Positron announced an $875 million financing package on September 10 to develop its next generation of AI inference hardware. Inference is the work of running an already trained model to answer a request. The company sells systems for cloud providers and other organizations serving those requests, rather than training a consumer chatbot of its own.
The workflow starts with a trained model. Positron’s product site describes uploading or linking a model file, then connecting an application to a compatible software endpoint to request its output. Atlas, its existing server product, packages accelerators, memory and inference software into a system customers can deploy. The company’s proposition is to change the hardware underneath an AI service without making its users rebuild the entire application.
Memory is central to the next design. A serving system must keep model information and working context available as it generates answers; more computing alone does not eliminate the problem of moving that information. Positron is developing Asimov chips and Titan systems around substantial memory capacity. These are future products: the latest announcement targets chip-design completion at the end of 2026 and production in the second half of 2027.
The company says it is deploying more than 50 racks of Atlas at Oracle Cloud Infrastructure, with partner Parasail using that capacity for its inference service. It also names Jump Trading and i3d.net as production customers. Those are company-reported deployment facts, distinct from the simulations used in comparisons for its future chips. A forecast for Asimov is not measured performance from a Titan system already in a customer’s data center.
The announced package comprises a $375 million Series C and a Series C-1 of up to $500 million, with a stated $5 billion post-money valuation. NEA, Atreides Management, Valor Equity Partners, Andra Capital, SemiAnalysis Capital and Jim Clark are among the lead investors across the financing. FinSMEs also reported the raise. The proceeds support chip development, engineering infrastructure and the production ramp; the up-to component should not be read as a separate completed raise.
Positron’s next test is carrying experience from deployed servers into a new chip and a manufacturable system. Customers need more than favorable modeled economics: hardware must arrive, run their models reliably and fit the power and cooling available in their facilities. That transition from design to dependable deployment is the business task the financing is meant to support.