AWS Educate gives students and teachers access to cloud credits and training. Google for Education provides Google Workspace free to millions of institutions. Microsoft Azure for Students offers $100 in credits with no credit card required. These programmes reach tens of millions of students worldwide — including future infrastructure architects, future CIOs, future technical decision-makers.

This is not philanthropy. It is customer acquisition at the lowest possible marginal cost. A student who learns to code on AWS learns AWS APIs, AWS conventions, AWS services. Their reflexes, code snippets, portfolio projects, certifications — all AWS. When they join a company five years later and recommend a tech stack, they recommend what they know. What they know is what was given to them for free.

The study Sir, What Is This Good For documented that school creates a rupture between symbol and matter. Free Tier adds another layer: it gives access to cloud abstractions without the student ever having configured a physical server, opened a network cabinet, or manipulated a real filesystem. The developer trained on Free Tier knows how to deploy a Lambda function. They do not know what runs underneath. They have never needed to — Free Tier removed that necessity.

AWS Free Tier: 750h EC2 t2.micro/month for 12 months · 5 GB S3 · 1 million Lambda requests/month · 25 GB DynamoDB. Documented — but no real-time alert in the developer's workflow. No red banner in the AWS console when approaching the threshold.

GCP Free Tier: 1 e2-micro instance/month · 5 GB Cloud Storage · 2 million Cloud Functions requests. Same principle: limits invisible in the workflow.

Azure Free Tier: 750h B1S virtual machines for 12 months · 5 GB Blob Storage · 1 million Functions requests.

The project grows. Users arrive. Traffic rises. The developer codes — not monitoring their invoice. Then: first invoice. Brutal. At that point, migrating costs more than paying. Code is in DynamoDB. Infrastructure is in CloudFormation. Developers know AWS — nothing else.

Free Tier is designed so the threshold is crossed exactly when attachment is maximal and exit is hardest. Not a design accident. The design.

A local authority IT department wants to modernise a public service. Budget constrained. A contractor proposes: "let's test on cloud — it's free to start." Project begins on AWS or Google Free Tier. Six months later: threshold exceeded. The authority is a paying customer of a US hyperscaler for a public service funded by local tax. No procurement process — the project started free and did not reach public contract thresholds. The contractor is gone. Internal migration skills do not exist.

Public Free Tier is the junction of two mechanisms documented separately. On one side, Finance Washing (MU'ALLAQA IV): public money funds hyperscalers through procurement. On the other, the Free Tier Bait: project starts free, threshold crossed when attachment is maximal. In the private sector, the decision-maker commits their company. In public, they commit taxpayer money — often without formal procedures. The taxpayer pays twice: once for the public service, once for an exit they cannot fund.

Beyond the consumer Free Tier, hyperscalers distribute massive cloud credit packages to early-stage startups: AWS Activate up to $100,000 · Google for Startups up to $200,000 · Microsoft for Startups up to $150,000. These credits are distributed through accelerators, incubators, and partner venture capital funds.

This is not Free Tier. It is disguised venture capital. The effect is radically different: a startup with $100,000 in cloud credits does not build frugally — it builds like a CAC40 firm. Microservices, proprietary PaaS (managed Kubernetes, database-as-a-service, managed AI), complex distributed architectures. Since it pays nothing, it does not optimise. It accumulates cloud technical debt without knowing it.

When credits run out — generally between 12 and 18 months — the startup discovers its true infrastructure cost. The "TCO crisis" is the moment when the real cost structure imposes itself brutally: an architecture built for free capital cannot be paid for with the real revenues of a growth-stage startup. Two outcomes: raise additional funds to pay infrastructure (entering additional financial dependency on VCs) or migrate under emergency conditions — taking months and risking breaking everything.

VC funds push to consume credits rapidly to accelerate "Time to Market." They have signed partnerships with hyperscalers — each startup in their portfolio consuming credits is a new acquisition for AWS. The VC, the startup, and the hyperscaler form a triangle whose first two fund the third without knowing it.

Hyperscalers offer data transfer into their cloud (Ingress) completely free — permanently, not limited to 12 months. Sending 100 TB into AWS S3: free. Sending 1 PB into GCP Cloud Storage: free. This is an endless Free Tier on data ingress.

The illusion of free works. The organisation migrates its data without financial friction. The decision seems rational — no immediate cost, no visible risk. The data enters. The data accumulates. Metadata is created. Indexes are built. Applications connect. Dependency installs itself silently.

When the organisation wants to exit or repatriate data to its physical hardware, pricing reverses brutally:

AWS: $0.09/GB for the first 10 TB outbound per month (to internet) · regional rates vary · free to other AWS services (creating additional inter-service dependency).

GCP: $0.08/GB for internet egress · free internally.

Azure: $0.087/GB for first 10 TB outbound.

On 100 TB of data: $8,000 to $9,000 in egress fees. On 1 PB: $80,000 to $90,000. This is not a service fee — it is an exit ransom. It was documented in Terms Under the Microscope as a contractual retention mechanism. It is also a reverse Free Tier: free entry funds the paid exit.

A computing teacher in a public university wants students to handle real workloads — servers, networks, databases, deployments. Two options.

Option A: use free consoles provided by AWS, Google or Microsoft. Students handle modern, real tools in functional environments. The course is credible. Students learn — vendor APIs, but they learn something.

Option B: not use proprietary consoles. In this case, students handle no modern tools. The university has no budget for lab servers, no room dedicated to physical experimentation, no cluster available for student projects.

The teacher does not choose between AWS and a sovereign alternative. They choose between AWS and nothing. Not negligence — institutional budget constraint.

AWS Educate does not exist out of philanthropy. It exists because Amazon identified that universities had no server budget — and that this void was an opportunity to acquire future captive clients. The State progressively eliminated physical pedagogical infrastructure budgets. Computer lab rooms with real servers disappeared. Computing clusters for student projects are rare. The void exists. Hyperscalers filled it with Free Tier.

Not a coincidence. A strategy. The solution documented in The Maths Teacher's Bag (satellite dish, screws, marble in the parabolic dish) applies here too: not a programme reform, but a physical server in every university. Not an AWS console — a server the student can touch, configure, and break without additional billing.

Student Free Tier · Invisible threshold · Public procurement · Startup credit · Data gravity trap · Budgetary abandonment · Causal chain — seven mechanisms, one logic: make cloud free or nearly free when dependency installs, and expensive or very expensive when exit is considered. Free Tier is not a commercial offer. It is an architecture of capture.