Anthropic's $100B AWS Bill Precedes Its Late-September IPO

Anthropic's decision to sign a $100 billion cloud contract before its IPO marks a new benchmark for growth-stage AI startups. The prospectus, expected after Labor Day, gives founders and investors a template for evaluating how compute-heavy AI businesses scale toward public markets.

Anthropic's forthcoming IPO prospectus could answer a question that Amazon's public filings cannot: whether the company behind the Claude AI models has the financial wherewithal to support a decade-long cloud computing commitment exceeding $100 billion. The Information reported in late August that Anthropic plans to publish its prospectus after Labor Day, with a listing possible as soon as late September or October. For Amazon shareholders, that document is unusually relevant because AWS already carries an estimated $496 billion backlog as of June 30, 2026, up from $364 billion in March and just $195 billion in mid-2025. The Anthropic deal, signed April 20, 2026, alone represents roughly one-fifth of that total and covers up to 5 gigawatts of capacity on Amazon's own silicon, including Graviton processors and Trainium2 through Trainium4 AI chips.

“

For Amazon shareholders, that document is unusually relevant because AWS already carries an estimated $496 billion backlog as of June 30, 2026, up from $364 billion in March and just $195 billion in mid-2025.

The scale of AWS's contracted future is striking. The segment's backlog grew 154% year over year, including a $132 billion jump in a single quarter. Anthropic was not the only driver: Amazon also disclosed a $100 billion, eight-year expansion of its existing $38 billion OpenAI commitment, announced a quarter earlier. Together these deals show that hyperscale AI developers are signing cloud contracts far longer and larger than the industry has historically seen. The weighted-average remaining life of AWS's long-term contracts stretched from 4.0 years to 6.4 years over 12 months, meaning more of the backlog now sits further away from revenue recognition.

For Amazon, these commitments provide durable revenue visibility, but they also introduce credit and concentration risk. A single customer's promise to spend more than $100 billion is only as good as that customer's liquidity and execution. If Anthropic's growth stalls or its unit economics deteriorate, AWS could find itself with capacity built for a contract that may be renegotiated or restructured. The prospectus will show Anthropic's cash reserves, revenue, cost structure and compute spending, allowing Amazon investors to assess the quality of the backlog rather than simply its size. That matters because AWS's backlog is one of the main ways investors now value Amazon's cloud segment.

For Anthropic, the IPO narrative will be defined by its ability to fund frontier AI model training. Committing to $100 billion over ten years implies average annual AWS spend of roughly $10 billion, an enormous figure relative to most peers' revenues. If the prospectus shows high gross margins and improving operating leverage, the market may accept compute-heavy economics as a path to dominance. If it shows heavy losses and reliance on ongoing capital raises, the commitment could become a value-destroying anchor. The disclosed option on future chip generations also suggests Anthropic is betting a portion of its roadmap on Amazon silicon rather than solely on Nvidia GPUs, a strategic shift with long-term cost and performance implications.