to a business unit (BU) with distinct goals, cost accounting, and resources, and the pool administrators/planners explicitly assign accelerators to teams within that BU. The demand assessment and prioritization procedure typically takes place quarterly or semi-annually and it d","cbCaik3tdDdrPMzz","https://ap.wps.com/l/cbCaik3tdDdrPMzz","pdf",515776,5,1,18,"English","en",105,"# Abstract\n# Introduction\n## ML and the training scheduling stack\n## Static vs dynamic pools\n## Quota Marketplace design overview","[{\"question\":\"What fairness and efficiency properties does the paper claim Quota Marketplace achieves?\",\"answer\":\"The paper establishes that its market-based approach attains Pareto efficiency and max-min fairness by combining value expression with dynamic, supply-demand-aware pricing.\"}]",1784211760,45,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":79,"head_meta":81,"extra_data":83,"updated_unix":28},"quota-marketplace-dynamic-pricing-for-efficient-allocation-of-ml-training-resources","",{"@graph":36,"@context":78},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/quota-marketplace-dynamic-pricing-for-efficient-allocation-of-ml-training-resources/86441/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-27","2026-07-16",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72],{"name":73,"@type":74,"acceptedAnswer":75},"What fairness and efficiency properties does the paper claim Quota Marketplace achieves?","Question",{"text":76,"@type":77},"The paper establishes that its market-based approach attains Pareto efficiency and max-min fairness by combining value expression with dynamic, supply-demand-aware pricing.","Answer","https://schema.org",{"og:url":52,"og:type":80,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":82,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":85},[86,90,94,98,102,107,112,115,120,123,127],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":87,"show_sort_weight":88,"slug":89},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":91,"show_sort_weight":92,"slug":93},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Exam",70,"exam",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Comic",60,"comic",{"id":103,"doc_module":4,"doc_module_name":46,"category_name":104,"show_sort_weight":105,"slug":106},6,"Technology",50,"technology",{"id":108,"doc_module":4,"doc_module_name":46,"category_name":109,"show_sort_weight":110,"slug":111},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":113,"slug":114},30,"research-report",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":46,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":20,"slug":130},19,"General","general"]