Farsight & Ornn bring AI compute in-house for banks

Farsight has partnered with Ornn to let banks run Farsight's AI platform on infrastructure inside their own systems. The arrangement is aimed at financial institutions that want to keep sensitive information within their own environment.

The setup combines Farsight's software with GPU capacity sourced through Ornn, letting banks run AI models on internal data and prior work without sending that material to external providers.

The move comes as banks face growing pressure to adopt AI tools for client work while managing restrictions around confidential information, deal pipelines and internal documents. Many institutions still rely on a small group of hosted model providers, leaving them exposed to external pricing, capacity constraints and product decisions.

For banks that want to operate models themselves, securing the necessary computing resources has also been a hurdle. Ornn's role in the partnership includes supplying access to GPU capacity and providing pricing information through the Ornn Compute Price Index, which it describes as a GPU price index based on transacted deals.

Farsight said its platform can route tasks across different models rather than depend on a single provider. It also said the system can learn a team's templates and standards so work reflects a firm's existing style.

In-house focus

The announcement follows Farsight's separate disclosure that Deutsche Bank is running AI workflows on-premises. The company cited that deployment as an example of the compliance and security approach supported by the new arrangement with Ornn.

Financial institutions have become more cautious about how they use generative AI for work that carries regulatory, reputational and client risk. While generic AI tools are now widely available, banks have been looking for ways to apply them to internal and client-facing tasks without losing control of data or becoming tied to a single infrastructure supplier.

Farsight said the platform is built for financial services and designed around a firm's own data, precedents, templates and standards. According to the company, teams use it across web, PowerPoint, Excel, Word and email for documents including pitch books, memoranda and financial models.

Ornn focuses on access to and pricing of GPU compute. It said it helps organisations obtain, finance, and hedge computing resources, reflecting a broader market effort to treat AI infrastructure as a procured, priced input rather than an opaque service bundle.

That distinction may matter for banks trying to forecast the cost of AI adoption. By separating software from compute procurement, firms may gain more visibility into infrastructure spending and their exposure to swings in supply and pricing.

"Today, a bank that wants to reduce its dependence on a single model provider has few practical alternatives," said Samir Dutta, Co-Founder and Chief Executive Officer of Farsight.

"Working with Ornn, we can offer institutions a solution, with models running on compute in their own environment and producing pitch books, memos and models grounded in the firm's own work," Dutta said.

Wayne Nelms, Co-Founder of Ornn, linked the partnership to cost management as banks increase their use of AI.

"Financial institutions want clarity on what they pay for compute today and the risks they take on as their AI needs grow," said Wayne Nelms, Co-Founder of Ornn.

"Working with Farsight, we're helping banks connect how they use AI with the cost and availability of the compute behind it, so they can make informed decisions today and plan for their exposure tomorrow," Nelms said.