Baseten showing up at GTC
Baseten is being name‑checked in coverage about cryptic SF billboards and LLM autoscaling, and had visibility in NVIDIA GTC reporting as an autoscaling/ML‑ops player. That positions them in the LLM production/autoscaling conversation where training reliability and throughput matter. (techspot.com)(brics-econ.org)(hyperight.com)
Baseten closed a $300 million financing that valued the company at $5 billion in late January 2026, with IVP and CapitalG listed as lead investors. (businesswire.com) Multiple outlets reported NVIDIA participated in that round with roughly $150 million of capital, signaling a strategic partnership between the GPU vendor and the inference platform. (pymnts.com) Baseten ran an official presence at NVIDIA GTC, promoting its inference stack at the March 16–19, 2026 event in San Jose. (baseten.co) GTC 2026 drew more than 30,000 developers, researchers, and executives to the show floor, amplifying Baseten’s visibility during major product and partner announcements. (deeperinsights.com) Baseten’s engineering docs describe an autoscaler that dynamically adjusts replica counts to balance latency, throughput and GPU‑hour cost for production endpoints. (docs.baseten.co) The company’s “How Baseten Works” documentation maps Truss packaging, build pipelines, request routing and deployment lifecycle steps that move models from push to running endpoints. (docs.baseten.co) An NVIDIA case study highlights Baseten’s use of TensorRT‑LLM and a custom model‑builder to speed up model compilation and optimize inference performance. (nvidia.com) Baseten publishes a model library including NVIDIA Nemotron 3 Super, GLM 5 and MiniMax M2.5, offers a multi‑cloud managed inference product, and cites customers such as Abridge, Writer and Patreon as production users. (baseten.co) Baseten’s multi‑cloud work with providers produced a Google Cloud case study claiming over 225% better cost‑performance for high‑throughput inference on A4 VMs and about 25% improvement for latency‑sensitive workloads. (cloud.google.com) Baseten’s marketing head Mike Bilodeau characterized the recent San Francisco cryptic billboard campaign as intentionally inscrutable and aimed at insider audiences rather than general consumers. (kedm.org)
