Anthropic is intensifying efforts to secure data center capacity across Europe, signaling a calculated move into a region where energy economics, regulation, and access to compute are becoming decisive factors in the global artificial intelligence race.

The company is hiring a senior dealmaker in London to lead negotiations for European infrastructure, a role it describes as “critical” to sourcing and executing agreements that will power its next generation of AI systems. The position, focused on commercial sourcing and transaction execution, highlights how infrastructure procurement has evolved into a specialized, high-stakes function, akin to energy trading or large-scale real estate finance.

The move meets a broader shift in the industry. As AI models grow more complex, the bottleneck is no longer just talent or algorithms, but access to reliable, high-density compute. Training and deploying frontier systems now requires vast amounts of electricity, specialized chips, and physical infrastructure—resources that are increasingly scarce and geographically uneven.

Anthropic has already moved aggressively to secure supply elsewhere. It recently committed more than $100 billion over a decade to Amazon Web Services, effectively locking in long-term access to cloud capacity. It has also expanded its partnership with Broadcom, securing about 3.5 gigawatts of compute—an amount comparable to the output of multiple large power plants.

Those commitments underline the scale of the challenge. Industry estimates suggest hyperscalers’ AI infrastructure spending will exceed $600 billion by 2026, with capital flowing not just into servers and chips, but into land acquisition, power generation, and cooling systems. In that context, Europe is emerging as both an opportunity and a constraint.

The region offers proximity to enterprise customers, strong regulatory frameworks, and increasing demand for sovereign or locally hosted AI services, particularly among governments and regulated industries. It also presents structural hurdles: high energy costs in key markets, lengthy permitting processes, and growing scrutiny over environmental impact.

Anthropic’s approach appears to be pragmatic as the London-based role will target established data center hubs, Frankfurt, London, Amsterdam, Paris, and Dublin, while also exploring newer markets in the Nordics and Southern Europe. This is seen as a dual strategy: securing capacity in core commercial centers while diversifying into regions where power is cheaper and more abundant.

The Nordics have become a focal point for that expansion. Countries such as Norway and Finland offer access to low-cost, renewable energy and cooler climates that reduce cooling expenses, making them attractive for large-scale AI workloads. Microsoft has already moved to expand its presence in Norway, while Nebius is developing a major facility in Finland.

Southern Europe is also gaining traction as a secondary growth corridor. Microsoft has committed billions to data center projects in Portugal and Spain, and Oracle is advancing infrastructure plans in Italy. These investments suggest a broader rebalancing of Europe’s data center map, driven by the search for lower energy costs and less congested grids.

Yet the constraints are too real to ignore. OpenAI recently halted its planned “Stargate” project in the United Kingdom, citing high electricity prices and regulatory complexity. The decision highlights a key risk for Anthropic: that Europe’s cost structure could limit the economic viability of large-scale AI deployments, even as demand grows.

This tension is shaping how deals are structured. Rather than relying solely on traditional cloud leasing, AI firms are increasingly exploring direct agreements with developers, long-term power purchase arrangements, and hybrid ownership models that provide greater control over costs and capacity. Anthropic’s reported evaluation of direct acquisition deals suggests it is moving in that direction.

There is also a geopolitical layer as European policymakers are pushing for greater technological sovereignty, seeking to reduce reliance on foreign-controlled infrastructure. That could create opportunities for companies willing to localize operations, but it may also introduce additional regulatory requirements and scrutiny over data handling and security.

Thus, the expansion into Europe does not seem optional for Anthropic. As competition with OpenAI intensifies, access to compute is becoming a defining advantage. Both companies are scaling their presence in the region, recognizing that enterprise adoption will increasingly depend on local infrastructure availability and compliance with regional regulations.

The hiring of a dedicated transaction lead signals that Anthropic is preparing for a sustained campaign rather than isolated deals. It also underscores that the AI race is being fought as much in power grids and data centers as in research labs.

What emerges is a picture of an industry entering a more capital-intensive, infrastructure-driven phase. Europe, with its mix of demand, policy ambition, and structural constraints, is set to play a pivotal role in that outcome.

ServiceNow delivered a decisive message to a market preoccupied with disruption: artificial intelligence is reinforcing its business rather than eroding it.

