SambaNova has completed the first close of a $1 billion Series F funding round at an $11 billion post-money valuation. General Atlantic leads the round, with significant investment from Seligman Ventures, T. Rowe Price Associates, Inc., and Capital Group. JPMorganChase is the latest enterprise customer to select SambaNova’s systems for fast, on-premises AI inference, according to SambaNova.
Why it matters
$11 billion. That’s a number that demands context. SambaNova’s previous round was a $350 million Series E, per the company, the new valuation represents a substantial step up from that prior mark, and it’s anchored by a specific enterprise deployment claim rather than a speculative pipeline. When a financial institution of JPMorganChase’s scale selects on-premises AI inference infrastructure, that’s a reference customer with real procurement weight behind it.
The catch is attribution. JPMorganChase’s selection of SambaNova systems is announced by SambaNova, there’s no independent JPMorgan confirmation in the available sources. That’s standard for this type of enterprise partnership announcement, but enterprise infrastructure buyers evaluating whether the JPMorgan reference is replicable should note the distinction between a vendor announcement and a joint press release.
The broader signal is the infrastructure thesis itself. SambaNova’s pitch is that inference, running AI models at production scale, needs purpose-built chips and on-premises deployment for regulated-industry enterprises. Financial services companies can’t always send sensitive data to a cloud API. SambaNova’s SN50, described as its fifth-generation AI chip and purpose-built for agentic inference, is the product being positioned against that need. The company’s sovereign AI deployments across providers in Australia, Europe, and the UK suggest the on-premises thesis is resonating beyond the US market.
Context
This round closes the same week Prime Intellect raised $130 million for open enterprise agent training infrastructure. Two AI infrastructure-layer deals totaling over $1.1 billion in 48 hours, both explicitly targeting enterprise production AI, is a pattern worth noting: infrastructure capital is moving faster than application-layer capital right now, and the bets are on companies that give enterprises an alternative to cloud-API dependency. The Hub’s prior analysis of collapsing inference costs maps the structural dynamic that makes on-premises alternatives more economically viable than they were two years ago.
What to watch
Watch the “first close” language carefully. A first close means additional investors can still come in before the round formally closes, the final total could exceed $1 billion. Watch whether any of the T. Rowe Price or Capital Group participation represents institutional asset management conviction parallel to what Vanguard and Blackstone showed in the Norm Ai round: buy-side validators carrying real AI deployment weight. And watch whether JPMorganChase issues any independent statement confirming the SambaNova relationship, that confirmation, if it comes, upgrades the vendor announcement to a joint validation with significant downstream sales effect.
What to Watch
TJS synthesis
The real story is that “first close at $1B” is doing two things simultaneously: it signals capital demand strong enough that SambaNova could close the round in stages rather than all at once, and it leaves room for the final number to be higher. General Atlantic’s lead isn’t a passive financial bet, they’re a growth equity firm that takes board seats and drives commercial execution. Watch whether SambaNova announces a second close within 90 days with additional sovereign AI or financial services anchor customers. If JPMorganChase is one reference customer, the question isn’t whether the model works, it’s whether SambaNova can replicate that relationship across four or five more tier-one financial institutions before Nvidia responds with its own on-premises inference stack.
Sources: TechCrunch, SambaNova.