AI hardware startup Tsavorite Scalable Intelligence has secured over $100 million in pre-orders from enterprises and cloud providers across the U.S., Asia, and Europe to deploy its new class of AI acceleration chips designed to scale intensive workloads. The announcement reflects growing global demand for efficient and cost-effective compute architectures that can handle increasingly complex AI workflows.
Founded in 2023 by former Intel engineers and semiconductor veterans, Tsavorite has developed the Omni Processing Unit (OPU) — a new compute architecture that merges CPU, GPU, memory, and connectivity into a single integrated system. This composable approach enables developers to dynamically configure hardware resources for a wide range of AI applications, from edge devices to large-scale data centers.
“We’ve built the first truly composable, developer-friendly AI platform that delivers step-change gains in efficiency, cost, and scale — from edge to hyperscale,” said Shalesh Thusoo, CEO of Tsavorite.
Composable Compute for the Next Wave of AI Workloads
AI workloads are becoming increasingly data-intensive and distributed, driving the need for architectures that optimize power efficiency, scalability, and total cost of ownership. Tsavorite’s OPU architecture allows components to be shared or isolated as needed, eliminating bottlenecks between memory and compute units.
The company’s chips are being fabricated on Samsung’s SF4X platform, a 4-nanometer process node optimized for high-performance computing. Tsavorite said the first enterprise-grade AI appliances supporting agentic and multi-modal AI workflows will be available by next year, aimed at customers seeking scalable, low-latency infrastructure for model deployment and inference.
Industry analysts say composable compute solutions like Tsavorite’s could help enterprises rationalize infrastructure costs as AI models continue to grow in size and complexity. By integrating multiple compute types into a single unit, companies can allocate resources more efficiently, particularly in multi-tenant or hybrid cloud environments.
AI Hardware Race Heats Up
The launch comes as global tech giants and startups alike race to address the AI infrastructure bottleneck. With GPU shortages and escalating data center energy demands, enterprises are exploring alternatives that deliver comparable performance with better energy and space efficiency.
Tsavorite’s pre-orders underscore confidence in the company’s ability to deliver production-ready silicon at scale, despite not disclosing its valuation or total funding to date. Its leadership, drawn from Intel, NVIDIA, and ARM alumni, positions the company as a contender in the fast-growing AI hardware ecosystem.
By combining programmability, integration, and efficiency, Tsavorite aims to challenge traditional accelerator paradigms and establish its OPU as a mainstream choice for developers and cloud operators looking to balance performance with sustainability.
