True random number generators (TRNGs) extract randomness from physical phenomena to produce inherently unpredictable bitstreams. Owing to their strong cryptographic properties, TRNGs are fundamental components for establishing trusted roots in secure systems. However, the throughput of current TRNGs falls short of meeting the increasing demands posed by high-speed encryption and rapidly growing data volumes. To solve this issue, in this article, we propose a multi-channel tetrahedral TRNG (MCT-TRNG) via metastability-enhanced entropy with ultra-high throughput. We first propose a novel entropy source structure of a metastability-enhanced ring oscillator, which extracts randomness in unstable signals by switching transmission paths. Then, we introduce a feedback XOR ring to improve the degree of signal chaos. On this basis, we propose a tetrahedral post-processing structure with four channels to produce independent parallel outputs. The experiments show that the generated random sequences have successfully passed the NIST and AIS-31 tests. The MCT-TRNG incurs only 18 LUTs with a throughput of 2.2 Gbps on Xilinx Artix-7 FPGA. Compared with existing works, our design has the highest throughput, yielding promising application potential.