Enterprises are stuck between hyperscaler lock-in and the complexity of managing their own AI infrastructure. Leaseweb has the assets to become the trusted alternative. This is the commercial strategy to make it happen.
"The question is not whether enterprises need an alternative to hyperscaler AI infrastructure. The question is whether Leaseweb can build the commercial engine fast enough to become that alternative before the window closes."
The infrastructure decision is becoming as strategic as the AI model choice itself. Enterprises are discovering that hyperscaler lock-in, unpredictable costs, and data sovereignty concerns are preventing AI from moving beyond pilot.
21% of enterprise workloads are being pulled off hyperscalers4 due to economics that do not work at scale, especially for GPU training and high-throughput inference. The CFO cannot approve what the CFO cannot predict.
EU AI Act and GDPR are making infrastructure jurisdiction a compliance requirement, not a preference. 61% of Western European CIOs now prioritise local providers.5 Leaseweb announced accelerated sovereign cloud development this week.6
GPU availability is improving, but enterprises need partners who understand workload configuration, cost modelling, and the path from pilot to production, not just hardware vendors reading spec sheets.
These are not hypothetical objections. These are the conversations I have had across 18 years of enterprise technology sales, adapted for the AI infrastructure decision. Each answer ends with a reframing question that changes the conversation.
The honest answer is: maybe you should not. If your workloads are tightly integrated into a hyperscaler ecosystem and the economics work at your current scale, migration creates risk without sufficient reward.
But here is what we see in practice. When enterprises scale AI from pilot to production, three things happen simultaneously: costs become unpredictable, performance becomes inconsistent due to shared infrastructure, and data sovereignty becomes a board-level concern.
Leaseweb's bare metal GPU servers, spanning NVIDIA L4, L40S, and H100 NVL7 across 28 data centres globally,8 typically deliver 30 to 50% cost savings3 for predictable, high-throughput AI workloads. With 99.99% uptime SLA.3
Leaseweb is European-headquartered, founded in Amsterdam in 1997.9 Our data centres across Europe operate under EU jurisdiction, which means your data stays within EU borders by design, not by workaround.
For organisations navigating the EU AI Act, this matters because the regulation requires transparency about where AI models are trained and where data is processed. Hyperscalers offer EU regions, but the corporate entity and data access sit under US jurisdiction. Leaseweb removes that complexity entirely.
We offer NVIDIA L4, L40S, and H100 NVL GPUs, available globally as dedicated servers.7 The L4 is also available in Public Cloud across the UK, US, Canada, Netherlands, and Germany, with G6 GPU-optimised instances offering 1 to 4 L4 GPUs with AMD EPYC v3 CPUs and 24GB VRAM.10
The choice depends on your workload profile, not your aspiration. The L4 is optimised for inference. The L40S handles heavier mixed workloads. The H100 NVL is for large-scale model training. Most enterprises over-specify initially, and a trusted advisor helps them right-size from day one.
Leaseweb provides standard infrastructure, standard APIs, and transparent pricing with no proprietary lock-in mechanisms. Your AI workloads run on standard NVIDIA GPUs with standard drivers.
Our model is simple: we earn your business every month by delivering performance, reliability, and value. Not by making it expensive to leave.
This is where the real partnership begins. Scaling from pilot to production is where 95% of enterprise AI projects fail, according to MIT research.2 The infrastructure is rarely the root cause, it is typically governance, cost predictability, and the confidence to commit.
Start with a single GPU server to validate. Once validated, scale to a dedicated cluster. We provide technical consultation throughout, plus quarterly business reviews and proactive capacity planning.
CoreWeave is GPU-native and Kubernetes-first, with a $66.8B revenue backlog11 and massive scale ambitions. But they are US-centric, heavily VC-funded, and their model depends on continued capital access.12 OVHcloud is a strong European player, but their AI-specific portfolio is less developed.
Leaseweb combines European sovereignty, a 27-year track record,9 a dedicated AI infrastructure portfolio spanning L4 to H100,7 global reach across 28 data centres in EU, US, and APAC,8 and the hybrid flexibility that enterprises actually need. No single competitor offers all five.
