The Dispatch

Writing on AI infrastructure

Deep, measured engineering on GPUs, multi-cloud architecture, LLM systems, and autonomous infrastructure operations.

7 articles

What is GPUVerse? The Intelligence Layer for AI Infrastructure

GPUVerse is the AI Infrastructure Engineer: a deterministic decision engine that compares 10 GPU clouds and 54 accelerators to recommend where and how to run AI.

GPUVerse Team

August 7, 202629 min read

Read story

Latest

Newest first

Comparisons

August 12, 20267 min read

GPU Cloud Pricing in 2026: What Every Accelerator Costs

Floor prices for 15 GPU and TPU types across 10 clouds as of August 2026, from $0.59/hr RTX 4090s to $6.99/hr GB200, plus the cost-per-token math that turns hourly rates into real unit economics.

Read article

Tutorials

August 12, 20267 min read

How to Choose a GPU for LLM Inference (2026 Guide)

A practical method for sizing GPUs for LLM inference: the VRAM math, which accelerator fits each model size, and how to compare on cost per million tokens.

Read article

Research

August 12, 20267 min read

What is ComputeOps? A Definition of the Discipline

ComputeOps is the discipline of deciding where and how AI workloads should run, deterministically, explainably, and across every provider. Here is the full definition, how it differs from DevOps, MLOps, and FinOps, and why AI compute needed its own practice.

Read article

Comparisons

August 9, 20268 min read

The 2026 GPU Cloud Comparison: 10 Providers, 54 Accelerators

An analysis of the GPU cloud landscape in 2026, hyperscalers, neoclouds, and the Vast.ai marketplace, examining pricing models, spot support, availability, and cost per token.

Read article

GPU Benchmarks

August 8, 20269 min read

H100 vs H200 vs B200: Choosing the Right GPU for Your AI Workload

A technical comparison of NVIDIA's H100, H200, and B200 for AI, specifications with sparsity qualifiers, catalog floor pricing, and cost-per-token arithmetic you can verify.

Read article

Tutorials

August 4, 202611 min read

The Hidden Cost of GPU Spot Instances: A Practical Guide to Preemption

Spot instances offer dramatic cost savings for GPU workloads, but the operational complexity can quickly offset the price advantage. Here's how to use spot instances effectively.

Read article

Browse