AWS booms as Amazon lifts AI capex to $220B and launches Forward Deployed Engineering

Amazon's latest results reframed the AI-infrastructure arms race: AWS is running at roughly a $169B annualized run rate with a nearly $500B cloud backlog, and CEO Andy Jassy raised 2026 capital spending to $220B — while cautioning it still won't fully meet AI demand. Reuters reported the AWS growth reassured investors worried about ballooning capex, sending Amazon shares higher. Bedrock consumption in Q2 exceeded all prior quarters combined, a signal that enterprise agentic workloads are moving from pilots to production at scale.

The most novel move is organizational: Amazon will invest $1B to build an AWS Forward Deployed Engineering team, embedding AI engineers directly inside customer organizations — including the NBA, NFL, and Southwest — to ship agentic systems in days rather than months. It mirrors Palantir's forward-deployed model and reflects a bet that services and hands-on integration, not just raw compute, capture AI's enterprise value.

The scale of the $220B commitment underscores how the hyperscaler capex war is now measured in hundreds of billions, with AWS, Microsoft, and Google each racing to secure chips, power, and data-center capacity. AWS's advantage is its installed enterprise base and Bedrock's model-agnostic catalog; the risk is whether demand justifies spending of this magnitude if the agentic-AI ROI curve flattens.

On LinkedIn, AWS's Matt Wood highlighted the PGA Tour's agentic AI platform on Bedrock, and Swami Sivasubramanian announced an expanded AI-strategy role, both reinforcing AWS's customer-momentum narrative. The number to watch next quarter is whether Bedrock's record consumption keeps compounding — the metric AWS is leaning on to justify the historic capex.