Tachyum revealed specifications for its groundbreaking 2nm Prodigy Universal Processor, promising to run AI models with far more parameters than any current solution while drastically cutting costs.
The company claims that the Prodigy Ultimate delivers up to 21.3x higher AI rack performance compared to Nvidia Rubin Ultra NVL576, while Prodigy Premium achieves up to 25.8x higher performance than Vera Rubin 144. The 2nm Prodigy, the first chip capable of surpassing 1,000 PFLOPs on inference, will publish detailed technical specifications within a week. For comparison, Nvidia Rubin currently provides 50 PFLOPs.
AI Authority Trend: FS Launches PicOS AI Switch System to Boost Scalable AI and HPC Networking
As global competition in AI intensifies particularly between China and the United States the demand for massive computational power grows. ChatGPT 4, for instance, has roughly 1.8 trillion parameters, while the human brain contains an estimated 150 trillion synapses. Emerging AI systems like BaGauLu already reach 174 trillion parameters, but researchers anticipate a major leap from models trained on the collective knowledge of humanity, potentially exceeding 100,000,000 trillion (10²⁰) parameters. Traditional large-scale AI infrastructure could cost more than $8 trillion and require over 276 gigawatts of power. In contrast, Tachyum projects its solution will achieve similar capabilities for $78 billion with just 1 gigawatt, making advanced AI accessible to multiple organizations and nations.
Tachyum also commits to open-sourcing all software and licensing its memory technology, which uses standard components to increase DIMM-based memory bandwidth tenfold. These innovations support JEDEC adoption, lowering costs and promoting widespread use. Previously, Tachyum released licensable TAI data types and its TPU core, with plans to open the Instruction Set Architecture (ISA) soon.
The Prodigy design has evolved to meet server, AI, and HPC demands, offering up to 5x integer performance, 16x AI performance, 8x DRAM bandwidth, 4x chip-to-chip I/O, 16-socket scalability, and 2x power efficiency. The 2nm upgrade significantly reduces power consumption, and each chiplet integrates 256 high-performance 64-bit cores. Backed by a recent $220 million investment, the 2nm Prodigy is on track for tape-out.
AI Authority Trend: Fourier Cooling Launches Scalable Prefabricated Data Centers for AI and HPC
Prodigy SKUs address AI, exascale computing, HPC, cloud, and big data workloads. Prodigy Ultimate integrates 1,024 cores, 24 DDR5 17.6GT/s memory controllers, and 128 PCIe 7.0 lanes, while Prodigy Premium scales from 512 to 128 cores with 16 DRAM channels. Entry-level Prodigy offers 32–128 cores with 4–8 DRAM controllers.
Tachyum also provides native software, compilers, libraries, applications, and AI frameworks, allowing customers to run unmodified Intel/AMD x86 binaries alongside native applications.
“With tape-out funding now secured after a long wait, the world’s first Universal Processor can proceed to production, designed to overcome the inherent limitations of today’s data centers,” said Dr. Radoslav Danilak, founder and CEO of Tachyum. “The distinct markets addressed by Prodigy are the AI, server, and HPC markets, requiring fast and efficient chips. Tachyum’s Prodigy Premium and Ultimate will supercharge workloads with superior performance at a lower cost than any other solution on the market.”
The Prodigy Universal Processor delivers unprecedented AI performance, triple the performance of leading x86 processors, and six times the HPC performance of the fastest GPGPUs, eliminating the need for dedicated AI hardware while drastically lowering data center costs.
AI Authority Trend: VDURA and AMD Launch Scalable Reference Architecture for AI and HPC
To share your insights, please write to us at info@intentamplify.com