"Is Leaseweb building a GPU rental business, or is it building an AI infrastructure advisory practice that monetises through compute? The answer determines whether we compete on price or on value."
Focused on enterprises where cost transparency, EU sovereignty, and bare metal performance create the strongest differentiation against hyperscalers and GPU-native competitors.
Companies with proven AI workloads that have outgrown hyperscaler economics.
Cloud bill shock: unpredictable costs that make CFO approval impossible for scaling.
Transparent pricing, 30-50% savings, dedicated bare metal with no noisy neighbours.
Healthcare, financial services, and government where data sovereignty is mandatory.
EU AI Act compliance that hyperscalers cannot fully address from US-headquartered infrastructure.
European-headquartered, full GDPR compliance, sovereign cloud within EU borders.
GPU-intensive rendering, generative content production, and video processing at scale.
Need burst GPU capacity for production deadlines without long-term hyperscaler commitments.
Rapid deployment (hours not weeks), flexible commitment models, dedicated GPU performance.
| Leaseweb | Hyperscalers | CoreWeave | Hetzner | |
|---|---|---|---|---|
| Cost transparency | ★★★★★ | ★★☆☆☆ | ★★★☆☆ | ★★★★★ |
| EU data sovereignty | ★★★★★ | ★★★☆☆ | ★☆☆☆☆ | ★★★★☆ |
| GPU range (L4 to H100) | ★★★★☆ | ★★★★★ | ★★★★★ | ★★★☆☆ |
| Enterprise support | ★★★★☆ | ★★★★★ | ★★★☆☆ | ★★☆☆☆ |
| Bare metal performance | ★★★★★ | ★★★☆☆ | ★★★★☆ | ★★★★★ |
| No lock-in | ★★★★★ | ★★☆☆☆ | ★★★☆☆ | ★★★★☆ |
A three-phase approach that builds the advisory foundation before scaling the commercial engine. Each phase delivers measurable outcomes, not just activity.
The BVF is a proprietary framework I built to help enterprises stop guessing and start measuring which AI initiatives deserve investment. It synthesises McKinsey, Gartner, BCG, Deloitte, and Forrester benchmarks into two actionable outputs: an Organisational Drag Cost and a Decision Confidence Score.
The quantified Euro value currently at risk due to structural and management constraints. This makes the status quo more expensive than the transformation, the single most powerful argument for a CFO.
Every AI initiative receives a classification: Accelerate (invest now), Fix (address barriers first), or Stop (redirect resources). Data-driven, not political.
Leaseweb's AI infrastructure customers are making exactly these decisions. A Business Development Manager who can walk into a prospect conversation with a diagnostic framework, rather than a price list, changes the nature of the relationship entirely. The BVF positions Leaseweb as a trusted advisor, not a commodity vendor.
"95% of GenAI pilots fail to deliver measurable P&L impact.2 The BVF exists to identify which 5% deserve scale, and which infrastructure decisions accelerate that journey."
A weekly newsletter I write and publish, connecting AI developments to enterprise transformation strategy. This is how I would build Leaseweb's thought leadership engine.
My proprietary five-layer methodology for diagnosing why AI transformations fail. The core insight: AI adoption is a management and structural problem, not a technical one. Until leadership addresses the underlying "Bureaucracy Debt," technology investment yields diminishing returns.
The foundation most frameworks skip. You cannot transform what you cannot be honest about.
Make the status quo more expensive than the transformation.
Transformation travels through relationships, not org charts.
Align the strategic layer with the operational layer before deploying AI at scale.
Desire before Knowledge. Training without buy-in is wasted capital.
AI transformation is not a project, it is a permanent operating rhythm.
18 years at HPE, Microsoft, Salesforce, and Atos. I know how vendors structure deals, where the hidden costs sit, and what makes a customer trust you enough to give you their most critical workloads.
In 2014, I suffered a cardiac arrest and was resuscitated. Since then, I have completed six desert ultramarathons including four finishes of the Marathon des Sables, one of the toughest footraces on earth, and received the "Never Give Up" award in 2025. Building a new AI business development function is a marathon, not a sprint. I know what it takes.
Every great strategic decision starts with a question that stops the room.
craig@craighortonadvisory.com
Amsterdam, Netherlands