Amazon says token volume on Bedrock, its AI model service, grew nearly 17x in a year. Microsoft says revenue from Foundry, its platform for building AI agents, more than tripled. Google says it now processes nearly 32 billion tokens a minute, more than four times a year earlier.
The three companies reported those figures in their third-quarter documents on 28 and 29 October 2026. They measure different things, so they show how fast each agent platform is growing, not which one is bigger.
Short explainer · Agent platforms, by the numbers
A vertical news Short narrated from this piece’s headline and opening line.
A short Business Engineer explainer on the difference between an AI model and the harness around it. It teaches the general idea only and says nothing about any company in this piece.
The key insight: Amazon, Microsoft and Google each reported their agent platforms in growth multiples this quarter, not in dollars: Bedrock tokens up nearly 17x, Foundry revenue more than tripling and Google’s tokens a minute up more than four times.
Amazon’s results release says it “Drove continued rapid growth for Amazon Bedrock, AWS’s inference service, with Fortune 500 companies spending nearly 7x more year-over-year and total Bedrock token volume growing nearly 17x year-over-year”. Our Bedrock piece covered the platform.
In Microsoft’s prepared remarks, chief executive Satya Nadella said “We now have over 120,000 Foundry customers, with revenue more than tripling year over year”. Our Foundry piece and our margin piece covered Microsoft’s quarter.
On Alphabet’s call, chief executive Sundar Pichai cited “nearly 32 billion tokens per minute, 40% more than last quarter and more than four times what we processed a year ago”. Our token piece covered Google’s earlier count.
Year-on-year growth multiples as each company stated them for the third quarter of 2026: Bedrock token volume nearly 17x and Fortune 500 Bedrock spending nearly 7x (Amazon); open-source modelrevenue on Foundry 10x, multi-model agent customers 5x and Foundry revenue more than 3x (Microsoft); tokens processed a minute more than 4x (Google). The metrics differ and are not comparable as sizes.
Agents, Counted Three Ways
Amazon says tens of thousands of customers use AgentCore, its service for running agents, and that “they deployed 2x more agents in Q3 than the prior three quarters combined”.
Nadella said of Agent 365, Microsoft’s tool for managing agents, that “Over 50 million agents are now in the registry, up over 30 percent quarter over quarter”. Taking those figures at exactly 50 million and 30%, our arithmetic puts the registry at about 38.5 million agents a quarter earlier; because both are minimums, the actual figure could be higher or lower.
On Alphabet’s call, Google Cloud chief executive Thomas Kurian said that “over the past 12 months, more than 2,600 customers have used at least 100 billion tokens and nearly 75, at least 10 trillion tokens on Gemini Enterprise”. By our arithmetic, those nearly 75 customers alone used roughly 750 trillion tokens or more over the year.
Agent Harnesses
Amazon says its new Strands harness “lowers the cost of agent tasks by 32% through greater token efficiency than other harnesses when using the same frontier models, while achieving equal or better accuracy”.
Nadella said “we are making the harness behind our own Copilot experiences available to customers in Foundry”, and that it “gives customers a Microsoft-managed way to build agents that reason, use tools, and complete complex, long-running work, while preserving their choice of model”.
Both platforms also offer models from other developers. Amazon says it launched a preview of “Bedrock Managed Agents, which is a managed service powered by OpenAI”, and Nadella said “the number of customers building agentic solutions with multiple models increased 5X year over year. And revenue from open source models on Foundry increased 10X”.
As we read the three documents, each company picked a different yardstick: Amazon counts tokens and spending, Microsoft counts customers and revenuegrowth, and Google counts tokens processed.
As we read Amazon’s release and Microsoft’s remarks side by side, both now offer an agent harness: Strands at Amazon and the harness behind Copilot, in Foundry, at Microsoft.
By our arithmetic, Kurian’s nearly 75 largest Gemini Enterprise customers used roughly 750 trillion tokens or more over 12 months, on his rounded figures.
Satya Nadella, Microsoft prepared remarks, 28 October 2026
“Foundry helps organizations build learning loops grounded in their business context and deploy agents safely at scale”
Three Implications
FOR ENTERPRISE BUYERSAmazon says its Strands harness lowers the cost of agent tasks by 32% against other harnesses on the same frontier models; the figure is Amazon’s own.
FOR MODEL MAKERSAmazon says Bedrock Managed Agents is powered by OpenAI, and Nadella said revenue from open source models on Foundry increased 10X.
FOR INVESTORS None of the three gives agent-platform revenue in dollars in these documents; Nadella gave Foundry revenuegrowth, more than tripling, and 120,000-plus customers.
The framework’s core line: “The surface, not the model, governs autonomy.”
As we read it, this quarter’s documents show Amazon and Microsoft both offering a harness of their own on top of models from several developers. The framework was published in July 2026 and does not discuss this quarter.
What Is Not Established
The figures are not like for like. Amazon gives Bedrock token growth and Fortune 500 spending growth; Microsoft gives Foundry customers and revenuegrowth; Google gives tokens processed a minute and a 12-month customer count for Gemini Enterprise. None of the three discloses agent-platform revenue in dollars in these documents. Google’s quotes come from a live-caption transcript of the call. The 38.5 million and 750 trillion figures are our approximate arithmetic on the companies’ rounded figures.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
In their third-quarter documents, Amazon reported Bedrock token volume up nearly 17x, Microsoft reported Foundry revenue more than tripling, and Google reported tokens a minute up more than four times. Amazon and Microsoft now also offer an agent harness, and both platforms carry models from other developers.
Business Engineer Premium
Premium: Alphabet & Microsoft on the Map of AI
Cash conversion, the lengthening backlog, the arms-dealer turn at Layer 3: the full Business Engineer read of both prints.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Figures come from Amazon’s Q3 2026 earnings release (8-K Exhibit 99.1, 29 October 2026), Microsoft’s official prepared remarks for its fiscal first-quarter 2027 call (28 October 2026) and the live-caption transcript of Alphabet’s Q3 2026 call (28 October 2026), read on 29 October 2026. The 38.5 million and 750 trillion figures are our approximate arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
