What Runpod changed

• Added a detailed technical guide on fine-tuning Large Language Models (LLMs) using Axolotl.
• Introduced new content regarding Global Volumes for Serverless endpoints, allowing model weights to be stored once and read anywhere.
• Removed several prominent case study links and customer testimonial sections from the main page.
• Removed the "Fall 2026 State of AI Compute" report and related blog post links.
• Simplified the page layout by removing large "Solution" and "Enterprise" marketing blocks.

Runpod is shifting focus toward educational content that helps developers move from raw data to live models. By providing step-by-step guides for popular tools like Axolotl, they are lowering the barrier to entry for fine-tuning and positioning their GPU pods as the go-to environment for custom AI development.

• Added a new announcement banner for ISO 27001 certification, replacing a webinar link.
• Added a G2 rating badge (4.6 stars) near the main headline.
• Updated the "What's New" section with a guide on SSH/SCP and a post about Runpod Enterprise.
• Added new customer case studies including Nodalview and Faceless Video.
• Updated customer quotes to emphasize cost savings compared to AWS and scalability.
• Changed the primary call-to-action button text from "Sign Up" to "Get Started."

Security certifications like ISO 27001 help remove sales friction for enterprise customers. It signals that the platform is maturing beyond a developer tool into a serious infrastructure provider for large companies.

• Added "Global Volumes" beta, providing elastic, region-independent storage that can be mounted in any data center.
• Added a new customer testimonial emphasizing ease of integration and long-term retention.
• Replaced a blog post about Qwen3.8-Flash-Next with a guide on Global Volumes.
• Replaced a blog post about private GPU pools with a benchmark study on GPU embedding workloads.
• Updated the primary contact phone number to +1 (888) 478-6012.

Runpod is expanding its infrastructure capabilities to include region-independent storage. This reduces friction for developers who need to move data between different data centers and signals a move toward more flexible, stateful AI workloads.