The company reported first-quarter subscription revenue of $3.67 billion, up 22% year over year, beating the high end of its guidance across topline growth and profitability metrics. Management also lifted its full-year outlook, projecting 2026 subscription revenue of $15.7 billion to $15.8 billion, implying sustained growth of roughly 22% to 22.5% and outpacing analyst expectations heading into the results.

The performance lands at a sensitive moment for the software sector. Over the past six months, valuations have come under pressure amid concerns that generative AI models from firms such as OpenAI and Anthropic could bypass traditional enterprise platforms by enabling companies to build workflows directly on top of large language models.

Chief executive Bill McDermott rejected that thesis, arguing that real-world enterprise adoption is revealing a different set of constraints — particularly around cost, governance, and operational reliability.

“The results speak a lot louder than the words. We’re now in another beat and raised quarter,” McDermott said, framing the quarter as evidence that demand remains intact even as the technology landscape shifts.

At the heart of ServiceNow’s positioning is its role as an orchestration layer — software that integrates AI capabilities into structured enterprise workflows rather than replacing them. That distinction is becoming increasingly important as companies move from experimentation to scaled deployment of AI tools.

McDermott pointed to accelerating uptake of the company’s AI offerings as a key driver of growth. ServiceNow had previously projected $1 billion in AI-related software revenue by 2026, but that figure has now been revised to at least $1.5 billion, with the potential for further upside.

“We’ll probably blow through that, too, because the acceptance of our AI solutions is just absolutely stunning,” he said.

Forward indicators support that narrative. Remaining performance obligations, a measure of contracted future revenue, rose 25% to $27.7 billion, while current RPO increased 22.5% to $12.64 billion. The expansion signals that large enterprises are not only maintaining spending but locking in multi-year commitments, even as macro conditions remain uneven.

The more consequential insight lies in how enterprises are evaluating AI economics. While direct access to large models offers flexibility, usage-based pricing structures can introduce significant cost volatility, particularly in high-volume operational environments.

McDermott said customers exploring model-centric architectures are encountering a mismatch between theoretical efficiency and practical cost. He cited a case where a chief information officer at a major client assessed using a direct AI model approach to run IT operations. According to McDermott, the model-driven setup would have cost roughly ten times more than deploying ServiceNow’s integrated AI tools.

The issue is not just pricing, but predictability. Enterprise IT budgets are typically structured around fixed or subscription-based costs, whereas AI model usage often scales with demand, making expenses harder to forecast. That unpredictability can become a constraint at scale, particularly in regulated industries where cost control and auditability are critical.

By embedding AI within its platform, ServiceNow is effectively converting variable AI costs into more predictable software spend, while also layering governance, compliance, and workflow management on top. This approach positions the company less as a competitor to model providers and more as an intermediary that translates raw AI capability into enterprise-ready applications.

McDermott was explicit in his critique of standalone AI offerings for enterprise use, describing them as “parlor tricks,” a characterization that underscores the gap between demonstration-level capability and production-grade deployment.

The broader implication is that the competitive landscape is shifting. Rather than a binary contest between traditional software and AI-native systems, the market is evolving into a layered architecture. At the base are model providers, competing on performance and scale. Above them sit platforms like ServiceNow, which integrate those models into business processes, enforce governance, and deliver measurable outcomes.

ServiceNow’s results suggest that this middle layer remains critical. Enterprises are not abandoning platforms; they are demanding that those platforms incorporate AI in ways that align with operational realities.

The company’s momentum also underpins a structural advantage: deep integration into mission-critical workflows such as IT service management, customer operations, and employee systems. These embedded positions make displacement more difficult, even as new technologies emerge.

However, the pace of AI model improvement could compress the value of intermediary layers if models become easier to deploy and manage directly. Pricing dynamics could also shift if model providers move toward more predictable enterprise licensing structures.

For now, however, ServiceNow appears to be benefiting from the transition phase. Its raised guidance, expanding AI revenue expectations, and strong forward bookings indicate that customers are prioritizing integration, reliability, and cost control over experimental flexibility.

In that context, AI is not dismantling the enterprise software stack. It is reshaping it — and, for companies able to absorb and operationalize the technology effectively, extending its growth cycle rather than ending it.