How Are EU Startups Supposed to Quit US Cloud?

Not all traps looks the same, some are used to catch bears, mice but free credits are just another trap hiding behind a fancy US branded logo

Drug dealers have understood this. The first hit is free. It is not charity. It is customer acquisition. Once the dependency is established, the customer does not comparison shop. They do not evaluate alternatives. They just keep paying, because the cost of withdrawal is worse than the cost of the next fix.

Social media companies perfected the digital version. Facebook, Instagram, TikTok, they gave you a free platform, let you build your audience, your identity, your business on top of it. Try leaving now. Your followers do not come with you. Your content is not portable. Your reach disappears overnight. The platform is free but the cost of leaving is everything you built.

Gambling works the same way. Casinos do not charge you to sit at the table. The drinks are free. The first few wins come easy. By the time you realise the house always wins, you are already chasing losses and the idea of walking away feels like giving up on money that is rightfully yours.

Vendor lock-in follows exactly the same pattern. AWS, Google Cloud, Microsoft Azure, and Oracle hand startups hundreds of thousands in free cloud credits. The product is genuinely excellent. The documentation is world-class. The services are easy to adopt. And by the time the credits run out, your entire architecture is built on proprietary APIs, managed databases, and deployment pipelines that only work on that one provider. You are not a customer anymore. You are captive.

Here is the full picture, US and EU side by side:

Region	Provider	Programme	Max Credits	Duration
US	AWS	Activate (Portfolio)	$100,000	12 to 24 months
US	AWS	Activate (AI Tier)	$300,000	12 to 24 months
US	Google Cloud	Google for Startups	$200,000	24 months
US	Google Cloud	Google for Startups (AI)	$350,000	24 months
US	Microsoft Azure	Founders Hub (Investor)	$150,000	Varies
US	Oracle Cloud	Oracle for Startups	Free credits + 70% discount	Ongoing
EU	OVHcloud (FR)	Startup Programme (Scale)	€100,000	12 months
EU	OVHcloud (FR)	Startup Programme (Start)	€10,000	12 months
EU	Scaleway (FR)	Growth Stage	€36,000	12 months
EU	Scaleway (FR)	Early Stage	€9,000	6 months
EU	IONOS (DE)	Startup Programme	€100,000	Up to 5 years

These are not theoretical maximums buried in fine print. AWS's Portfolio tier is available to any startup with a VC or accelerator affiliation. Google's AI tier actively recruits companies building on foundation models. Microsoft's investor offer routinely starts at $100,000. For a pre-revenue company, these credits are not a marketing perk, they are the difference between twelve months of runway and eighteen.

Now look at the other side. OVHcloud, the largest independent European cloud provider, runs a startup programme with two tiers: a Start level at €10,000 in credits, and a Scale level at up to €100,000, but you need to have closed a Series A to qualify for that. Scaleway, the developer-friendly French provider loved by the Hacker News crowd, tops out at €36,000 for its Growth Stage programme, aimed at seed to Series A companies. IONOS, the German provider backed by United Internet, offers up to €100,000 but spread over five years, with an initial SPARK credit of €5,000.

The gap is not just in headline numbers. The US programmes come bundled with dedicated startup engineering teams, prioritised support tickets, AI and ML service credits, marketplace listing opportunities, and crucially, direct connections to the VC networks that decide your next funding round. When your startup is in the AWS or Google Cloud portfolio programme, you are signalling to investors that you are building on infrastructure they understand, trust, and have portfolio-wide pricing agreements with.

EU providers offer technical consultation hours and some partner perks, which is genuinely useful. But no European cloud provider can pick up the phone and connect you to Sequoia or Andreessen Horowitz. The credit programmes are not just about compute, they are about ecosystem lock-in disguised as generosity.

Spotify (Sweden, on Google Cloud). Sweden's most famous tech export and one of Europe's biggest success stories, ran on AWS for years before migrating its entire backend to Google Cloud Platform in 2016. The migration involved moving 1.5 billion files. Today, Spotify uses BigQuery, Pub/Sub, and Dataflow as core infrastructure. A Swedish company, serving European users, processing European data, running on American cloud. Spotify chose Google for the depth of its data analytics tooling, not because European alternatives existed, because at the time, and arguably still, nothing in Europe comes close to BigQuery at that scale.

