Hosted.ai secures $19M to tackle GPU efficiency bottlenecks
Hosted.ai, a new infrastructure startup founded by former Nvidia engineers, has raised a $19 million funding round led by European venture firm Creandum. The company is building software to dramatically improve GPU efficiency for large-scale AI workloads, a critical pain point for cloud providers, enterprises and AI labs racing to deploy advanced models.
Ex-Nvidia talent targets AI infrastructure pain points
The founding team behind Hosted.ai brings deep experience in GPU architectures, high‑performance computing and large‑scale AI training from their time at Nvidia. By focusing on how GPUs are orchestrated, scheduled and utilized across clusters, the startup aims to help customers squeeze more performance out of scarce and expensive hardware.
At the core of its platform, Hosted.ai is developing advanced resource orchestration and workload optimization tools that can reduce idle time, balance heterogeneous GPU fleets and prioritize the most critical machine learning jobs. This promises higher throughput and lower unit costs for training and inference, without requiring customers to redesign their models.
Why GPU efficiency matters for AI economics
As demand for generative AI and large foundation models surges, the global shortage of high‑end GPUs has become a defining constraint for the industry. Cloud platforms and AI‑first startups are spending heavily on AI infrastructure, but underutilization of clusters can erode margins and slow product rollouts.
By improving GPU utilization rates, Hosted.ai wants to shift the economics of AI deployment. Better efficiency can translate into faster experimentation cycles, more predictable capacity planning and reduced reliance on continuously buying new hardware.
Creandum backs next wave of AI infrastructure
Lead investor Creandum, known for early bets on European and global technology leaders, is positioning the round as a strategic move into the AI infrastructure layer. The fresh capital will allow Hosted.ai to expand its engineering team, deepen integrations with major cloud providers and scale pilots with early enterprise customers.
As competition intensifies to control the stack beneath modern AI applications, the success of Hosted.ai will be a key test of how much value can be unlocked through smarter software rather than simply more silicon.