In brief: GPUs have memory limitations when facing the demands of AI and HPC applications. There are ways around this bottleneck, but the solutions can be expensive and cumbersome. Now, a startup ...