Oct 23, 2026, 11:40 AM ETInterview, Fireside Chat

Amazon CEO Andy Jassy on the $220 Billion AI Buildout

$220 billion CapEx Spend: Amazon plans to spend $220 billion on capital expenditures this year (approximately $600 million daily), though CEO Andy Jassy notes this still leaves "a meaningful amount of demand on the table."
Capacity Sold-Out Forecast: AWS compute capacity is sold out for 2026 ("26 is obviously sold out"), largely sold out for 2027, with significant portions of 2028 already committed.
Revenue Growth Metrics:
- Overall top-line revenue is roughly 80% higher in 2024 compared to when Jassy became CEO in 2021.
- AWS revenue is nearly three times larger than in 2021.
- Advertising revenue is approximately 2.5 times larger than in 2021.
- AWS grew 36.7% year-over-year in the last quarter, reaching a $169 billion annual revenue run rate.
AI Revenue Explosion: The AI-specific segment within AWS reached a $25 billion run rate in just three years, growing several hundred times faster than AWS did in its first three years (which was a $60 million run rate).
Long-Term Projection: Jassy anticipates the AWS business could eventually reach "a few hundred billion dollars of revenue" over a 10-year period.

Pandemic Fulfillment Network Overbuild: To meet pandemic-era demand, Amazon doubled its US fulfillment network size in 18 months and built a last-mile network comparable to UPS in 20 months, initially leading to suboptimal cost structures.
Strategic Pivot to Regionalization: The company "regionalized a flat national network," reducing transportation distances to lower costs and increase speed, a move that "unleashed speed of delivery and a lot of demand."
Capex Cash Cycle Dynamics:
- Data center infrastructure requires ~2 years of lead time to monetize and has a 30-year useful life.
- Server and networking gear require only months of lead time, allowing for demand-driven purchasing.
- Heavy CapEx creates short-term free cash flow headwinds before facilities are monetized and break even (typically ~3 years).
Supply Chain Constraints: Jassy predicts supply constraints (chips, memory, power, cooling) will persist beyond 2027 due to the unprecedented scale of AI demand compared to historical infrastructure builds.

Political and Regulatory Hurdles: Jassy characterizes US politics as uniquely "anti-data center," noting that building the necessary infrastructure is comparable to the 1950s highway and bridge builds in economic importance.
Water Usage Reality Check: Jassy refutes environmental concerns by noting the golf industry consumes 200 times more water than Amazon's data centers; one golf course uses more water in two days than Amazon's centers use in a year.
Grid Investment Catalyst: The surge in AI demand is the primary driver forcing investment in US power grids, which Jassy argues the country "has not invested in... the way we've needed."
Community Investment: Amazon paid approximately $3 billion in taxes in St. Joseph County, Indiana, and committed an additional $1 billion to community programs like "Built Together" over the next five years.
Customer Allocation Policy: Despite "big labs" willing to pay almost anything for power and chips, Amazon refuses to displace core enterprise customers; the strategy is to "have as much capacity as we can" rather than squeezing existing customers to favor new, high-paying AI demand.

Annapurna Labs Acquisition: Bought to vertically integrate and reduce reliance on external vendors like Intel, originally to offload virtualization and networking from EC2 servers.
Graviton Processors: Built by the Annapurna team, these CPUs offer 40% better price performance than x86 processors and are used by 99% of Amazon's top 1,000 EC2 customers.
Trainium AI Accelerators: A custom AI chip developed to compete with dominant providers, offering 30% to 40% better price performance to lower the cost of AI inference.
Inference Cost Reduction: Jassy identifies the high cost of inference as the primary barrier preventing pervasive enterprise adoption of AI; custom chips are central to lowering this cost by an order of magnitude.

Strategic Rationale: Investments ($18 billion in Anthropic, ~$50 billion in OpenAI) were driven by the teams' quality, the transformative potential of the technology, and the belief that these companies will be market leaders long-term.
Technical Integration:
- Anthropic: Deep partnership involving running models on Bedrock and training on Amazon's Tranium chips.
- OpenAI: Partnership enabled by a new "stateful runtime environment" allowing customers to run OpenAI models on AWS, bypassing prior exclusivity deals.
Risk Mitigation: Jassy acknowledges the market likened these moves to the dot-com bubble but argues the speed of growth and strategic necessity of partnering with top-tier labs justify the capital deployment.

Team Size Reduction: Jassy predicts organizations will have "fewer humans doing the work," with smaller, cross-functional teams where AI agents handle 80% of coding, customer service, and research tasks.
Reinvention vs. Optimization: Amazon is reimagining legacy products (Alexa, Prime Video, Fire TV) from scratch rather than incrementally improving them, believing that "AI-native" reinvention is required to remain competitive.
Alexa Transformation: The "brains" of Alexa are being rewired to leverage generative AI, a difficult internal transition requiring teams to abandon past successes for new capabilities.
Security Imperative: Immediate security risks involve AI finding vulnerabilities faster than humans; applications must be rewritten to be "agentically capable" to deploy security patches in minutes rather than weeks.
Code Rewrite Necessity: Jassy predicts companies will need to "rewrite almost all of their code to be agentically capable" to allow agents to update systems autonomously.

Triple Transformation Era: Jassy defines the current moment as the convergence of AI, robotics, and space development, all of which are reshaping customer experiences.
Education Strategy: For young people and universities, Jassy advises against banning AI in education; instead, students must learn to use it while retaining "critical reasoning and thinking" skills, as AI cannot yet effectively synthesize strategy or invention.
Diversification Strategy: Amazon maintains a "decentralized single-threaded" organization to pursue diverse bets (satellites, healthcare, robotics, autonomous driving) simultaneously without spreading leadership too thin, relying on specific leaders for each vertical.
Sustainability Goal: The core mission unifying all divisions remains "trying to make customers' lives better and easier every day," balancing customer success with long-term sustainable business growth.