CoreWeave, Inc. (NASDAQ:CRWV) stock rose Thursday after unveiling an AI-native object storage platform designed to move data to GPUs with minimal friction and without extra egress or transaction fees.
The company introduced CoreWeave AI Object Storage, a fully managed service built to keep training and inference pipelines fed at scale.
The system uses CoreWeave’s Local Object Transport Accelerator, or LOTA to make a single dataset instantly reachable across regions, clouds, and on-prem environments.
Modern AI models live or die on data throughput and proximity to compute.
Traditional object stores can throttle GPU utilization with latency, duplication, and cost sprawl.
CoreWeave’s architecture aims to maintain high, predictable bandwidth so developers can train and serve models without shuffling copies or paying unpredictable fees.
CoreWeave says performance scales with workload size and sustains throughput to distributed GPU nodes “from any region, on any cloud, or on-premises,” aided by private interconnects, cloud peering, and ports rated up to 400 Gbps.
The company positions this backbone as a way to keep global datasets coherent while avoiding replication-heavy operations.
The storage lineup debuts with three automatic, usage-based tiers. CoreWeave targets more than 75% lower storage costs for typical AI patterns, while eliminating egress, request, and tiering charges.
The company pitches this as simpler budgeting and more transparent unit economics than many legacy storage schemes.
“As the essential cloud for AI, every decision at every layer is focused on optimizing for efficiency and performance,” said Peter Salanki, Co-Founder and CTO, CoreWeave.
The launch extends CoreWeave’s software push after recent steps like ServerlessRL for reinforcement learning.
The company has been building a broader AI stack via CoreWeave Ventures and acquisitions including OpenPipe and Weights & Biases, with a pending deal for Monolith AI.
Price Action: CRWV shares were trading higher by 4.29% to $145.21 at last check Thursday.
Read Next:
Photo via Shutterstock
