Together AI and Y Combinator partner to launch the first dedicated GPU cluster for the YC community

Together AI and Y Combinator announced a partnership to deliver the first dedicated YC GPU cluster, intended to give YC's portfolio of AI-native startups easier access to compute for building and scaling.

According to the announcement, securing compute capacity has become one of the hardest problems for young companies, and for many startups the upfront cost of two years of compute exceeds their entire cash balance. Together AI built the cluster and a developer experience specifically for YC portfolio companies to get compute for inference and training quickly and cost-effectively. Instead of long-term commitments, startups can spin up GPUs for short-term sprints while benefiting from long-term rates.

The cluster supports needs ranging from single-node compute to companies scaling up. YC founders also get access to learnings and best practices from Together's researchers, engineers, and customer experience teams. Startups reserve, provision, and manage their own GPUs through Together's self-service portal with their own billing support, so scaling compute does not route through YC. Founders get GPUs ready in minutes, the announcement said. The cluster is running at full utilization today, while individual companies can still plan compute needs months in advance.

Together AI said it was founded four years ago and works with over 8,000 customers, including Cursor, Decagon, Cognition, and Eleven Labs. Its research includes work on attention mechanisms and the Mamba architecture now used in models like NVIDIA's Nemotron.

The two organizations plan to expand the cluster, and Together will support companies as they graduate from the batch and scale. Together said founders can also use the rest of its platform, from inference to fine-tuning and training. YC's next batch is taking applications for its Fall 2026 cycle, and YC said it is especially interested in funding companies developing new research breakthroughs that require GPU compute.