Customers will run the largest open-weight models on Titan through the same Parasail API, committing to output instead of hardware
Parasail, the first multi-silicon inference cloud, today announced it will deploy 5 megawatts of Positron AI's next-generation Titan inference systems. Parasail expects the Titan capacity, which comes online in late 2027, to serve trillions of tokens a day for companies building on frontier-scale open-weight models. The deal expands a partnership in which Positron’s first-generation Atlas systems already serve Parasail customers.
Titan runs multi-trillion-parameter models on low-power LPDDR5X, a commodity DRAM, in place of the high-bandwidth memory GPUs rely on. Compared to traditional GPUs, Titan delivers 2.4x the tokens per dollar, or, at the same cost per token, 4.1x the speed. Customers will reach Titan through the same Parasail account and OpenAI-compatible API they already use. Titan joins Parasail’s multi-silicon fleet that is already serving trillions of tokens a day.
“No single chip is best at every AI workload, so building an inference cloud around just one never made sense," said Mike Henry, co-founder and CEO of Parasail. “Positron fills a critical part of our fleet for token traffic that is highly sensitive to cost and needs titanic scale up. We see their Titan Systems being instrumental in long-running agents, agent swarms, and physical AI and robotics use cases that have to run in power constrained environments.”
“Agents need the largest models, and running the largest models is a memory problem,” said Mitesh Agrawal, CEO of Positron AI. “We built Titan to hold multi-trillion-parameter models in low-power commodity memory. Parasail already runs our Atlas systems in production, and it knows how to get new silicon to customers at scale.”
About Parasail
Parasail is the first multi-silicon AI inference cloud designed to intelligently route different AI workloads to the hardware best suited to run them. Combining intelligent orchestration with advanced kernel optimization, Parasail builds on its established GPU infrastructure to deliver high-performance, cost-efficient inference through a unified platform. As AI computing technology advances, Parasail is designed to incorporate those innovations so customers can benefit from ongoing improvements in performance and efficiency without managing multiple hardware providers or having to replatform. Parasail is headquartered in San Mateo, California, and backed by Kindred Ventures, Touring Capital, Basis Set, and others. Learn more at parasail.io.
About Positron AI
Positron AI builds hardware and software to make AI inference dramatically cheaper and more energy efficient. The company’s shipping product, Atlas, is deployed at hyperscaler scale today, and its next-generation custom silicon, Asimov, tapes out in late 2026 with production in the second half of 2027. Asimov powers Titan, a multi-terabyte-memory inference system for long-context and next-generation AI workloads. Positron’s energy-efficient architecture allows its systems to be deployed in air-cooled or liquid-cooled data centers at varying rack densities, without dependence on constrained HBM or CoWoS supply chains. Learn more at positron.ai.
View source version on businesswire.com: https://www.businesswire.com/news/home/20261119916374/en/
No single chip is best at every AI workload, so building an inference cloud around just one never made sense.
Contacts
Media contacts
Parasail: parasail@aircoverpr.com
Positron: Cbackus@positron.ai