Klarna (Sweden, on AWS). The Swedish fintech valued at over $6 billion, selected AWS as its preferred cloud provider in 2019. The deal was explicit: Klarna built its core banking platform on AWS from the ground up, uses Amazon EKS for container orchestration, and relies on SageMaker for real-time credit scoring. Klarna processes payments for 60 million customers across 17 countries on American infrastructure. As a fully licensed European bank, it works with AWS compliance teams to meet regulatory requirements, but the data flows through US-owned pipes.

Delivery Hero (Germany, on AWS). The Berlin-based food delivery giant operating across 70 countries, has been cloud-native on AWS since its founding in 2011. It processes 10 million daily orders using Amazon EventBridge, runs Kubernetes on EKS, and even migrated from Google Cloud to AWS for its image service. Delivery Hero's entire operational backbone, route optimisation, real-time order tracking, rider allocation, runs on American infrastructure.

Revolut (UK / EU-licensed, on Google Cloud). The London-founded (now EU-licensed) digital bank with 40+ million users, chose Google Cloud after evaluating multiple providers. It uses GCP for the scalability needed to support its explosive growth across European markets.

These are not fringe cases. They are Europe's flagship tech companies. Every one of them made rational infrastructure decisions based on what was available, affordable, and technically superior at the time they were scaling. And every one of them is now deeply locked into US cloud ecosystems in ways that would cost hundreds of millions to reverse.

The economics of cloud credits create a dependency spiral that is fiendishly effective. Here is how it works in practice:

Phase 1: The gift. A startup receives $100,000 to $350,000 in free credits. Engineers build on the provider's managed services: Aurora instead of self-hosted Postgres, BigQuery instead of a self-managed data warehouse, Lambda or Cloud Functions instead of a standard compute layer. These services are fast to adopt, well-documented, and genuinely excellent. The startup ships faster than it would on raw VMs.

Phase 2: The habit. Twelve months in, the credits are running out. The startup is now deeply integrated with provider-specific APIs, IAM patterns, monitoring dashboards, and deployment tooling. Engineers have built muscle memory around the provider's CLI and console. Infrastructure-as-code templates reference hundreds of provider-specific resources. The cost of migrating is no longer zero. It is measured in engineering months.

Phase 3: The bill. Credits expire. The startup is now paying full price, or more commonly, negotiating an enterprise discount agreement (EDA). The cloud bill becomes one of the top three operating costs. Migrating would mean rewriting data pipelines, retraining operations teams, negotiating new compliance certifications, and accepting months of reduced engineering velocity. The startup stays.

Phase 4: The permanence. By Series B or C, the cloud architecture is load-bearing infrastructure. Migrating is a multi-year programme, not a project. Spotify's move from AWS to GCP took over two years with a dedicated platform team. Most companies never attempt it at all.

This is not a conspiracy. It is a business model. The credits are a customer acquisition cost, and the lifetime value of a startup that grows into a unicorn paying seven or eight figures annually in cloud spend makes $350,000 in free credits look like rounding error. US hyperscalers can fund these programmes because they operate at a scale that no European provider approaches.

The credit gap is a symptom of a deeper structural issue: Europe does not have a hyperscaler. Not one. AWS, Azure, and Google Cloud together control roughly 70% of the European cloud market. European providers collectively hold about 15%. The entire EU cloud industry is a rounding error in Amazon's quarterly earnings.

This matters for startup credits because these programmes are loss-leaders funded by massive operating margins at scale. AWS generated over $90 billion in revenue in 2023. It can afford to give away $300,000 to every promising AI startup on the continent and barely notice. OVHcloud's total annual revenue is around €870 million. Scaleway does not publicly disclose revenue but is estimated in the low hundreds of millions. These companies literally cannot afford to match US credit programmes without betting the company.

The EU's response has been institutional rather than commercial. The 8ra initiative (formerly IPCEI-CIS), backed by Deutsche Telekom, SAP, IONOS, and Schwarz Digits, aims to build sovereign European cloud infrastructure, but its focus is on large-scale AI data centres and government workloads, not startup credit programmes. The European Commission has announced €20 billion in funding for AI-focused data centres, which is significant but targeted at infrastructure buildout, not startup acquisition.

