NuAxiSemiconductor trade intelligence
Next in line to challenge Nvidia: Taalas hardwires Llama into silicon, claims 17,000 tokens per second
What happened
AI chip startup Taalas announced its HC1 inference chip, claiming it can generate 17,000 tokens per second for Llama 3.1 8B, which it benchmarks at 48x the performance of Nvidia's B200. This represents an early-stage, unproven technology challenge to established GPU-based inference. Neutral: No immediate impact on supply, demand, or pricing for existing electronic components.
Impact and signal
Low impactImpact score: 22/100
Affected components and suppliers
AI Inference Chip
Taalas
startupbenchmarkinference
Source and time
Source: DigiTimes Asia · 2/24/2026 · View original source
Open the NuAxi live feed →
Real-time pricing, supply signals and AI analysis for semiconductor traders