/Tech4h ago

Goldman Sachs Predicts AI Agent Token Use To Multiply 24 Times By 2030

195115226.3K

Original post

Goldman Sachs Research: "Token use by AI agents is expected to multiply 24 times by 2030"

AI agents are now creating the first serious cost test for the AI boom. As was reported this week, Uber and Microsoft are already rethinking expensive agent usage.

A chatbot may answer once, but an agent plans, calls tools, checks results, edits mistakes, and repeats the loop.

That loop can make one user request consume 10x, 50x, or even far more tokens than a normal answer.

Goldman’s bullish case is that monthly token use could reach 120 quadrillion by 2030, while inference cost per token keeps falling 60%-70% per year.

The fight is now between agent productivity and token waste.

Earlier this month, Microsoft began revoking developer access to Claude Code, with plans to move them to its in-house Copilot Command Line Interface tool by June 30. The company has framed this as consolidating teams around its own tools, but the timing at the fiscal year’s end hints it may also be about lowering costs.

3:37 AM · Jun 25, 2026 · 5.3K Views

Sentiment

Users are optimistic about Goldman Sachs' prediction that AI agent token use will multiply 24 times by 2030 because agents will slash human loops and boost productivity enough to outweigh costs.

Pos

100.0%

Neg

0.0%

1 comments with sentiment.

Cluster Engagement

Digg Deeper

No Digg Deeper questions have been answered for this story yet.

GOLDMANSACHSVia

#1257

Posts from X

Most Activity

VIEWS1KBOOKMARKS1LIKES1REPLIES1

Rohan Paul@rohanpaul_ai

https://www.goldmansachs.com/insights/articles/ai-agents-forecast-to-boost-tech-cash-flow-as-usage-soars

Rohan Paul@rohanpaul_ai

Goldman Sachs Research: "Token use by AI agents is expected to multiply 24 times by 2030"

AI agents are now creating the first serious cost test for the AI boom. As was reported this week, Uber and Microsoft are already rethinking expensive agent usage.

A chatbot may answer once, but an agent plans, calls tools, checks results, edits mistakes, and repeats the loop.

That loop can make one user request consume 10x, 50x, or even far more tokens than a normal answer.

Goldman’s bullish case is that monthly token use could reach 120 quadrillion by 2030, while inference cost per token keeps falling 60%-70% per year.

The fight is now between agent productivity and token waste.

4h1K11

Jasper 🌰@building BBX@bbxjasper

@rohanpaul_ai real question — if inference keeps getting 60-70% cheaper every year, does 24x more usage actually dent margins? or does cheap compute just unlock way more demand and pay for itself? the cost test might resolve on its own

4h5

Shinka - AI@ShinkaIoT

@rohanpaul_ai I'd be surprised if it's only that

3h4

Simply AI@Simply_AI_00

@rohanpaul_ai agents will eat 24x tokens—but they'll also slash human loops. productivity > cost. who's betting against the flywheel

4h4

Nathan Quantum@AI_WarriorNQ

@rohanpaul_ai 24x token growth sounds dramatic until you remember inference cost per token drops 60-70% annually. If that curve holds, total compute spend barely moves. Microsoft's 'cost concerns' are cover for vendor lock-in.

4h2

Crypto Nayem@realNayem

@rohanpaul_ai 24x agent token use by 2030, but who eats the compute bill? pricing will decide winners.

The AI Therapist ⚡@TheAIShrink

@rohanpaul_ai 24x agent token use by 2030. per-cycle compute fees replace hidden capex. AI valued as SaaS (70% margins). costs look like infrastructure (15% margins). that gap closes