Aranya Raises $11M for AI GPU Cluster Operations

Aranya has announced $11M in financing to expand the operating layer that turns bare-metal GPU servers into production-ready AI clusters. The package includes a $9M Seed round led by First Round Capital and an earlier $2M Pre-Seed led by Asylum Ventures.

The San Francisco company is building around a practical infrastructure gap. Buying GPUs gives an AI company or data center valuable hardware, but it does not complete the networking, storage, orchestration, observability, scheduling, and operational work required to run inference or training reliably. Aranya says its software can compress that conversion into less than 48 hours.

That distinction explains why the financing matters beyond another AI-infrastructure round. Capital spending has put more accelerators into more facilities, while the usable output of those machines still depends on teams that can make heterogeneous hardware behave like a coherent system. Aranya is betting that the control layer between a rack and a workload becomes its own durable market.

What Aranya Announced

Aranya's September 1 announcement describes $11M across 2 rounds. First Round Capital led the $9M Seed, with BoxGroup, Vermilion Cliffs, and Asylum Ventures participating. Asylum Ventures led the earlier $2M Pre-Seed, joined by Founder Collective, Parable VC, and Uncommon Ventures.

The accounting deserves one clean distinction. Aranya announced a $9M Seed rather than a new $11M Seed. The larger headline combines that round with the prior Pre-Seed. A July 28 Form D reports $11,310,983 sold in a $12,010,973 offering and says the amount includes conversion of previously issued convertible securities. The filing and announcement describe related capital through different reporting conventions.

Aranya did not disclose its valuation, revenue, customer identities, or individual investor allocations. The company says it will use the financing to expand engineering and go-to-market hiring, serve a growing customer base, and launch an AI-native interface for operating multiple clusters.

The Product Begins After the GPU Purchase

Aranya's open-source engine, clusterdOS, is built on Kubernetes and uses a GitOps model to coordinate the software around a GPU cluster. The company's published stack includes networking, storage, monitoring, inference tooling, virtual machines, and Slurm environments alongside Kubernetes workloads.

That collection matters because a data center rarely hands an AI company a standardized appliance. Facilities differ in networking, storage, hardware generations, firmware, operating practices, and failure modes. A team can own enough compute for a serious workload and still spend weeks turning those machines into an environment that developers can use without an infrastructure specialist standing nearby.

Aranya's proposition is to absorb that heterogeneity. The company specifies requirements with the facility, configures the operating stack for the workload, and manages clusters through a common multicluster layer. In plain business terms, the customer is buying less time between capital expenditure and productive compute.

The Evidence Aranya Has Put Forward

Before its first anniversary, Aranya says it is managing more than $500M worth of GPUs. A company case study with Hydra Host reports a joint footprint of more than 1,700 GPUs, production-ready deployments in 24 to 48 hours, 99.99% Kubernetes control-plane uptime, and a 90% reduction in cluster issue-resolution time.

Those figures are useful operating evidence, but they remain company-reported. Aranya has not published an independent audit, a broad customer cohort, or enough economic detail to compare total operating cost against hyperscaler services, generic managed Kubernetes, or a customer's internal platform team. The numbers establish the shape of the claim; a larger deployment base will determine how portable the result is.

That is the right boundary for an early infrastructure company. A 1,700-GPU footprint is more meaningful than a clean laboratory demo, while repeatability across different facilities, workloads, and customer teams is the evidence that turns an implementation advantage into a platform.

Why Inference Changes the Infrastructure Sale

Training attracted the first giant clusters because model development concentrated huge jobs in relatively controlled environments. Inference spreads the operational burden. Production traffic changes, latency matters, capacity must be available when users arrive, and hardware can sit across providers or facilities with different economics and technical constraints.

The result is a new handoff problem. Data centers can supply bare metal. AI companies can supply models and demand. Somebody still has to turn the hardware into a dependable service, then keep the fleet visible and usable when components fail or workloads move.

Aranya is trying to own that handoff. Natural-language controls and proactive remediation may make the interface more accessible, but the deeper product is operational consistency. A polished command surface has value only when the system beneath it diagnoses the right failure, respects permissions, and changes infrastructure without creating a new incident.

The Team Behind Aranya

Aranya was founded in 2025 by Christian Bhatia Ondaatje, Sasivarnan Kanaghasalam Sathyapriya, and Aryamika Bhatia Ondaatje. Christian Bhatia Ondaatje serves as co-founder and CEO. The Form D lists all 3 founders as executive officers and directors, while public source material does not consistently establish precise current C-suite titles for the other 2 founders.

The company's current hiring plan reflects the work ahead. Aranya is recruiting around platform engineering, frontend development, DevOps and site reliability, marketing, and developer relations. That mix suggests the next phase is not only deeper infrastructure automation. It is also productizing the interface, supporting deployments, and building an open-source community that can widen adoption.

What the $11M Changes

Aranya now has capital to move from promising deployments toward a repeatable operating system for GPU fleets. The $9M Seed gives the company room to hire the engineers who can handle facility-specific complexity while making the customer experience feel less bespoke. The earlier $2M Pre-Seed funded the foundation that reached the current announcement.

The market question is moving with the same hardware. Scarcity once made GPU access the headline. As more capacity reaches data centers, utilization, reliability, and time-to-production become harder to hide behind rack counts. The winning infrastructure provider may be the one that helps customers convert expensive inventory into dependable work before the next generation of chips changes the configuration again.

Aranya is entering that race with a clear promise: any bare metal, configured for the workload, production-ready in under 48 hours. The financing gives the company more people and product surface to carry that promise. Each new facility will show whether the operating layer can travel as quickly as the capital buying the machines.