Hosted.ai Raises $19M Seed for GPUaaS Platform
Hosted.ai, a San Jose, Costa Rica-based provider of multi-tenant GPU-as-a-Service (GPUaaS) software, has raised $19M in seed funding led by Creandum. The platform uses software-defined GPU pooling, orchestration, and monetization tools to enable cloud service providers (CSPs), telcos, and hosting providers to achieve 100% GPU utilization with 2x-10x overcommit and 5x higher profits per GPU compared to traditional passthrough models. The capital will expand engineering and customer support teams while scaling the full hyperconverged IaaS stack integrating GPUaaS with VMs, storage, and networking.
GPUaaS Contracts Explode Amid GTC Buzz
The timing aligns with surging demand for GPU infrastructure. Duos Technologies secured a $176M GPUaaS contract on March 13, 2026, its largest deal ever projecting over $50M annual revenue. Meta signed a $27B deal with Nebius for GPUaaS capacity the same week as Hosted.ai's announcement following NVIDIA GTC 2026, where GPU virtualization emerged as a key theme. Hosted.ai's software addresses this by empowering tier-2 providers to monetize idle hardware without massive CAPEX.
AI Workloads Leave GPUs 60% Idle
AI model training and inference workloads typically leave GPUs idle 60% of the time due to static resource allocation. Service providers struggle with low utilization rates of 20-40%, limiting profitability amid hardware shortages. Current solutions like passthrough models fail to enable multi-tenancy, forcing providers to underutilize expensive assets. This waste problem, not scarcity, hampers neocloud builders serving AI developers and enterprises.
Software Pooling Unlocks 5x Profits
Hosted.ai's turnkey control panel orchestrates multi-tenant GPU sharing like VMware for CPUs, with dynamic scheduling and overcommit sliders balancing performance and security. Providers gain tools for billing, budget controls, and auto-shutdown, pushing utilization to 80-90%. Early metrics from packet.ai dashboard show 22% visit-to-signup conversion, $172.74 ARPU, and 100% retention.
As CEO Ditlev Bredahl noted:
"The GPU infrastructure powering AI today is incredibly wasteful. Customers are forced to rent GPU as a static resource… AI workloads leave about 60% of that GPU idle, on average."
Creandum Validates OnApp Team Reunion
Creandum, with AI infra bets like Cast AI and Modal, led the round alongside Repeat VC, People Ventures, Z21 Ventures, Golden Sparrow Ventures, Hersir Ventures, and Tekton Ventures. The investor mix signals conviction in Hosted.ai's GPU economics for CSPs, backed by existing investors following on. Creandum's pattern of early-stage deep tech plays underscores strategic validation for Hosted.ai's neocloud focus.
GPUaaS Market Grows to $26B by 2030
The GPUaaS market stands at $8.21B in 2025, projected to reach $26.62B by 2030 at 26.5% CAGR, per MarketsandMarkets. Trends like AI inference demand and sovereign cloud needs drive this expansion. Competitors like CoreWeave ($7.5B+ raised) focus on massive clusters, while Lambda Labs ($320M) and Runpod ($20M+) offer on-demand pods. Hosted.ai differentiates as software-only for existing hosts.
OnApp Exit Team Scales GPUaaS
Founders Ditlev Bredahl (CEO), James Withall (COO), Julian Chesterfield (CTO), and Narendar Shankar (CCO) reunite from OnApp, the cloud management platform acquired by Virtuozzo in 2021 after raising $25M+. Bredahl's serial exits and investor role via KickAss Capital add domain depth for CSP pain points. Their track record directly translates to Hosted.ai's GPU orchestration for telcos and hosters.
Neocloud Launches Target Tier-2 CSPs
Hosted.ai plans to launch "Neocloud in a Box" at CloudFest 2026, offering zero-CAPEX trials for full-stack GPUaaS. Partnerships like Maerifa Solutions combine hardware financing with software. Expansion includes engineering hires, NVIDIA GTC presence, and FuriosaAI NPU integration to support global neocloud growth amid Big Tech's $600B+ 2026 AI capex.