Amazon CEO Andy Jassy has said that Amazon Web Services' cloud computing capacity is largely sold out through 2027, with significant customer commitments already extending into 2028. The remarks, made in a video podcast with Andreessen Horowitz co-founder Ben Horowitz and general partner Erik Torenberg and reported by Reuters on October 23, 2026, capture the central tension of the AI infrastructure boom: demand is growing faster than even the largest cloud provider can build.
"2027 is largely sold out. We always leave a certain amount for on demand and our inference service Bedrock, or EC2, but largely sold out all the big chunks. A good chunk of 2028 is sold out," Jassy said, according to Reuters. EC2 is AWS's service providing cloud servers for running applications, while Bedrock gives customers access to AI models for building generative AI applications.
$220 billion — and still not enough
Earlier in 2026, Amazon said it would spend about $220 billion on capital expenditure for the year, raised from around $200 billion largely because of higher memory costs. Jassy said the company "could invest a lot more and it wouldn't be a bad use of capital," given that cloud-computing capacity for 2026 is already sold out.
The scale of the spending is difficult to grasp. In the second quarter of 2026, AWS generated about $42 billion in revenue, up roughly 36–37% year over year, with about $16 billion in operating income — an annualized run rate of about $169 billion, its fastest quarterly expansion in 18 quarters. In the same quarter, Amazon's cash capital expenditures totalled just over $53 billion, primarily related to AWS and generative AI. Spending and revenue cover different activities and time horizons — data centres are long-lived assets — but the comparison shows a company investing ahead of revenue at extraordinary scale.
Jassy has previously said that AWS contracted commitments jumped from $364 billion to $496 billion in a single quarter earlier in 2026. That backlog — revenue from signed contracts not yet recognised — is the financial mirror of the capacity shortage: customers are signing for compute Amazon has not yet built.
Why demand is outrunning supply
The drivers are familiar but worth stating precisely. AI laboratories training frontier models need enormous, concentrated compute. Enterprises deploying AI agents and generative applications need inference capacity that scales with usage. Both compete for the same scarce inputs: advanced chips, high-bandwidth memory, power, and data-centre sites with grid connections.
Memory costs are a particular bottleneck. Jassy attributed the $20 billion capex increase largely to higher memory prices, and memory is also the constraint behind new architectures like pooled and optical memory systems. Power is another: Amazon recently backed a 7.65 GW off-grid Texas gas plant, consistent with a company that believes it is genuinely power- and capacity-bound, not just chip-bound.
What "sold out" actually means
Jassy's phrasing needs care. "Largely sold out" does not mean no capacity is available. Amazon reserves capacity for on-demand use, EC2, and Bedrock inference. What is sold out is the large, long-term blocks that hyperscale customers commit to in advance. For a startup, that distinction matters: on-demand and inference capacity remains accessible, but negotiating a large reserved commitment on favourable terms is a different market than it was two years ago.
It is also worth noting that these booking claims are company framing from a podcast and earnings commentary, not audited backlog figures. Amazon is scheduled to report third-quarter results on October 29, 2026, which will give investors a chance to test the narrative against reported numbers.
What it means for founders and investors
For founders, the capacity shortage cuts both ways. On one hand, compute scarcity raises costs and lengthens procurement timelines for AI-heavy products. Architecture choices — efficient models, caching, batching, and inference optimisation — are now cost strategy, not engineering polish. On the other hand, scarcity creates opportunity for companies that help others use less compute, find spare capacity, or run workloads where capacity exists.
For investors, the AWS story is a case study in how AI demand is reshaping capital allocation. When the world's largest cloud provider says $220 billion is not enough, the investment question shifts from "is AI demand real?" to "where are the bottlenecks, and who owns them?" Memory, power, interconnects, and data-centre construction are all candidates — and each has its own cycle, lead time, and margin structure.
The deeper question is durability. Jassy has said he sees AWS becoming a $1 trillion business in time. Whether that happens depends on whether today's contracted demand converts into sustained usage revenue, or whether some of it reflects customers over-committing in a scarce market. The October 29 earnings report, and the quarters after it, will be the test.
Angeled.com connects founders with investors.
Sources: Reuters via https://wixx.com/2026/10/23/amazon-cloud-capacity-sold-out-through-2027-ceo-jassy-says/ , https://catenaa.com/industries/technology/amazon-says-aws-cloud-capacity-largely-booked-through-2027/ , https://www.latestly.com/technology/amazon-ceo-andy-jassy-says-aws-capacity-sold-out-through-2027-7655371.html
Read Next
OpenAI Releases Codex Security as Open-Source CLI and SDK
Databricks Unveils Always-On Genie, an AI Agent That Stops Waiting to Be Asked
DeepMind Launches AlphaGenome 2, a DNA Model That Reads Genomes Cell by Cell
ElevenLabs Unveils 13 New Products and Upgrades at Its New York Summit
Google Launches Gemini Decision 1, a Model That Answers Instead of Writing
Runway Open-Sources Praxis-1, Its Robot Model That Learned by Watching Video



