The Thought Experiment
Last week I published a piece on Amazon’s vertical sovereignty play in logistics. The thesis was simple: Amazon is building a stack so complete that once you’re in, leaving becomes structurally expensive. Several people reached out and asked the obvious next question — so what would it actually look like to go all-in?
Fair question. I run my own infrastructure on Google Cloud. I chose that path deliberately, and I’ve written about why. But intellectual honesty requires stress-testing your own decisions. So here’s the exercise: take a real-shaped business and rebuild it entirely on Amazon’s stack. Not as a hypothetical. As a genuine evaluation of where Amazon is genuinely impressive and where the walls start closing in.
Meet Ridgeline Services
To make this concrete, let’s build a company. Ridgeline Services is a 22-person regional facilities management company operating across three metro areas. They handle commercial building maintenance — HVAC, plumbing, electrical, janitorial coordination — for property management firms. They have a small warehouse for equipment and supplies, a fleet of service vehicles, and a growing need for both cloud infrastructure and physical logistics.
Ridgeline is the kind of mid-market services company that exists in every region of the country. They’re past startup chaos but not yet at enterprise scale. They have real operational complexity — scheduling, procurement, fleet management, customer communication, compliance documentation — and they’re growing fast enough that their current patchwork of tools is starting to crack.
The question: what happens if Ridgeline rebuilds everything on Amazon?
Layer 1: Cloud Infrastructure (AWS)
This is where Amazon’s case is strongest, and it’s not particularly close.
AWS remains the largest cloud provider by market share. For Ridgeline, the relevant services are straightforward: EC2 or ECS for hosting their job management platform, RDS for their PostgreSQL database, S3 for document storage (inspection reports, photos, compliance records), and CloudFront for their customer-facing portal.
The honest assessment: AWS is excellent here. The breadth of services is unmatched. If Ridgeline’s CTO wants managed Kubernetes, it’s there. If they need a simple managed database, it’s there. If they want serverless functions for automated notifications, Lambda handles it cleanly.
Where it gets interesting is AI. Amazon Bedrock gives Ridgeline access to foundation models from Anthropic, Meta, Mistral, and Amazon’s own Nova family through a single API. They could build an AI assistant that reads inspection reports, flags compliance issues, and drafts customer communications — all within their existing AWS environment. Bedrock’s Intelligent Prompt Routing can reduce costs by routing simpler queries to cheaper models automatically.
Verdict: Genuine strength. AWS for compute and AI infrastructure is a defensible choice for a company like Ridgeline. The lock-in exists at the service level (good luck migrating a complex Lambda architecture to another cloud), but the value proposition is real.
Layer 2: Procurement (Amazon Business)
Here’s where the stack starts getting interesting. Ridgeline buys a lot of stuff — HVAC filters, plumbing fittings, electrical components, cleaning supplies, safety equipment, uniforms. Their current process is probably a mess of distributor accounts, local hardware store runs, and someone’s personal Amazon account with a company card.
Amazon Business replaces all of that with a single procurement platform. Approval workflows so the warehouse manager can’t order without the ops director signing off on purchases above a threshold. Integration with accounting systems through connections to platforms like Coupa and SAP Ariba. Business Prime for free two-day shipping on eligible items. Guided Buying to surface preferred suppliers and products that meet organizational standards. Spend Visibility dashboards that show exactly where money is going across all three metro locations.
For a 22-person company managing multiple locations, this is genuinely useful. The approval workflows alone solve a real problem — Ridgeline’s ops director currently has no visibility into what each location is ordering until the credit card statement arrives.
Verdict: Genuinely useful, with a catch. Amazon Business solves real procurement pain for mid-market companies. The catch is that once your approval workflows, supplier preferences, and spend history live inside Amazon’s system, switching costs are high. Not because of a contract — because of accumulated organizational knowledge embedded in a proprietary platform.
Layer 3: Logistics (Amazon Freight and Supply Chain Services)
This is the layer that prompted the original sovereignty article, and it’s the one that changed most recently.
In June 2026, Amazon opened its LTL freight service to all domestic destinations — not just inbound to Amazon facilities. Ridgeline can now use Amazon Freight to move equipment between their three locations, ship palletized supplies from distributors to their warehouse, and deliver materials to job sites. The service includes next-day live pickup for orders placed by 5 p.m., real-time GPS tracking from pickup through delivery, automated appointment scheduling at receiving facilities, and electronic proof of delivery.
Amazon Supply Chain Services (ASCS), launched in May 2026, goes further. Ridgeline gets access to Amazon’s fleet of more than 80,000 trailers, 24,000 intermodal containers, and 100 aircraft. For a facilities management company that occasionally needs to move heavy equipment between metros or receive bulk supply shipments, this is infrastructure they could never build themselves.
Companies like Procter & Gamble, 3M, and American Eagle Outfitters have already signed on to ASCS. Peter Larsen, VP of Amazon Supply Chain Services, explicitly compared the play to what AWS did for cloud computing — taking Amazon’s internal infrastructure and selling it to everyone.
