Abstract: The Java virtual machine is stack based and object oriented. These features are not suitable for the hardware to execute the bytecodes directly, and the traditional just in time （JIT） compilers can’t meet the requirements of embedded system too. It is really a great challenge to design a good embedded Java chip. Introduced in this paper is the approach used in embedded Java chip JC401. With the support of the stack architecture and complicated instructions of Java virtual machine, and close matched with the JIT compiler and other software techniques, the problems in Java chip design have been overcome. The result of simulation indicates that the compiled Java programs in JC401 are 1.2 to 1.9 times as fast as the picoJava II which has the best performance in the existing Java chips, while the compiling cost can be ignored and JC401 improves considerably in hardware complex, native performance, and memory cost, compared with picoJava II. So the JC401 is suitable for embedded systems, and can provide high Java performance for embedded applications.