The bet
Together AI sells inference and fine-tuning for open-source models: rent the compute, keep control of the model and the data. The round signals that investors see a durable second lane next to the closed frontier labs, one where enterprises that cannot or will not send data to OpenAI or Anthropic still get frontier-class capability.
Why the money is flowing here
Open-weight models have kept closing the gap with proprietary ones, and every capability jump makes the run-it-yourself lane more credible. The constraint is not model quality anymore but reliable, affordable serving at scale, which is exactly the layer Together occupies.
What it means for founders
Two practical takeaways. First, the price of serious open-model infrastructure keeps falling as this segment gets capitalized, which strengthens the case for a hybrid model strategy instead of a single-vendor dependency. Second, if your product’s pitch includes data sovereignty, the ecosystem you depend on just got healthier. Recheck your inference costs quarterly; the floor is moving.