Emerging AI hardware company Tsavorite Scalable Intelligence announced it has locked in more than $100 million in pre-orders from enterprise clients and cloud providers across the United States, Asia, and Europe.

Founded in 2023 by veteran chip-industry engineers, Tsavorite is developing its signature product, the “Omni Processing Unit” (OPU), a unified architecture combining CPU, GPU, memory, and connectivity in a single module. The company says this design is optimised for the growing demands of AI workloads, offering scalability, lower power consumption and adaptable deployment.

The pre-orders reflect a broader wave of interest in specialised AI infrastructure as organisations of all sizes seek hardware that can handle complex workflows from edge devices to hyperscale data centres. Tsavorite will manufacture its OPU using Samsung’s SF4X platform, and expects delivery to begin next year.

Chief Executive Officer Shalesh Thusoo said the company’s goal is to “deliver step-change gains in efficiency, cost and scale from edge to hyperscale,” underscoring the ambition to serve a wide spectrum of customers from AI model developers to large cloud operators.

While the $100 million figure is substantial for a startup at this stage, Tsavorite has not revealed its current valuation or full fundraising history. That lack of transparency adds to the intrigue around the company’s strategy and growth trajectory.

Analysts see this development as an indicator of how AI hardware demand is fragmenting away from general-purpose chips to more tailored solutions. The OPU’s composable architecture suggests a shift toward flexibility: customers can reconfigure hardware to match specific AI tasks rather than rely solely on existing designs built primarily for graphics or traditional compute.

At the same time, questions remain about production risk, supply-chain constraints and how Tsavorite will compete in a market dominated by major players with deep fabrication and ecosystem advantages. Manufacturing via Samsung’s platform gives credibility, but scaling to fulfil large orders and delivering on performance claims will be true tests.

For businesses deploying AI at scale, whether in cloud data centres, autonomous systems, or real-time analytics, the promise of lower power consumption and higher workload efficiency is compelling. Tsavorite’s early traction suggests that demand is real. If the company can deliver on its roadmap, it may help reshape how the industry defines and purchases “AI infrastructure.”