{"slug": "neurophos-opu", "title": "Neurophos OPU", "summary": "Neurophos announced its Optical Processing Unit (OPU), a photonic AI chip achieving 0.47 ExaOPS at 300 TOPS/W, claiming 10,000x size reduction over traditional photonic tensor cores. The company says a single tray of 8 OPUs outperforms a 3,000 lbs GPU rack while consuming only 10 kW, with evaluations starting in 2026 and production in 2028.", "body_md": "top of page\n\n# THE OPU\n\nCompressing a rack into the size and power draw of a single GPU.\n\n# 0.47\n\nExaOPS\n\n# The most efficient AI chip on the planet\n\n# 300\n\nTOPS/W\n\n# THE BREAKTHROUGH\n\n# Photonic Tensor Cores Are Enormous\n\n# Neurophos Made Them\n\n10,000x Smaller\n\n# The elements required for ExaOPS speeds occupy 1m x 1m.\n\n# Demonstrated. Protected by 300 patents.\n\n# 30\n\n# YEAR\n\nLEAP\n\nA LEAP EQUIVALENT TO 30 YEARS OF TRANSISTOR SCALING\n\n# THE WORLD DEMANDS TERAWATT COMPUTE\n\n# This requires more power than the United States produces today\n\n# With\n\nGPUs\n\n# With\n\nOPUs\n\n-\n# 1000+ nuclear plants\n\n-\n# $20 Trillion for GPUs\n\n-\n# 10 nuclear plants\n\n-\n# 100x fewer XPUs\n\n# 10 kW\n\n# 2 ExaFLOPS\n\nSingle tray. 8 OPUs. Faster than a 3,000 lbs GPU rack.\n\n# TERAWATT COMPUTE.\n\nGIGAWATT POWER.\n\nEvaluations begin 2026.\n\nFirst systems targeted for early 2028.\n\nProduction ramp begins mid 2028.\n\nbottom of page", "url": "https://wpnews.pro/news/neurophos-opu", "canonical_source": "https://www.neurophos.com", "published_at": "2026-06-14 23:51:14+00:00", "updated_at": "2026-06-15 00:12:31.376795+00:00", "lang": "en", "topics": ["ai-chips", "ai-infrastructure"], "entities": ["Neurophos", "OPU", "GPU"], "alternates": {"html": "https://wpnews.pro/news/neurophos-opu", "markdown": "https://wpnews.pro/news/neurophos-opu.md", "text": "https://wpnews.pro/news/neurophos-opu.txt", "jsonld": "https://wpnews.pro/news/neurophos-opu.jsonld"}}