Here’s a plot twist nobody saw coming: JPMorgan Chase, the bank that’s supposed to be all-in on cloud computing, just decided to keep some of its most important AI work locked behind its own firewall.
This week, SambaNova Systems—an AI chip startup that Intel apparently tried to buy for $1.6 billion less than a year ago—raised $1 billion at an $11 billion valuation. The kicker? JPMorgan is the anchor customer, deploying SambaNova’s systems to run AI inference on its own servers, not in the cloud.
For the past two years, the AI infrastructure story has been pretty straightforward: everyone moves to the cloud. AWS, Azure, Google Cloud—pick your poison, pay per token, let someone else worry about the servers. It’s clean, it’s simple, and it’s been mostly true.
But JPMorgan just added a massive asterisk.
## The Data Problem Nobody Wants to Talk About
Here’s the thing: not all data is created equal. Banks sit on client information and proprietary trading strategies they literally cannot send to a third-party server. Hospitals manage patient records that federal law requires them to protect. Defense contractors and government agencies face outright restrictions on running sensitive workloads on commercial cloud infrastructure.
Cloud economics look great on a spreadsheet. But when your data is the crown jewels, that architecture comes with a risk profile that’s just not acceptable.
SambaNova’s CEO basically sent a signal to the entire banking industry: if you want control over your most sensitive AI work, we’ve got the hardware for it. And apparently, JPMorgan was listening.
## The Inference Supercycle Just Got Complicated
The real story here isn’t about one bank making one deal. It’s about the AI infrastructure market splitting into two distinct lanes.
The hyperscalers—AWS, Azure, Google Cloud—will still capture the majority of AI inference demand. That’s not changing. But within regulated industries, a structurally different market is forming for secure, on-premises inference infrastructure.
Banks, hospital systems, and government agencies don’t move fast. But when they do, they move at scale, under long-term contracts, with infrastructure budgets that stick around for years. That’s the kind of revenue stream that makes private capital very, very interested.
The companies positioned to win here are the ones selling the hardware, the networking, the storage, and the software stack that makes on-premises inference work. Dell’s AI Factory already has over 4,000 enterprise customers. Everpure (formerly Pure Storage) has rebuilt its entire platform to make enterprise data accessible to AI workloads without the overhead of replication.
JPMorgan’s decision just made their pitch to the next bank a lot easier.
## The Bottom Line
SambaNova’s valuation jump from a rumored $1.6 billion acquisition target to $11 billion in under a year isn’t random. It reflects something real: private capital has decided that secure, on-premises enterprise AI inference is a durable market, and the price of entry has changed.
The frontier labs and hyperscalers drove phase one of this trade. The enterprise deployment wave is phase two—and within regulated industries, it plays by different rules.
The inference supercycle is real. But the cloud-only thesis? That’s officially dead.
