Amazon 2006: Launching AWS

Situation

It is early 2006. Amazon is a $10.7 billion revenue company — a retailer of books, electronics, and consumer goods. It is not a technology services company. It has no enterprise sales force. It has no pricing model for selling computing infrastructure. And it is about to launch Amazon Web Services — specifically Amazon S3 (storage) in March 2006 and Amazon EC2 (compute) in August 2006.

The origin story matters: in 2002-2003, Amazon engineers realized that every internal team was rebuilding the same foundational infrastructure — storage, databases, messaging, payment processing — from scratch. Jeff Bezos mandated that all teams expose their services via APIs and architect as if they were going to sell those services externally. This was a forcing function for operational excellence.

By 2005, Amazon's internal infrastructure was modular, scalable, and well-documented. Andy Jassy was assigned to figure out whether this capability had external commercial value.

The external market context:

1. Starting a tech company in 2004 requires massive upfront capital. A startup needs to buy servers, rent datacenter space, hire operations engineers, and build before they can ship. Most great ideas never get funded because the infrastructure barrier is too high.
2. Enterprise IT is locked into proprietary hardware/software stacks. IBM, HP, and Sun sell long-cycle infrastructure contracts at high margins. CIOs budget 3 years in advance.
3. Broadband and internet infrastructure are maturing. The same forces that are making Netflix's streaming viable are making hosted computing infrastructure viable.

Jassy's insight: charge by the hour for computing and by the gigabyte for storage. No contracts. No minimums. Infinite scale on demand. The price will be a fraction of what startups pay for equivalent capacity.

The decision moment

It is January 2006. Bezos has approved the AWS launch for Q1. Three decisions remain:

1. Pricing. S3 is priced at $0.15 per GB/month stored, $0.20 per GB for data transfer out. EC2 is priced at $0.10 per hour for a "small instance." These prices are set to be profitable at scale — but the margin will be thin for years. Every traditional enterprise vendor Bezos has studied charges 10-100x more. Is Amazon pricing to win market share, or leaving money on the table?
2. Who is the customer? Jassy wants to target startups — low friction, self-service, no sales force required. Enterprise IT buyers want contracts, SLAs, and account managers. Which customer do you build for first?
3. What does this do to Amazon's retail brand? Amazon is a consumer brand. Announcing that Amazon is selling cloud computing infrastructure will confuse consumers, confuse investors, and invite ridicule from enterprise IT analysts who will say "why would I trust my data center to a bookseller?"

You are Jeff Bezos.

Key financial datapoints (for reference)

Frameworks invoked

• Platform Strategy. AWS is not a product — it is a platform. Every startup that builds on AWS creates switching costs, generates usage data, and eventually brings enterprise customers who want to run alongside the startups they acquired. The platform value grows with each customer added.
• Resource-Based View. Amazon's core resource is operational infrastructure built at scale. The insight is that this resource has value outside Amazon's own retail operations. The competitive advantage is not the idea — it is the decade of operational experience that created the infrastructure.
• Two-Sided Markets. AWS creates a network of developers who build on it, and enterprises that want access to what those developers built. Each side makes the other more valuable.
• Economies of Scale. Cloud infrastructure is a business of massive fixed costs and near-zero marginal costs at scale. The company that reaches scale first can price everyone else out. This is a winner-take-most market — but it takes years for the scale advantages to be visible.

Discussion questions

1. Amazon's retail margins are ~3%. AWS will pressure those margins further for several years before it generates meaningful profit. How do you make the financial case to investors for an investment that will hurt short-term earnings?
2. Bezos chose to target startups first (self-service, low friction) rather than enterprise IT (high contract value, complex sales). Was this the right sequencing? What are the long-term consequences of this choice?
3. IBM and HP had vastly more enterprise relationships, sales forces, and credibility in corporate IT. Why couldn't they replicate what AWS built, even with 5-10 years of warning?
4. The "bookseller sells cloud computing" framing is a branding liability. But it also means no incumbent takes the threat seriously for 5+ years. Was the confusion a cost or an asset?
5. AWS's pricing at launch is $0.10/hr for compute. By 2023, comparable compute has dropped to ~$0.002/hr. How do you think about a business model where your price must decline 98% over 17 years to remain competitive, and your growth thesis depends on volume making up the difference?

The real outcome (revealed at session end)

2006-2008: AWS adoption is slow among enterprises but explosive among startups. Dropbox, Reddit, Airbnb, Netflix, and NASA are early customers.

2010: Netflix begins moving its entire infrastructure to AWS — a competitor of Amazon's Prime Video. AWS serves competitors without editorial discrimination. This policy is radical and counterintuitive but reinforces developer trust.

2013: AWS is profitable. Amazon begins disclosing AWS revenue separately from retail. Analysts realize for the first time that AWS is a higher-margin business than Amazon's retail operations.

2015: AWS is run-rating at $7.8B — the fastest-growing major enterprise software business in history. Microsoft and Google are competing but years behind.

2023: AWS revenue reaches $90.8 billion. AWS operating income: $22.8 billion — the vast majority of Amazon's total operating profit. Without AWS, Amazon's retail business is marginally profitable.

The lesson: The biggest strategic bets often look like distractions at the time. Bezos bet that an internal operational capability — reliable, scalable infrastructure — was worth selling externally before any market existed for it. The 17-year compounding of that insight created the most profitable business in Amazon's history.

Sources

• Brad Stone, The Everything Store (2013).
• Andy Jassy, various recorded interviews on AWS origins.
• Amazon 10-K filings 2006-2023.
• Ben Thompson, "AWS and the Frozen Middle" (Stratechery).
• HBS case: "Amazon Web Services 2006" and "Amazon in 2017" (2017).