The next generation magnetic spectrometer in space, AMS-100, is designed to have a geometrical acceptance of $100\,\mathrm{m}^2\,\mathrm{sr}$ and to be operated for at least ten years at the Sun-Earth Lagrange Point 2. Compared to existing experiments, it will improve the sensitivity for the observation of new phenomena in cosmic rays, and in particular in cosmic antimatter, by at least a factor of 1000. The magnet design is based on high temperature superconductor tapes, which allow the construction of a thin solenoid with a homogeneous magnetic field of 1 Tesla inside. The inner volume is instrumented with a silicon tracker reaching a maximum detectable rigidity of 100 TV and a calorimeter system that is 70 radiation lengths deep, equivalent to four nuclear interaction lengths, which extends the energy reach for cosmic-ray nuclei up to the PeV scale, i.e. beyond the cosmic-ray knee. Covering most of the sky continuously, AMS-100 will detect high-energy gamma rays in the calorimeter system and by pair conversion in the thin solenoid, reconstructed with excellent angular resolution in the silicon tracker.

The AMS-02 detector will measure cosmic rays on the International Space Station. This contribution will cover production, testing, space qualification and integration of the AMS-02 anticoincidence counter. The anticoincidence counter is needed to to assure a clean track reconstruction for the charge determination and to reduce the trigger rate during periods of high flux.

For cosmic particle spectroscopy on the International Space Station the AMS experiment will be equipped with a Transition Radiation Detector (TRD) to improve particle identification. The TRD has 20 layers of fleece radiator with Xe/CO2 proportional mode straw tube chambers. They are supported in a conically shaped octagon structure made of CFC-Al-honeycomb. For low power consumption VA analog multiplexers are used as front-end readout. A 20 layer prototype built from final design components has achieved proton rejections from 100 to 2000 at 90% electron efficiency for proton beam energies up to 250 GeV with cluster counting, likelihood and neural net selection algorithms.