Meanwhile, SAP has committed €20 billion over the next decade to European sovereign cloud. Deutsche Telekom's T Cloud offers sovereign cloud solutions from German data centres. These are real investments, but they serve enterprise and government customers, not a two-person startup in Tallinn trying to decide between Firebase and self-hosted Supabase.

Matching US credit programmes euro-for-dollar is probably not realistic, and may not be necessary. But the current gap is so wide that it does not even register as a competitive consideration for most founders. Here is what could shift the equation:

A pan-European startup credit fund. The EIC (European Innovation Council) already disburses billions in grants. A dedicated programme offering €100,000 to €250,000 in cloud credits redeemable at any qualifying EU cloud provider (OVHcloud, Scaleway, IONOS, Hetzner, Exoscale) would let startups choose sovereign infrastructure without a financial penalty. The credits could be funded jointly by the Commission and participating providers, similar to how AWS credits are co-funded by VC partnerships.

VC pressure and procurement requirements. European investors could require or incentivise portfolio companies to evaluate EU cloud providers as part of due diligence. Public sector procurement, which is already moving toward sovereignty requirements under EUCS and NIS2, could be extended to grant-funded startups: if you take EU money, host on EU infrastructure.

Technical parity on managed services. Credits alone will not solve the problem if EU providers do not offer the managed services that make hyperscalers sticky. Scaleway's Kapsule (managed Kubernetes) and OVHcloud's Public Cloud are solid, but neither has an equivalent to BigQuery, SageMaker, or DynamoDB. Open-source alternatives exist (ClickHouse, MLflow, ScyllaDB) but the "managed" part matters enormously when you have three engineers and no dedicated platform team.

Honest marketing. European cloud providers need to stop competing on sovereignty alone. Founders do not wake up thinking about the CLOUD Act. They wake up thinking about shipping features, controlling burn rate, and not going bankrupt. The pitch needs to be: we are cheaper per unit than AWS (Hetzner demonstrably is), our support is faster (smaller providers often are), and here are credits to get started.

Every year, thousands of European startups make their cloud decision in the first weeks of existence. Most of them pick AWS, Google, or Azure, not because they evaluated alternatives, but because the credits showed up in their inbox before they had finished incorporating. By the time anyone asks them about digital sovereignty, they are two years and 200 microservices deep into an ecosystem they cannot leave without bleeding engineering time they do not have.

Here is the uncomfortable truth that nobody in Brussels wants to say out loud: we are trying to build digital sovereignty on American foundations. We pass GDPR, we draft the AI Act, we design certification schemes like EUCS, we talk endlessly about strategic autonomy. But the brick and mortar underneath all of it, the actual servers, the actual compute, the actual storage where European data lives and European innovation runs, belongs to American companies. You cannot build a sovereign house on someone else's land. And right now, 96% of the cloud capacity used by Europeans sits on US infrastructure, subject to US law, at the discretion of US companies.

Innovation in Europe is real. Mistral AI is valued at €14 billion. Aleph Alpha is positioning itself as the sovereign AI choice for European governments. The talent is here. The ideas are here. The regulation is here. What is missing is the foundation. These startups still train their models and serve their customers on infrastructure that the EU has no control over. We are building the top floors of a skyscraper while ignoring the fact that we do not own the ground floor.

Fixing this requires two things, and we are falling short on both. First, financial incentives. We need to make it economically rational for a startup to choose EU cloud. Right now it is a sacrifice. A founder picking Scaleway over AWS is leaving $200,000 on the table and getting worse managed services in return. That is not a fair fight. We need EU-level credit programmes, co-funded by the Commission and EU cloud providers, that close the gap enough to make the decision competitive.

Second, features. Incentives get startups through the door, but they will not stay if the product is not good enough. EU cloud providers need to invest in the managed services that make hyperscalers sticky: serverless compute, managed databases, AI/ML tooling, analytics engines. Sovereignty is a compelling argument for government procurement, but a startup founder cares about shipping features, controlling burn rate, and not going bankrupt. The pitch cannot just be "we are European." It has to be "we are European and our product is good enough that you would choose it anyway."

This is not a problem that regulation can solve. You cannot GDPR your way into a competitive cloud market. It requires commercial action, real money, real product investment, and a realistic acknowledgment that the playing field is absurdly tilted. We built GDPR to protect European data. We might want to also build the infrastructure to keep it here.