Anthropic’s IPO filing in the US has drawn attention to a less visible part of the AI market: the energy and infrastructure needed to run large-scale generative AI systems. In the filing, the company says competition for electricity supply and data centre locations with enough available power is intense, and it notes growing reliance on natural gas for some data centre expansion.

For European businesses, this matters because AI adoption increasingly depends not only on software models and APIs, but also on energy availability, data centre capacity, supplier resilience and infrastructure costs. Companies using external AI services may not see these constraints directly, but they can still affect pricing, service availability, procurement decisions and sustainability reporting.

What happened

According to reporting on Anthropic’s S-1 filing, the company disclosed that access to electrical supply and suitable data centre locations has become highly competitive. It also acknowledged reliance on natural gas in parts of the data centre buildout supporting AI growth, alongside exposure to fuel cost increases and supply chain constraints.

These statements are notable because they come from a major AI provider preparing to go public, where infrastructure risks must be described to investors. The filing does not create new rules for customers, but it offers a clearer view of operational dependencies behind the AI boom.

Why it matters for European businesses

Many European SMEs and larger digital teams consume AI through cloud platforms, SaaS products or API providers rather than running their own models. Even so, upstream infrastructure constraints can have downstream business effects.

• Cost pressure: If power, fuel or data centre capacity becomes more expensive, AI providers may face higher operating costs that can eventually affect enterprise pricing.
• Capacity and availability risk: Tight power and data centre availability can slow expansion of AI services, especially during periods of high demand.
• Vendor concentration: Businesses that rely heavily on a small number of AI providers may face greater exposure if infrastructure bottlenecks affect those suppliers.
• Sustainability scrutiny: Companies promoting AI-enabled products or internal automation may face more questions about the environmental footprint of the tools they use, particularly in regulated or sustainability-conscious sectors.
• Procurement complexity: Infrastructure resilience is becoming part of AI due diligence, alongside data protection, security and compliance.

This is particularly relevant in Europe, where energy costs, ESG expectations, cloud strategy and digital sovereignty concerns already influence technology purchasing decisions.

Who may be affected

• SMEs adopting generative AI tools for customer service, content, workflow automation or internal productivity.
• E-commerce companies using AI for product content, merchandising, support or personalisation at scale.
• Marketing teams and agencies that depend on AI platforms for campaign production, search visibility, creative generation or analytics support.
• IT and procurement teams assessing long-term AI vendors, cloud dependencies and operational resilience.
• Regulated or sustainability-sensitive businesses that need clearer supplier information for risk, reporting or governance purposes.

What companies should consider

• Review AI supplier concentration: Identify where critical workflows depend on one provider and assess fallback options.
• Ask infrastructure questions in procurement: Request information on service resilience, regional hosting, scaling constraints and major operational dependencies where relevant.
• Monitor total AI cost: Budget for potential pricing changes rather than assuming current API or platform costs will remain stable.
• Check sustainability claims carefully: Avoid making broad environmental claims about AI-enabled services unless they are supported by credible supplier information.
• Prioritise use cases with clear ROI: Where infrastructure costs across the market are rising, practical automation and measurable business value become more important than experimental adoption.
• Include infrastructure risk in AI governance: AI policies should not focus only on model outputs and data protection, but also on continuity, dependency and supplier risk.

The main takeaway is that AI adoption is increasingly shaped by physical infrastructure constraints, not just software innovation. For European businesses, that makes vendor resilience, cost control and realistic AI planning more important as the market matures.