Record capital floods AI as Amazon exits AGI race

The week of 18–22 August brought a cascade of headline-sized financial commitments to AI infrastructure, alongside a significant strategic retreat from one of the industry's largest players.

Nvidia Guarantees $105 Billion for OpenAI's Ohio Data Centre

Nvidia has committed a $105 billion financial guarantee to support OpenAI's planned data centre campus in Ohio — one of the largest single infrastructure backing arrangements in technology history. The agreement positions Nvidia as more than a chip supplier; it becomes a financial underwriter for the compute capacity that frontier AI models require. The Ohio facility is expected to run Nvidia hardware at scale and will serve as a key node in OpenAI's expanded US infrastructure footprint. The deal reflects how tightly the AI supply chain is becoming integrated, with hardware manufacturers taking on new roles financing the facilities their products power.

Databricks Raises $5 Billion at a $190 Billion Valuation

Databricks, the enterprise data and AI platform, closed a $5 billion funding round this week, pushing its valuation to $190 billion — among the highest ever for a private technology company. The round follows an earlier 2026 raise and reflects sustained investor appetite for platforms that manage the data pipelines AI systems depend on. Unlike frontier model labs, Databricks builds data infrastructure rather than large language models, serving enterprises that want to deploy AI reliably at scale. At $190 billion, the company exceeds the public market capitalisation of many established technology firms, and a listing is widely anticipated.

Anthropic in Discussions for a Potential $2 Trillion IPO

Anthropic, maker of the Claude family of AI assistants, is reported to be in early discussions about a public listing that could value the company at up to $2 trillion — which would be the largest initial public offering in financial history if completed at that figure. The reported valuation reflects a stark contrast with OpenAI, whose revenue growth has reportedly slowed this year while Anthropic's enterprise business has continued to expand. Anthropic has not confirmed a timeline or target valuation, and public market conditions will shape whether and when any listing proceeds. The reports signal that institutional investors view frontier AI safety research and commercial deployment as compatible — and potentially very profitable — long-term bets.

Anthropic Shifts Enterprise Data Retention to Client Private Clouds

In a separate development, Anthropic announced it will allow enterprise customers to store conversation and usage data within their own private cloud environments rather than on Anthropic's infrastructure. The shift responds to longstanding corporate concerns about data sovereignty — concerns that have slowed AI adoption in regulated industries such as finance, healthcare, and legal services. The change is notable because AI safety research has historically depended on access to model interaction data; keeping it in customer silos makes it harder for labs to monitor for emergent model behaviours. Anthropic has signalled it will work with enterprise clients on alternative safety-monitoring arrangements, though the specifics have not yet been published.

Amazon Closes Its AGI Lab and Pledges $220 Billion to Infrastructure

Amazon has confirmed it is closing its internal artificial general intelligence research laboratory and redirecting its AI strategy entirely. Rather than competing to build frontier models, Amazon is committing $220 billion to cloud and AI infrastructure — data centres, specialised chips, and the networking fabric that connects them. The decision makes Amazon the single largest capital investor in AI infrastructure among major technology companies, while stepping back from the model-building race it entered quietly in recent years. For enterprise customers, the pivot means AWS will continue to host and serve models built by others — including Anthropic, in which Amazon has previously invested — rather than developing its own flagship AI system.