- SambaNova Systems announced a partnership with Emergys on October 29, 2026, adding professional and managed services for agentic AI deployments, a deliberate move beyond hardware sales into the operational layer where most enterprise pilots stall.
- SambaNova’s SN50 RDU, unveiled in February 2026, will be hosted in Oracle Cloud Infrastructure data centers following an October 27, 2026 agreement, giving OCI customers an alternative to Nvidia GPUs for high-throughput inference.
- JPMorganChase selected SambaNova as an inference infrastructure partner in July 2026, deploying SN40 and SN50 systems on-premises, the kind of regulated-sector win the Emergys partnership is designed to replicate at scale.
SambaNova Systems closed a $1 billion Series F at an $11 billion valuation in July 2026, then spent the following weeks announcing the partnerships that explain why: two deals in three days that reframe the company from chip vendor to full-stack inference provider. The gap SambaNova is betting on is not performance, it’s deployment.
What the SN50 Actually Does
SambaNova Systems, founded in 2017, builds custom Reconfigurable Dataflow Units (RDUs) rather than adapting general-purpose GPUs for AI workloads. The SN50, its fifth-generation chip unveiled in February 2026, targets agentic workloads specifically, pipelines that chain reasoning steps and tool calls, where token latency compounds across each hop. SambaNova claims the SN50 runs up to five times faster than competing chips, with a total cost of ownership up to three times lower, though independent benchmarking has not confirmed those figures across diverse workloads. The architecture supports linking up to 256 accelerators and models exceeding 10 trillion parameters. At an average of 20 kW per rack, it also fits within the power envelope of existing air-cooled data centers, a practical constraint that matters as much as raw throughput for enterprises planning capacity.
Inference Demand Behind the Pivot
Agentic AI systems, multi-step reasoning pipelines running at production scale, put different pressure on infrastructure than single-turn LLM calls. Token generation speed and per-token cost become the deciding variables when a system is fielding millions of agent interactions rather than thousands of chat completions. SambaNova co-founder and CEO Rodrigo Liang has said agents are driving the next wave of inference demand, requiring fast, consistent tokens at a viable cost. The SN50’s architecture is SambaNova’s answer to that constraint, and the Oracle and Emergys deals are the distribution mechanism.
Oracle Cloud as the Distribution Channel
On October 27, 2026, SambaNova announced it would deploy SN50 inference chips inside Oracle Cloud Infrastructure data centers, with Oracle hosting SambaRack SN50 systems for customers needing high-throughput inference at scale. The move expands OCI’s AI infrastructure options beyond Nvidia GPUs, relevant for customers who have hit capacity or cost ceilings on GPU-based inference. Oracle has been active in positioning OCI for agentic workloads, and adding SambaNova’s RDU capacity extends that without requiring Oracle to build its own custom silicon program.
Deployment Expertise Becomes the Bottleneck
Two days after the Oracle announcement, SambaNova disclosed a partnership with Emergys, a digital transformation services firm, to provide professional services, managed services, AI platform operations and customer success support to SambaNova clients. The target markets are enterprise, regulated industries and sovereign AI deployments, precisely the environments where procurement is the easy part and operationalisation is where projects stall. Emergys covers model onboarding, capacity planning, monitoring and round-the-clock service operations. For customers in data-sensitive sectors, that operational wrapper also addresses sovereignty and security requirements that pure cloud deployments cannot meet.
Where the Customer Base Stands
JPMorganChase selected SambaNova as an inference infrastructure partner in July 2026, deploying SN40 and SN50 systems for secure on-premises AI inference, a reference deployment that carries weight in regulated-sector sales conversations. OVHcloud, which chose SambaNova for its AI Endpoints service in November 2025, represents the European in-country deployment pattern: customers who need performance but cannot route data through US-based hyperscaler infrastructure. SambaNova’s product stack is built around these use cases. SambaCloud serves developers; SambaStack and SambaManaged target operators running private inference clouds; SambaRack handles on-premises builds. The Emergys partnership layers services across all of them.
The Funding Behind the Strategy
The Series F first close in July 2026 was led by General Atlantic, following a Series E of over $350 million in February 2026, the same month the SN50 launched. Intel Capital holds roughly 9% of SambaNova after a planned $15 million investment in April 2026. The company also beat its sales plan for the fiscal year ending January 2026. That combination, strong sales execution, two rounds of funding inside six months, and an $11 billion valuation, gives SambaNova the runway to build out the services layer the Emergys deal represents, rather than competing on chip specs alone. Inference cost pressure across the enterprise market makes that services layer increasingly the purchase criterion.
