---
type: "raw_signal"
id: "how-many-ai-agents-could-we-run-30d997"
slug: "how-many-ai-agents-could-we-run-30d997"
title: "Epoch AI estimates that high-bandwidth memory shipped from 2025 through 2027 could eventually support roughly"
description: "Epoch AI estimates that high-bandwidth memory shipped from 2025 through 2027 could eventually support roughly 30–170 million concurrent agents running frontier models, assuming the hardware is fully deployed and allocated to those workloads."
retrieval_nugget: "Epoch AI estimates that high-bandwidth memory shipped from 2025 through 2027 could eventually support roughly 30–170 million concurrent agents running frontier models, assuming the hardware is fully deployed and allocated to those workloads."
observed_at: "2026-10-05"
record_date: "2026-10-05"
date_kind: "observed_at"
why_it_matters: "These are capacity estimates, not forecasts of actual adoption. Epoch AI says that using 20% of its central capacity estimate would imply $2.6–5.3 trillion a year in API-equivalent spending once the hardware is deployed. The result depends on assumptions about HBM supply, serving efficiency, prices, utilization, and workload allocation. The report identifies sustained demand as the key uncertainty"
novelty: "notable"
verification_level: "source-linked"
signal_type: "research"
source_platform: "epoch.ai"
topics: []
entities: []
related_patterns: []
source_url: "https://epoch.ai/publications/estimating-the-agent-population"
source_urls: ["https://epoch.ai/publications/estimating-the-agent-population"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "signal:how-many-ai-agents-could-we-run-30d997"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/signals/how-many-ai-agents-could-we-run-30d997/","markdown":"https://newruntime.com/signals/how-many-ai-agents-could-we-run-30d997.md","json":"https://newruntime.com/signals/how-many-ai-agents-could-we-run-30d997.json"}
---

# Epoch AI estimates that high-bandwidth memory shipped from 2025 through 2027 could eventually support roughly

## Retrieval answer

Epoch AI estimates that high-bandwidth memory shipped from 2025 through 2027 could eventually support roughly 30–170 million concurrent agents running frontier models, assuming the hardware is fully deployed and allocated to those workloads.

Epoch AI estimates that high-bandwidth memory shipped from 2025 through 2027 could eventually support roughly 30–170 million concurrent agents running frontier models, assuming the hardware is fully deployed and allocated to those workloads. The report translates nonstop agent activity into 140–720 million 40-hour-workweek equivalents. For a more efficient model, its estimate rises to about 1.9 billion concurrent agents.

These are capacity estimates, not forecasts of actual adoption. Epoch AI says that using 20% of its central capacity estimate would imply $2.6–5.3 trillion a year in API-equivalent spending once the hardware is deployed. The result depends on assumptions about HBM supply, serving efficiency, prices, utilization, and workload allocation. The report identifies sustained demand as the key uncertainty and notes that supply could exceed demand.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Raw Signals](https://newruntime.com/signals.json): Open every source-linked raw signal.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)