New serverless orchestration layer turns Vast.ai's global GPU platform into an on‑demand efficiency engine for every AI team.
LOS ANGELES, Dec. 10, 2025 /PRNewswire-PRWeb/ -- Vast.ai, the leader in GPU rental platforms, today announced Vast Serverless, the industry's first serverless system that can auto benchmark, optimize and predict GPU load across multiple auto-groups. Vast.ai's Serverless transforms the GPU cloud into a self-optimizing fabric by auto-benchmarking instances across multiple GPU groups and intelligently routing workloads in real time to deliver the best performance per dollar.
As AI teams look beyond traditional clouds for more cost‑effective compute, Vast.ai has grown into one of the world's largest and most affordable GPU clouds. Vast.ai has assembled a global marketplace of 17,000 GPUs from 1,300 independent providers, providing customers with access to a diverse fleet of consumer and data-center GPUs across hundreds of locations. More than 123,000 customers now use Vast.ai to run training and inference workloads, often at a fraction of the cost of conventional GPU cloud options.
Vast & Vast Serverless at a Glance
- 17,000+ GPUs
- ~123,000 paying customers
- 68 GPU types and 50+ hardware parameters to choose from
- 1,300+ providers across a globally distributed fleet
- SOC 2–compliant with an optional "Secure Cloud" mode
"Vast Serverless changes the game in cost‑effective compute," said Jake Cannell, CEO of Vast.ai. "It automates the provisioning and orchestration of the most price‑efficient GPUs from a globally distributed supply, matching your code to the machines that have the best performance per dollar at that moment in time. Any AI team with functions or models to deploy can tap into this global, liquid market for GPU compute instead of wrestling with fixed, expensive cloud instances."
Vast Serverless is a container‑native orchestration layer that abstracts away server selection into a few key parameters developers control. As workloads grow, Vast's market engine automatically benchmarks, ranks, and selects the optimal GPUs from a heterogeneous fleet, including highly cost‑effective consumer GPUs, so teams get maximum throughput for their budget.
"We needed to enrich 100,000 documents every two hours using LLMs, something that was prohibitively expensive on other clouds," said Anna Bosch, VP of Data Intelligence at Launchmetrics. "With Vast Serverless, we scaled up to 46 H100 servers on demand and completed the job in just 38 minutes at a quarter of the cost. It enabled us to move to production with confidence."
Key Advantages of Vast Serverless:
- Always‑on price/performance optimization: Vast Serverless continuously benchmarks GPU instances while workloads are running, constantly moving jobs to the most efficient machines for the task. Combined with predictive scaling, the system automatically optimizes for the best performance per dollar.
- Enterprise‑grade security on open infrastructure: Vast.ai is SOC 2 compliant. For customers with stricter requirements, Vast Secure Cloud ensures workloads run only in data centers.
- Radical pricing transparency: Developers leverage the world's most cost‑effective GPU rental platform with simple, straightforward pricing. There are no hidden premiums or special tiers for specific GPU models or worker types—customers pay the same price no matter how they provision instances.
- Unrestricted selection and control: Unlike providers that offer a handful of fixed SKUs, Vast Serverless exposes full marketplace flexibility. Teams can filter across 68 GPU types and more than 50 hardware parameters, including reliability, network speed, and CPU RAM. Workloads can be deployed across a globally distributed fleet of more than 500 locations, reducing latency and improving resilience.
- Universal workload support and real debugging: Vast Serverless is engineered for the full spectrum of GPU‑intensive tasks. Users can start from pre‑built autoscaler templates for popular frameworks like TGI and ComfyUI to run LLM, image, video, and TTS workloads. Developers retain access to real debugging tools such as logs, metrics, and direct Jupyter/SSH access, so they can troubleshoot quickly instead of treating infrastructure as a black box.
"Vast Serverless gives developers the best of both worlds," Cannell added. "You keep full control over the kind of compute you want, but our market algorithms handle sourcing, ranking, and provisioning the best machines globally. It's the next step in democratizing AI: turning the world's GPU capacity into a single, efficient, liquid market that anyone can tap into."
About Vast.ai
Founded in 2018, Vast.ai is the leading cloud GPU platform for AI and machine learning workloads. The service connects data centers and professionals running the Vast hosting software with users who can quickly find the best deals for AI compute according to their specific requirements. With GPU rental prices 3–5x lower than other cloud providers, Vast.ai offers a cost-effective solution for scalable AI and ML infrastructure. For press related inquiries, contact Anne Ward at [email protected]
Media Contact
Anne Ward, Vast.ai, 1 (323) 483-5256, [email protected], https://vast.ai/
SOURCE Vast.ai
Share this article