Always good to see more competition in the inference chip space, especially from Europe. The specs look solid, but the real test will be how mature the software stack is and whether teams can get models running without a lot of friction. If they can make that part smooth, it could become a practical option for workloads that want local control.