Case 05Growth EconomicsSuccess

Amazon flywheel and AWS

Amazon.com, Inc. · 1997–2020

Should Amazon remain storage and API led, launch a bounded external-compute beta, or commit broadly to external infrastructure?

At the decision boundary

Amazon external infrastructure adjacency decision

Decision time
August 24, 2006
Knowledge cutoff
August 24, 2006
Recommended path
Buy information about the storage-to-compute adjacency through a capacity- and loss-capped beta, while preserving the ability to stop before broad fixed commitments.
Confidence
Moderate

What happened

Amazon launched EC2 in beta as external compute, later opened the beta to all developers and added instance types, and moved EC2 to general availability with a regional service-level commitment.

The cutoff recommendation favored a bounded compute beta because S3 and developer-interest signals justified learning while standalone economics and fixed-capacity requirements remained unknown. The observed beta-to-open-beta-to-general-availability sequence is directionally consistent with that process. Later segment success supports the plausibility of the platform mechanism but does not prove that the exact initial capacity, pricing, timing or investment envelope was optimal.

Case inventory

What is inside

13source records
6financial tables
23material claims
1candidate rules

Transfer with care

Rule hypotheses from this case

All rule hypotheses →
Candidatemoderate confidence

rule.stage-platform-adjacency-under-unknown-unit-economics

Launch a bounded beta with explicit cohort, contribution, reliability, support and capital gates; expand access and scope only as evidence clears those gates.

A limited launch converts an adjacency thesis into customer and operating evidence while preserving the option to stop before broad fixed commitments.

Read against

A contrasting case sharpens the boundary.