Nscale and Anthropic in Talks for Second $45 Billion AI Compute Mega-Deal

TL;DR

• Anthropic and UK-based Nscale are reportedly in talks for a second $45 billion AI compute deal, which would double their partnership to $90 billion in total committed infrastructure just months after the first agreement.
• The expanded deal would massively scale GPU capacity and data center buildouts across the US, UK, and Europe to train and serve next-generation Claude models and meet soaring enterprise demand.
• If finalized, the agreement would cement Nscale as one of Anthropic's most critical infrastructure partners and intensify the AI arms race with OpenAI, Microsoft, Google, and xAI over scarce compute.

A Staggering Sequel Just Months Later

Anthropic and Nscale are back at the negotiating table, and the numbers are jaw-dropping once again.

According to recent reports, the AI safety and research company and the UK-based neocloud provider are discussing a second $45 billion compute agreement, following closely on the heels of their first $45 billion deal announced in August. If completed, the back-to-back deals would bring the total value of the partnership to an eye-watering $90 billion.

The speed of the sequel underscores just how ravenous the appetite for AI compute has become. Anthropic, the maker of the Claude family of models, has seen explosive growth in enterprise adoption, API usage, and agentic workloads. Securing long-term, dedicated infrastructure has become a top strategic priority, and Nscale has emerged as a willing and rapidly scaling partner to deliver it.

Neither company has officially confirmed the second deal, but sources familiar with the talks suggest negotiations are advanced and focused on capacity expansion timelines, power availability, and GPU deployments.

Recap: What The First $45 Billion Covered

The original $45 billion partnership, unveiled in August, was already one of the largest AI infrastructure commitments ever announced by a European cloud provider.

Under that agreement, Nscale committed to providing Anthropic with massive-scale AI compute capacity powered by tens of thousands of next-generation GPUs, including Nvidia's Blackwell and Rubin-class chips. The deal included development of large-scale data center campuses with dedicated power, advanced liquid cooling, and high-bandwidth networking optimized for frontier model training and inference.

The first deal positioned Nscale as a cornerstone of Anthropic's Project Rainier and broader push to diversify beyond traditional hyperscalers like Amazon Web Services and Google Cloud. For Nscale, backed by major investors and rapidly expanding across Europe and North America, it was a breakout moment that vaulted it into the top tier of AI neoclouds alongside CoreWeave, Crusoe, and Lambda.

Why Anthropic Needs Even More Compute Now

The push for a second massive tranche comes down to three converging pressures: bigger models, bigger demand, and bigger rivals.

First, training next-generation Claude models requires exponentially more compute. Frontier labs have shifted from single giant training runs to continuous pre-training, reinforcement learning, synthetic data generation, and large-scale evaluation — all of which devour GPU hours around the clock.

Second, inference demand is exploding. Claude is now deeply embedded in coding assistants, customer support agents, legal and financial workflows, and autonomous multi-step agents via Anthropic's API and Amazon Bedrock. Enterprise customers want guaranteed capacity, low latency, and data sovereignty in the US and Europe, which requires physically more data centers.

Third, competition is relentless. OpenAI's Stargate project with Oracle and SoftBank, xAI's Colossus supercomputer expansion, and Google DeepMind's TPU buildout have reset expectations for scale. To stay in the frontier race, Anthropic cannot afford to be compute-constrained.

The Infrastructure Play: Gigawatts, GPUs, and Greenfield Campuses

A second $45 billion deal would translate into gigawatts of new power capacity and potentially hundreds of thousands of additional AI accelerators.

Industry analysts expect the expanded partnership to focus on:

Large greenfield data center campuses in the US Midwest and Southeast, where land and power are more readily available, alongside continued expansion in the UK and Norway where Nscale already has a footprint.

Multi-gigawatt power procurement, including grid connections, natural gas turbines as bridge power, and long-term renewable and nuclear power purchase agreements to meet sustainability goals.

Next-gen hardware deployments, likely centered on Nvidia Vera Rubin NVL72 systems and future platforms, with advanced liquid cooling and Infiniband-scale networking.

Sovereign AI capacity in Europe and the UK, a key selling point for Nscale as governments push to keep AI infrastructure and data within national borders.

In practical terms, this is not just renting servers — it is co-building the industrial backbone for the next decade of AI.

What It Means For Nscale's Meteoric Rise

For Nscale, a second mega-deal would be transformative.

Founded as a relatively low-profile GPU cloud provider, Nscale has positioned itself as Europe's answer to CoreWeave — offering purpose-built AI infrastructure at a time when hyperscalers are capacity-constrained. Landing one $45 billion commitment from Anthropic put it on the map. Landing two would make it one of the most valuable private infrastructure companies in the world.

The deal would also likely unlock massive debt financing and project finance for data center construction, following the model pioneered by other neoclouds that use long-term customer contracts to fund expansion. Backers and lenders view Anthropic's revenue growth and $60-plus billion valuation as strong credit support for such buildouts.

Challenges remain, however, including securing permits, grid interconnections, chips, and skilled labor at unprecedented speed — bottlenecks that have delayed projects across the industry.

The Bigger Picture: The $90 Billion Bet On AI's Future

If finalized, the combined $90 billion Anthropic-Nscale partnership would rank among the largest private AI infrastructure commitments in history, rivaling the scale of Stargate and other hyperscale initiatives.

It signals a fundamental shift in how AI labs operate: from cloud tenants to co-architects of power plants, fiber networks, and silicon supply chains. Compute is no longer just an operational expense — it is the moat.

For the AI industry, the message is clear. The race to power next-generation Claude, GPT, Gemini, and Grok models will not be won by algorithms alone, but by who can secure the most GPUs, watts, and data center square footage the fastest.

With Anthropic and Nscale reportedly moving quickly toward a second mega-deal, that race just got a lot more intense — and a lot more expensive.

AndroGuider Team

Articles written by the AndroGuider team. We try to make them thorough and informational while being easy to read.