Hosted.ai Raises $19M Seed for GPU Efficiency

Hosted.ai has successfully closed a $19 million Seed funding round, signaling a significant injection of capital aimed at revolutionizing how Graphics Processing Units (GPUs) are utilized within the burgeoning artificial intelligence sector. The investment was spearheaded by Creandum, with crucial participation from People Ventures and Repeat VC. This funding will accelerate the deployment of Hosted.ai's innovative GPU-as-a-Service (GPUaaS) technology, designed to dramatically increase the output from existing hardware for 'neocloud' service providers.

The current AI infrastructure build-out, while unprecedented in scale and speed, faces a critical bottleneck in GPU compute accessibility and efficiency. Traditional models compel businesses to lease entire server units, even when only a fraction of their processing power is required. This inefficiency disproportionately impacts neocloud operators, who often find their substantial hardware investments underutilized, with average utilization rates hovering around 40%. This idle capacity translates directly into compounding financial losses, a challenge Hosted.ai aims to resolve.

Founded by a seasoned team with over 15 years of collective experience across multiple infrastructure generations, Hosted.ai's platform acts as an intelligent intermediary. It dynamically manages GPU memory in real-time, enabling multiple distinct AI workloads to share the same physical GPU simultaneously without compromising performance. This sophisticated orchestration is key to unlocking latent capacity within data centers.

For infrastructure operators, achieving higher utilization through Hosted.ai's solution means their existing hardware can support a significantly larger volume of AI tasks. This enhanced efficiency is projected to bring operators, many of whom operate on thin or negative margins, to profitability within a 12-month timeframe, crucially without necessitating further capital expenditure on new hardware. The ultimate goal is to enable AI developers to access GPU resources on a consumption-based model, leading to substantial cost reductions.

Beyond economic benefits, Hosted.ai's technology is poised to foster greater diversity in the global AI infrastructure market, which is currently heavily concentrated among a few major US hyperscalers. This is particularly relevant for European entities, including businesses and public sector organizations, grappling with data sovereignty regulations. By enabling more efficient regional infrastructure, Hosted.ai offers a viable alternative to relying on offshore providers. Furthermore, its federated GPU marketplace allows service providers to offer their surplus capacity and seamlessly tap into resources from other participants, effectively extending their reach without the burden of direct ownership.

Ditlev Bredahl, CEO of Hosted.ai, emphasized the inherent waste in current GPU provisioning, stating, "Customers are compelled to rent GPU as a static resource, as that's the only way operators can monetize it. Consequently, AI workloads leave approximately 60% of that GPU capacity dormant. Our platform leverages dynamic memory management to grant workloads access to the full GPU potential, intelligently optimizing task placement to maximize the utilization of every GPU in the data center. This results in elastic, efficient GPU compute that fundamentally transforms the economics for both operators and their clients."

Carl Fritjofsson, Partner at Creandum, highlighted the exceptional founder-market fit, noting, "Ditlev and his team have a proven track record, having successfully built similar solutions for the dot-com and cloud eras. Their unparalleled technical expertise positions them perfectly to lead in the GPU era, a market anticipated to dwarf its predecessors in scale."