AI Platforms & Infrastructure
One of our seven practice areas — every brief we’ve published in this coverage area, newest first.
-
Private AI Platforms: VMware VCF vs. Red Hat OpenShift AI
Broadcom’s VCF Private AI Services and Red Hat OpenShift AI both promise NVIDIA-validated private AI. A decision-ready comparison with sizing rules, audited benchmarks, and honest downsides for both stacks.
-
NVIDIA AI Factories: What the Reference Designs Signal
NVIDIA’s Enterprise Reference Architectures define three tiers of on-prem AI factory. We decode RTX PRO, HGX, and GB300 NVL72 into a sizing decision tree — and flag the lock-in trade-offs.
-
Bedrock vs. Vertex AI vs. Azure AI Foundry in 2026
The three hyperscaler AI platforms have converged on features, so the 2026 decision turns on model exclusivity, FedRAMP status, and Gemini’s billing traps.
-
What an On-Prem AI Cluster Really Costs in 2026
Consolidated 2026 price anchors for on-prem AI: NVIDIA B200 and DGX B300 system costs, the $2-5M liquid-cooling retrofit nobody quotes, and the software stack from Broadcom/VMware or Red Hat.
-
Buy or Rent GPUs? The 2026 Break-Even Math
The buy-vs-rent GPU crossover sits at roughly 60–70% sustained utilization in 2026. The break-even math, AWS and Google Cloud Blackwell pricing, and a decision rule finance will accept.
-
Self-Hosted LLM vs. API in 2026: The Real Cost per Token
The break-even math on self-hosting LLMs vs. paying per token, with real mid-2026 numbers: where 8x H100 pods win, where falling API prices win, and the hidden costs on both sides.