Claude Platform on AWS is two offerings sold through one contract. The compute was already there. The story is the enterprise vendor-onboarding tax Anthropic just stopped paying.

Anthropic launched Claude Platform on AWS this week. The Hacker News thread hit 204 points before the announcement was a day old, and most of the substantive comments did not talk about model quality, latency, or pricing. They talked about procurement.

"For any reasonably sized org, setting up new contracts with new vendors involves a lot of procurement, lawyers, negotiations, etc," one commenter wrote. "If a team can just click a button in AWS, there's no issue." Another was more direct: "I think the idea is that you can launder your team or product AI spend through your AWS account. This matters in Enterprise."

That is the announcement. The technology has been available for two years. What changed is the contract.

Anthropic now offers two distinct ways for AWS customers to use Claude. Claude Platform on AWS is the full Anthropic product, operated by Anthropic, with data processed outside the AWS boundary. Claude on Amazon Bedrock keeps data inside AWS infrastructure for customers with strict residency requirements. Both are billed through AWS Marketplace.

Managed Agents, Anthropic's hosted agent infrastructure, is included in the platform offering. So is feature parity with the direct Anthropic Claude Platform: file processing, the Files API, the long-context handling that powers Claude Code and the Sonnet 4.6 long-running workloads.

The technical substrate has not moved. Bedrock has hosted Claude since 2023. The new piece is the contractual wrapper around it. Anthropic now appears in AWS Marketplace as a first-class line item, with spend flowing through AWS Enterprise Discount Program commitments rather than direct Anthropic contracts.

Enterprise vendor onboarding is the quiet tax on every frontier AI deployment. Legal review, security questionnaires, data processing agreements, vendor risk assessments, finance setup, ongoing audit. Even for a company that wants Claude tomorrow, three to six months is a realistic baseline before the first production call lands. For regulated industries the timeline runs longer.

Cloud marketplaces collapse that timeline because the procurement work is already done. The AWS contract is signed. The DPA is in place. Security review has been completed at the cloud-provider tier. Adding Claude under the AWS Marketplace umbrella becomes a budget allocation decision, not a vendor onboarding decision. One Hacker News commenter described being unable to get Claude Platform access through internal review despite having a business case, while Bedrock access was already approved. That gap is the entire commercial logic of this announcement.

The cost to enterprises is real and rarely reported. G2's Enterprise AI Agents Report found that 57 percent of companies already have AI agents in production. Nearly 95 percent of IT leaders cite integration as a hurdle. The technology works. The buying process is where most of the delay lives.

AWS Marketplace has a structural limit. The widely-cited figure is that AWS Enterprise Discount Program customers can route up to 25 percent of their committed cloud spend through marketplace purchases. That cap is the only ceiling on how much enterprise AI spend AWS can absorb on Anthropic's behalf. It is also why Anthropic will need every other cloud channel it can sign.

Two SKUs through one channel is the part of the announcement to read twice.

The Platform option gives full feature parity and ships Anthropic's roadmap as it lands. The Bedrock option gives in-boundary data handling for customers whose compliance posture cannot accept data leaving AWS. Anthropic is not asking enterprises to choose between speed of feature access and compliance. It is selling both at the same shelf, priced through the same contract, with the customer self-selecting based on its own risk appetite.

That structure does two things at once. It widens the addressable market without compromising the flagship product. And it forces no one inside the enterprise to make a vendor-selection decision: the question is now which SKU, not which vendor. Anthropic keeps the relationship. AWS keeps the channel. The customer keeps the contract simplicity.

The Claude on AWS announcement does not sit alone. It is the second half of a sentence Anthropic started writing four weeks ago.

The first half was supply. As we covered in Anthropic's $40 billion Project Deal with Google, Anthropic has been stacking compute commitments across every hyperscaler that can deliver gigawatts on a credible timeline. The total contracted capacity now runs north of 11 gigawatts across Amazon, Google, Microsoft, Broadcom, NVIDIA, Fluidstack, and SpaceX. The compute crunch we flagged in April has not eased. Anthropic is buying its way through it.

The second half is distribution. Yesterday's AWS-Stripe-Coinbase deal put AWS at the center of agent-to-agent payments. Today's announcement puts Anthropic inside the same procurement layer that lets AWS sell to most of the Fortune 500. AWS is becoming the enterprise spine for agentic infrastructure: payments, model access, marketplace billing, and the contractual relationships that make all of it auditable.

For Anthropic, the calculation is clean. Direct sales are slow, expensive, and unevenly distributed. Marketplace sales are fast, channel-funded, and ride the buyer's existing budget. Even at AWS's marketplace fee structure, the unit economics favor breadth over margin for a frontier lab racing to recover capex.

Three near-term changes for enterprises buying Claude or evaluating model providers.

Procurement friction drops. Teams that have been waiting on internal approval to add Anthropic as a vendor can now route the same spend through AWS Marketplace with substantially less governance overhead. Expect the Claude enterprise adoption curve, which SaaStr reported at 128 percent year-on-year growth, to steepen further. The bottleneck was the contract, not the model.

Channel parity becomes a real consideration. Buyers running multi-model strategies should expect equivalent procurement parity from competing labs. OpenAI's Azure relationship already provides much of this through Microsoft commitments. Google's Vertex AI offers the same logic for Gemini. The labs that lack a hyperscaler-grade distribution channel will feel that gap acutely in the next four quarters.

Pricing is going to look different. Marketplace pricing follows different mechanics than direct contracts. Discount structures are subject to AWS's commitment math, EDP terms, and partner spread. Enterprise procurement teams should not assume marketplace prices match direct-negotiated rates either up or down. The first wave of large Claude-on-AWS contracts will set the public reference points, and they will not be on the Anthropic website.

The deeper pattern matters more than the immediate pricing. The dependencies underneath agentic AI are stacking up at every layer: compute supply, model access, payments rails, procurement contracts. State of the Stack tracked $690 billion in hyperscaler capex committed over the next 24 months. The Anthropic-AWS deal is one example of how that capex starts producing recurring revenue: through the contractual relationships the cloud providers already own.

At what point does "buying AI through your cloud provider" become indistinguishable from "buying AI from your cloud provider"?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.