Verdict: Impressive infrastructure, sovereignty risk intensifying. The logistics layer is where the vertical stack thesis becomes most visible. Amazon is now your cloud provider, your procurement platform, and your freight carrier. Each layer is individually competitive. Together, they create an integrated dependency that would be extremely painful to unwind.
Layer 4: Customer Communication (Amazon Connect)
Ridgeline’s customer communication is probably a disaster. Property managers call a main office number, someone writes the request on a sticky note, and it may or may not make it to the right technician. For a growing company, this breaks fast.
Amazon Connect — recently rebranded to Amazon Connect Customer — is AWS’s cloud contact center service. It handles inbound and outbound calls, chat, email, and task routing. In April 2026, AWS expanded the portfolio to include Amazon Connect Decisions for supply chain workflows and announced 29 agentic AI features including pre-built autonomous AI agents that can handle routine customer interactions without human intervention.
For Ridgeline, this means a property manager calls in, an AI agent captures the issue details, checks technician availability against the scheduling system, and either books the appointment directly or routes to a human dispatcher for complex situations. The system integrates natively with other AWS services — the call transcript goes to S3, the AI processing runs on Bedrock, the customer record updates in their RDS database.
Verdict: Powerful, and deeply entangling. Connect is a genuinely good contact center product. It’s also the layer where Amazon’s vertical integration becomes most seamless — and most difficult to extract. Your call recordings, AI training data, workflow automations, and customer interaction history all live in the AWS ecosystem. Moving to Twilio or a competing platform means rebuilding every automation from scratch.
Layer 5: Payments (Amazon Pay and Business Credit)
This is where the stack gets thinner. Amazon Pay is primarily designed for e-commerce checkout — letting customers pay on third-party websites using their Amazon credentials. It’s supported by more than 720,000 merchants, but it’s fundamentally a consumer checkout tool.
For Ridgeline, which invoices property management companies for services rendered, Amazon Pay doesn’t solve the core problem. They need accounts receivable, net-30 invoicing, and integration with their accounting system. Amazon’s recent rebrand of “Pay by Invoice” to “Business Credit Account” shows they’re moving in this direction, but the offering is still oriented around purchasing from Amazon, not general B2B invoicing.
Verdict: Gap in the stack. This is where the Amazon-only thought experiment breaks down for a services business. Ridgeline still needs Stripe or a traditional payment processor for customer invoicing, and QuickBooks or similar for accounting. Amazon hasn’t built the B2B financial layer that would complete the sovereignty loop for a company like this.
Layer 6: The Integration Tax
Here’s what you don’t see in any individual product evaluation: the integration tax paid by companies that don’t go all-in on one stack.
If Ridgeline uses AWS for infrastructure, Amazon Business for procurement, Amazon Freight for logistics, and Amazon Connect for customer communication — those four systems talk to each other with minimal friction. Procurement data flows into spend dashboards that inform logistics decisions. Customer calls trigger workflows that check inventory levels sourced from procurement data. AI models trained on call transcripts improve the automated responses that run on the same cloud infrastructure.
The moment Ridgeline picks a non-Amazon tool for any layer — say, Twilio for communications or a traditional freight broker for logistics — they inherit an integration burden. APIs to maintain, data to sync, authentication to manage, and failure modes that multiply with each connection point.
This is the actual mechanism of sovereignty capture. It’s not that any single Amazon service is irreplaceable. It’s that the integrated stack creates compound convenience that makes piecemeal alternatives feel expensive and fragile by comparison.
Where I Actually Landed
After walking through this exercise honestly, here’s what I think:
Amazon wins on logistics and procurement for a company shaped like Ridgeline. The combination of Amazon Business and Amazon Supply Chain Services solves real operational pain that mid-market companies currently address with duct tape and spreadsheets. No other single vendor offers this combination.
AWS wins on breadth but not uniquely on depth. Google Cloud and Azure are legitimate alternatives for compute and AI. The choice between them is real, not a formality. I chose Google Cloud for my own stack because of Vertex AI’s model garden and the integration with Google’s broader ecosystem. Ridgeline could make a credible case for any of the three.
The sovereignty risk is real but not uniform. Logistics and procurement lock-in happens through accumulated operational data and workflow dependencies. Cloud lock-in happens through service-specific architectures. Payments is the one layer where Amazon hasn’t closed the loop, which means Ridgeline still needs external financial infrastructure regardless.
The honest conclusion: building entirely on Amazon is more viable in 2026 than it was even six months ago. The ASCS launch and LTL expansion filled the biggest gaps. But “more viable” isn’t the same as “advisable.” The same operational convenience that makes the stack attractive is the mechanism that makes leaving expensive. You’re not buying services — you’re joining an ecosystem. And ecosystems have gravity.
That’s not a reason to avoid Amazon’s services categorically. Some of them — particularly ASCS for logistics — are genuinely best-in-class. The discipline is in choosing deliberately: use the layers where Amazon demonstrably wins, maintain alternatives where the switching costs are highest, and never mistake integration convenience for strategic advantage.
The companies that thrive in this environment won’t be the ones that went all-in on any single stack. They’ll be the ones that understood which layers to rent and which ones to own.