JPMorganChase will run AI workloads on chips from SambaNova inside its own data centres, a test of whether the largest US bank can keep artificial intelligence off both the public cloud and Nvidia’s hardware, as the chipmaker banked a first $1 billion of a Series F round at an $11 billion valuation.
The bank has named SambaNova an inference infrastructure partner and will deploy its SN40 and SN50 systems to run AI “behind the bank’s own walls”, keeping the data and an auditable trail inside its perimeter rather than sending workloads to a shared cloud. It is a trial, not a production rollout.
“We’re excited to deploy SambaNova’s RDU architecture and looking forward to testing its speed and security for on-prem inference in our demanding enterprise AI workloads,” said Darrin Alves, chief information officer for infrastructure platforms at JPMorganChase. AI infrastructure at the bank “has to meet a very high bar for performance, control and reliability”, he added.
Where regulated finance runs its AI
The question the test poses is where AI actually runs once a bank moves it from experiment into production. Cloud GPUs are plentiful but send data outside a firm’s own environment. On-prem AI inference keeps sensitive customer and market data inside the bank’s own kit, which matters for data-residency rules, latency and audit.
That is the pitch SambaNova makes to banks and governments that will not move sensitive workloads off their own infrastructure. It is one of a small group of chipmakers selling an alternative to renting Nvidia GPUs in the cloud, and a tier-one bank putting its name to a trial is the reference customer that pitch needs.
A finance-heavy cap table
The investor list is the other signal. The first close was led by General Atlantic, with T. Rowe Price Associates, Capital Group and Seligman Ventures, and a roster weighted towards financial institutions: BlackRock, the Qatar Investment Authority, Vista Equity Partners and Intel Capital. The $1 billion is a first close, not the full round.
“SambaNova’s $11 billion valuation highlights the central role that fast inference now plays in the enterprise AI stack,” said Rodrigo Liang, the company’s co-founder and chief executive. Martín Escobari, co-president and head of global growth equity at General Atlantic, said demand for inference was “accelerating well ahead of supply”.
Founded in 2017 and based in San Jose, California, SambaNova sells chips, systems and cloud services for AI inference. It said the capital would go towards expanding capacity, product development and scaling deployments across enterprises, sovereign AI programmes and cloud providers. In February it unveiled the SN50 chip and raised more than $350 million alongside a collaboration with Intel.
What SambaNova has not disclosed is how many of those deployments run in production rather than trial, and JPMorgan’s is the highest-profile name in the pilot column. Convert it to a production contract and the on-prem case has its proof point for every other bank weighing the same move. Leave it a test, and the $11 billion rests on intent.
