Amazon's recent $110 million investment, with Trainium chips directly entering campuses, has seen Berkeley, CMU, and UIUC all receive the hardware. The core pain point is very real—no matter how powerful the hardware, if efficient code can't be written, it's all for nothing. Looking forward to TeAAL and Autocomp truly lowering the barriers.

Amazon launches Build on Trainium with $110 million, opening Trainium chips to multiple universities and open-sourcing research achievements, covering over 10,000 students, focusing on machine learning, healthcare, and quantum computing. Berkeley develops the TeAAL framework and Autocomp, CMU optimizes FlashAttention, UIUC receives thousands of new-generation chips. The core challenge is writing code that efficiently utilizes specialized chips; the project provides dedicated clusters and expert support.