Penguin Solutions, Inc. — AI supply-chain exposure

Penguin Solutions, formerly SMART Global Holdings, is an AI-infrastructure specialist that designs, builds and operates large GPU clusters for others, including deployments for Meta. It is a systems integrator and operator rather than a chip or component maker — it takes accelerators, networking, storage and cooling and assembles them into working AI factories. Its value lies in the hard, unglamorous engineering of making tens of thousands of GPUs run reliably as one machine.

The structural hook is that owning silicon is not the same as running it. Standing up and keeping a large training cluster healthy — the interconnect, the power and thermal envelope, the failure management — is scarce expertise, and hyperscalers and enterprises increasingly outsource it. That positions Penguin as a supplier of the AI factory itself rather than any single part, exposed to the pace of cluster build-out. The model weights it modestly: a builder and operator sitting beside the hyperscalers, not the components they buy.