What Exactly Is the 'Flywheel'?
Imagine a massive, heavy wheel that’s hard to get moving. But with each push, it gains momentum until it spins on its own, creating more energy than it consumes. That’s the Amazon flywheel. Famously sketched on a napkin, the idea is a self-reinforcing
loop. It starts with low prices and a great customer experience. This attracts more customers, which in turn attracts more third-party sellers who want access to that audience. More sellers mean a wider selection of products, which improves the customer experience even more, driving further growth. This ever-increasing scale allows Amazon to lower its cost structure, which funds even lower prices. It’s a cycle of perpetual motion that powered Amazon from an online bookstore into a global behemoth.
The Bezos Era of Perpetual Motion
Jeff Bezos was the master of feeding the flywheel. He showed a legendary willingness to sacrifice short-term profits for long-term dominance, reinvesting every bit of energy back into making the wheel spin faster. Prime membership, with its promise of fast, 'free' shipping, was a huge push. The expansion into a third-party marketplace was another. But the most powerful push of all was Amazon Web Services (AWS). Originally built to serve Amazon's own infrastructure needs, AWS became a wildly profitable engine in its own right. The cash it generated was poured back into the retail business, subsidizing low prices and fast delivery, which crushed less-diversified competitors. Every part of the business was designed to reinforce the others.
Enter Jassy and the AI Mandate
Andy Jassy, who previously ran AWS, inherited the keys to this finely tuned machine. His challenge isn’t to build a new flywheel, but to find a new, sustainable fuel for it. That fuel is artificial intelligence. The problem is that AI is a fundamentally different economic proposition. The Bezos flywheel was built on tangible, measurable inputs: logistics, inventory, and server space. The path from investment to customer value was clear. AI, on the other hand, is built on massive, speculative, upfront investment. The capital required for computing power, data centers, and research is staggering, with CEO Andy Jassy confirming plans for around $220 billion in capital expenditures for 2026, with meaningfully more expected in 2027.
The High Cost of a Smarter Flywheel
Jassy is placing enormous bets, most notably with a potential investment of up to $25 billion or more in the AI startup Anthropic. In return, Anthropic has committed to spending over $100 billion on AWS cloud services over the next decade. This circular investment structure, where Amazon funds a company that in turn becomes a massive customer, highlights the strange economics of the AI race. While AWS is seeing incredible growth, with its AI business passing a $25 billion annual revenue run rate, the costs are immense. Amazon's free cash flow has turned negative, driven by huge increases in equipment purchases for AI. Jassy himself has noted that the cost of running AI models is still about ten times too high for widespread adoption, which is why Amazon is developing its own custom chips like Trainium to bring costs down. The central question remains: will these colossal AI investments accelerate the flywheel by creating new services customers will pay for, or will they become a capital-draining vortex that slows it down?
