Impala AI emerged from stealth on Wednesday with $11 million in seed funding led by Viola Ventures and NFX to develop a new AI stack for large language model inference. The platform is designed to help enterprises run AI at unlimited scale while cutting costs and maintaining control. The funds will be used to expand the team, enhance product capabilities and accelerate adoption among enterprise AI builders.
Led by CEO Noam Salinger, a former Granulate executive, Impala is pioneering an inference platform that enables enterprises to run AI directly within their own virtual private cloud environments. The system offers a serverless experience while managing GPU capacity without compromising control or flexibility.
At the center of Impala’s technology is a proprietary inference engine built to handle large-scale model deployment. Its first use case focuses on data processing, achieving up to a 13-fold reduction in cost per token on unmodified models without rate limits or capacity constraints.
“We’re at the dawn of a new era in AI, where inference will drive the next wave of innovation,” Salinger said. “We’re not just building another platform but the infrastructure that will power AI at real-world scale. Our mission is to empower teams to unlock the full potential of their models more efficiently than ever before.” As AI infrastructure demand continues to exceed supply, enterprises face growing challenges in running inference workloads efficiently. Impala aims to address these bottlenecks by expanding GPU capacity beyond the limits of current providers.
With the AI market shifting from model training to inference as the main operational cost, Impala positions itself to capitalize on the trend. Industry analysts project the inference market will reach $106.15 billion by 2025 and grow to $254.98 billion by 2030. Technology research firm Canalys noted that unlike training, which is a one-time expense, inference represents a recurring operational cost that can constrain AI commercialization.
“Our vision is to make inference invisible,” Salinger said. “When teams connect Impala to their cloud, they shouldn’t have to think about provisioning, scaling or optimizing GPU clusters. We handle that behind the scenes so they can focus on building products.”
Impala is already working with Fortune 500 companies and plans to expand its customer base globally.
“The demand for AI is accelerating, yet enterprises are held back by the cost and complexity of scaling it,” said Alex Shmulovich, partner at Viola Ventures. “Impala makes large-scale adoption seamless by cutting costs, protecting data and reducing friction while maintaining flexibility.”
“Inference is where the real battle for AI adoption will be won,” said Sarai Bronfeld, partner at NFX. “Impala offers a scalable, cost-effective way to bring models into production. The team is building the backbone of the inference economy.”
About Impala: Impala AI is an inference platform designed for large-scale language model usage. It delivers up to 13 times lower cost per token compared to current platforms while maintaining enterprise-grade reliability. The company’s multi-cloud, multi-region solution deploys directly into a customer’s virtual private cloud, ensuring full control over data, spending and cloud provider choice.