Pomegra
AI infrastructure startup Aranya closed $11M in seed funding led by First Round Capital to automate bare-metal GPU cluster deployments for inference providers.
- Aranya raised a $9M seed led by First Round Capital, plus an earlier $2M pre-seed, bringing total raised to $11M.
- The company's software converts bare-metal servers into production-ready GPU clusters in under 48 hours.
- Aranya already manages more than $500M in GPU hardware for AI inference providers and data centers, despite operating for less than a year.
Lead
Aranya, an AI infrastructure startup founded less than a year ago, announced on September 1, 2026 that it has raised $11 million across two rounds to automate the provisioning of bare-metal servers into production-ready GPU clusters. The $9 million seed was led by First Round Capital, with participation from BoxGroup, Vermilion Cliffs, and Asylum Ventures. An earlier $2 million pre-seed was led by Asylum Ventures, with contributions from Founder Collective, Parable VC, and Uncommon Ventures. Valuation was not disclosed.
The company was co-founded by Christian Bhatia Ondaatje, Sasivarnan Kanaghasalam Sathyapriya, and Aryamika Bhatia Ondaatje, and sits between raw data center hardware and the AI companies that consume it for inference and training workloads.
What Does Aranya Actually Do?
Aranya's core product is an open-source engine called clusterdOS, built on Kubernetes, which deploys and maintains GPU infrastructure through declarative configuration files. A bare-metal server rack can become a production-ready cluster in under 48 hours, eliminating the manual configuration work - networking, storage, drivers, monitoring, security hardening - that typically consumes weeks of engineering time.
The company says it already manages more than $500 million worth of GPU hardware. That number is striking for an organization operating for less than 12 months, and positions the funding round less as a proof-of-concept milestone and more as an acceleration of something already underway. Clients are predominantly AI inference providers and data centers, segments where compute availability directly determines revenue.
Why Is Bare-Metal Provisioning Still a Problem in 2026?
The GPU supply crunch has pushed a growing share of AI companies toward bare-metal contracts rather than hyperscaler cloud instances. Bare metal offers better performance-per-dollar and avoids the virtualization overhead that public clouds bundle in by default - but it arrives without the orchestration tooling those same clouds provide.
Inference providers in particular operate on thin margins and cannot absorb weeks of setup latency without losing customers to faster competitors. Aranya's bet is that clusterdOS eliminates that delay at a price point that makes outsourcing the provisioning problem rational.
The broader dynamic is not subtle. As Nvidia hardware allocation queues stretch into months, whoever controls the fastest path from delivery dock to production workload controls a genuine bottleneck. That bottleneck will not be permanent. Cloud providers are expanding bare-metal offerings, and established infrastructure-as-code platforms are extending into GPU territory. The window is open now; Aranya is moving to establish itself before it narrows.
Competitive Context
Aranya operates in territory that adjacent players have circled for years. Managed bare-metal providers offer hardware without the software complexity but typically bundle proprietary management tools and create lock-in. Infrastructure platforms like Terraform and Pulumi handle configuration but were not designed for GPU cluster lifecycle management at scale. The startup cohort targeting GPU orchestration has grown since 2023, though most focus on scheduling workloads on existing clusters rather than building the clusters from scratch.
Aranya's stated differentiation is the combination of deployment speed and an open-source engine designed to avoid proprietary dependency. clusterdOS being open-source is a deliberate posture - it lowers the adoption barrier for inference providers that are sensitive to vendor risk given how much capital is tied up in the hardware itself.
Investor Thesis
First Round Capital leading a seed of this size reflects conviction in the infrastructure layer beneath AI applications - a tier that received comparatively little venture attention during the application-layer frenzy of 2024 and 2025. Asylum Ventures participating in both the pre-seed and the seed indicates continuity from the earliest backers, a signal worth noting when the company's traction figures are this early-stage.
The $11 million total is modest relative to the hardware value Aranya already manages. That $500 million figure implies the company's software is trusted with significant customer assets, which functions as both a marketing claim and an implicit argument for the undisclosed valuation. Whether the assets under management translate into durable software revenue - rather than one-time provisioning fees - is the question this round will help answer.
Outlook
Aranya enters the next phase with a small war chest and an outsized implied customer footprint. The $9 million seed will need to cover engineering headcount, go-to-market expansion, and continued development of clusterdOS. The immediate priority is converting early hardware management contracts into defensible software relationships with recurring revenue characteristics. The near-term demand environment from inference providers running bare-metal fleets is favorable. The longer-term challenge is the infrastructure layer attracting capital quickly from larger players with more resources. Execution speed - the same quality Aranya sells to its customers - will determine whether the company leads or becomes a footnote in the GPU cluster automation story.
