NextSilicon, the Israel-based startup whose intriguing software-defined hardware has been talked about in concept for years by the company, today unveiled what it said is the first Intelligent Compute Accelerator (ICA).
Called Maverick-2 ICA, NextSilicon said it delivers more than 4x the performance-per-watt of traditional GPUs and 20x that of high-end CPUs, while reducing operational costs by more than half. The product, targeting HPC-AI, “redefines accelerator architectures…, moving beyond the limitations of traditional fixed GPU designs, which are based upon the eight-decade-old Von Neumann architecture.”
The product uses telemetry and NextSilicon’s adaptive algorithms to identify the “likely flow” of an HPC and AI application and, in accordance, “dynamically tailor acceleration in real-time.”
“Its self-optimizing architecture delivers immediate, scalable performance improvements,” the company said. “This is achieved by creating software-defined processor cores tailored to the performance needs of each application and then executed on the accelerator hardware. As your application runs, telemetry data is fed to NextSilicon’s intelligent algorithms to continuously self-optimize performance, power consumption and utilization in real-time. The result is efficient performance at scale for the most demanding HPC applications, while consuming 50-80 percent less power than traditional GPUs.”
The product is shipping to “dozens” of customers, according to the company, with volume shipments to begin early next year. NextSilicon said customers include the US Department of Energy and academic research institutions, along with organizations in financial services, energy, manufacturing and life sciences. Early adopter programs for new customers are now available through partners Penguin Solutions and Dell Technologies.
“Sandia National Laboratories is already using new Maverick-2 accelerators in anticipation of our deployment later this year of our new integrated system – named Spectra – which will be delivered in 2025,” said Sandia’s James H. Laros III, distinguished member of the technical staff and Vanguard Program lead. “Our partnership with NextSilicon, which began over four years ago, has been an outstanding example of how the NNSA laboratories can work with industry partners to mature novel emerging technologies. NextSilicon’s continued focus on HPC made them a prime candidate for the Vanguard program.”
NextSilicon said alternative accelerators have failed to unseat traditional GPUs and CPUs because of application porting and vendor-specific framework complexities, whichcreate lock-in.

source: NextSilicon
“NextSilicon’s Maverick-2 ICA is the first to break through these barriers with a powerful, user-friendly software suite,” the company said. The product natively supports languages such as C/C++, FORTRAN, OpenMP and Kokkos, enabling many applications to run on Maverick-2 without modification, according to NextSilicon. This saves months in development effort, the company said, and it plans upcoming integrations for CUDA, HIP/ROCm, and leading AI frameworks.
“Traditional CPU and GPU architectures are often constrained by high-latency pipelines and limited scalability,” said Steve Conway, senior analyst at Intersect360 Research. “There is a clear need for reduced energy waste and unnecessary computations within HPC and AI infrastructures. NextSilicon addresses these important issues with Maverick-2, a novel architecture purpose-built for the unique demands of HPC and AI. Maverick-2 is designed to break down barriers to adoption and migration from legacy architectures.”
The company also said Maverick-2 can “be easily adapted to new use cases for converged HPC-AI workloads, vector database applications, breakthroughs in AI models, and emerging agentic and generative AI applications.”
“We are excited to support the launch of the Maverick-2 Intelligent Compute Accelerator and its potential to drive new levels of performance and efficiency in the HPC market,” said Phil Pokorny, CTO of Penguin Solutions. “At Penguin, we are adept at turning cutting-edge ideas into outcomes—faster, and at any scale, enabling our customers to tackle the most demanding workloads. Maverick-2’s intelligent, software-defined approach complements the wide array of technologies we provide, offering organizations a powerful option to optimize their infrastructure and achieve breakthrough results in HPC and AI. As complexity in applications grows, innovations like Maverick-2 open new possibilities for our customers to accelerate scientific and commercial advancements.”
NextSilicon said it has raised a total of $303 million in funding from investors such as Aleph, Amiti, Liberty Technology VC, Playground Global, Standard Investments, StepStone, and Third Point Ventures.