SambaNova Systems is launching a $3.5 billion [1] AI cloud designed to reduce the cost of artificial intelligence for enterprise customers.
Lowering the cost of AI inference is critical for businesses attempting to scale operations. High operational expenses currently act as a barrier to the widespread deployment of AI agents across corporate workflows.
Rodrigo Liang, CEO of SambaNova Systems, said that inference represents the most significant cost challenge facing enterprise AI today [1]. To address this, the company is implementing a disaggregated architecture. This system separates GPUs, CPUs, and specialized AI processors to optimize performance and reduce spending [1].
The new cloud initiative is backed by Vista Equity Partners and Cambium Capital [1]. By shifting away from traditional monolithic hardware setups, SambaNova aims to make the technology more affordable and speed the adoption of autonomous AI agents [1].
Separate from the cloud launch, Intel has increased its investment in SambaNova by $15 million [2]. This additional funding brings Intel's total ownership stake in the company to about nine percent [2].
Liang said the goal is to move beyond the current expensive infrastructure models to enable more sustainable AI growth [1]. The company believes that specializing the hardware layer is the only way to achieve the efficiency required for next-generation enterprise applications [1].
“Inference is the biggest cost challenge in enterprise AI.”
The shift toward disaggregated architecture suggests a move away from the industry's reliance on general-purpose GPUs. By decoupling processing components, SambaNova is attempting to solve the 'inference bottleneck,' which could lower the financial entry point for companies deploying AI agents at scale.
