{"city":null,"name":"Tensordyne","slug":"tensordyne","facts":[{"date":"2026-06-15","type":"revenue_band","value":"pre-revenue","source":"Tensordyne release, Reuters, DCD"},{"date":"2026-06-15","type":"contract","value":"Forecast demand from LOIs exceeding $200M","source":"Tensordyne release, Reuters, DCD"},{"date":"2026-06-15","type":"product_launch","value":"Napier inference system announced; chip in production at TSMC 3nm","source":"Tensordyne release, Reuters, DCD"},{"date":"2025-09-16","type":"ownership_change","value":"Recogni rebranded as Tensordyne and pivoted from automotive edge chips to data-center LLM token generation","source":"EE Times"},{"date":"2025-09-16","type":"product_launch","value":"Automated quantization toolchain, HuggingFace model zoo, and Triton support planned","source":"EE Times"},{"date":"2025-09-16","type":"partnership","value":"Juniper Networks-assisted any-to-any interconnect for planned 144-chip data-center domains","source":"EE Times"}],"peers":[],"stage":null,"state":null,"aliases":["Recogni","Napier","TDN"],"country":"United States","domains":[{"why":"Tensordyne designs data-center AI chips specialized for high-throughput LLM inference, including its Napier system, which uses HBM3e and logarithmic arithmetic.","list":null,"rank":null,"tier":"watchlist","layer":null,"total":55,"domain":"AI Core","scores":{"leverage":8,"position":4,"progress":9,"firepower":6,"commercial":3,"deployment":2,"production":6,"technology":6},"edition":null,"segment":null}],"founded":null,"signals":[{"date":"2026-06-16","header":"Tensordyne has taped out its Napier AI inference compute engine after raising $176 million.","outlet":"The Next Platform"},{"date":"2026-06-16","header":"Tensordyne’s Napier architecture combines logarithmic compute cores, vector engines, and RISC-V cores.","outlet":"The Next Platform"},{"date":"2026-06-16","header":"Broadcom is serving as the chip shepherd for Tensordyne’s Napier compute engine.","outlet":"The Next Platform"},{"date":"2026-06-16","header":"Tensordyne’s stated rack comparison targets six times the token throughput with 30% less power.","outlet":"The Next Platform"},{"date":"2026-06-16","header":"Tensordyne plans to offer cloud access to Napier engines by the end of 2026.","outlet":"The Next Platform"},{"date":"2026-06-16","header":"Napier delivers 2.1 petaflops of dense FP8 performance with 144 GB of HBM4.","outlet":"The Next Platform"}],"website":"tensordyne.ai","category":"AI Infra · Chips & accelerators","categories":[{"name":"Inference-specialist chips","domain":"AI Core","parent":"AI Infra · Chips & accelerators"}],"last_round":null,"updated_at":"2026-09-22T22:20:52.182+00:00","facts_total":16,"positioning":"Logarithmic-math inference system maker (formerly Recogni; Sunnyvale and Munich) whose 3nm Napier chip, built with Broadcom and Juniper, is in production at TSMC; over $200M of forecast demand from a dozen letters of intent, Cirrascale as first named customer.","sub_category":"Inference-specialist chips","signals_total":29,"url":"https://www.frontiertechintelligence.com/companies/tensordyne","title":"Tensordyne · ai core, United States · scores, facts and signals · Frontier Tech Intelligence","description":"Tensordyne designs data-center AI chips specialized for high-throughput LLM inference, including its Napier system, which uses HBM3e and logarithmic arithmetic. AI Core watchlist (score 55) 29 signals and 16 facts with sources.","publisher":"Frontier Tech Intelligence (Future Frontiers, a trade name of Essential Cognition Inc.)","note":"Assessment by Frontier Tech Intelligence; facts and signals carry their sources."}