# Blog

Insights on cloud and AI cost optimization, FinOps best practices, and the Xosphere IQ platform.

## [Introducing TokenIQ: FinOps for AI](/content/blog/introducing-tokeniq/index.html)

Today we are launching TokenIQ, a new product that turns enterprise AI spend into measurable outcomes through unified governance, attribution, productivity correlation, and automated optimization, anchored on a defensible new metric called AIR (AI Return).

Alan Hand

8 min read

## [Introducing AIR: A Practical Way to Measure AI ROI](/content/blog/air-the-ai-return-metric/index.html)

AI ROI is fuzzy, costs are climbing, and "we feel more productive" is not a defensible answer. AIR (AI Return) is the metric we built to address that.

## [Right-Sizing Is a Control Loop, Not a Quarterly Report](/content/blog/right-sizing-automation/index.html)

Quarterly right-sizing reviews fail because workloads change faster than tickets close. The fix isn't better reports — it's treating right-sizing as a continuous control loop with the right guardrails.

## [Xogroups: Running Standalone EC2 Instances on Spot](/content/blog/xogroups-introduction/index.html)

ASGs assume your workload is fungible and replicated. A lot of real EC2 capacity isn't. Xogroups extend Spot replacement to standalone instances by tagging them as a group and replacing them one at a time, with the same private IP, EIP, EBS volumes, and ENI settings.

## [Cloud Cost Visibility: What Actually Moves the Needle](/content/blog/cloud-cost-visibility/index.html)

Most cost dashboards stop at the service level and call it visibility. The interesting questions start one layer deeper — where the NAT Gateway charges, idle commitments, and unallocated shared infra actually live.

## [One ASG, Many Instance Types, Some Spot: Reading the New Mixed Instances Policy](/content/blog/asg-mixed-instances-policy/index.html)

Auto Scaling Groups can now mix instance types and blend Spot with On-Demand in a single resource. The dual-ASG pattern is on its way out. What still needs an orchestrator on top is more interesting than what does not.

## [Running Production on Spot: A Field Guide for Skeptics](/content/blog/spot-production-workloads/index.html)

The advice to keep production off Spot is a decade old and rarely revisited. Here is what actually breaks, what does not, and the operational pattern that lets serious workloads run 60-90% on Spot without a 3am page.

## [EC2 Instance Type Selection: An Operator's Decision Framework](/content/blog/instance-type-selection/index.html)

AWS publishes more than 150 EC2 instance types across families, generations, and sizes. Picking well isn't a spec-sheet exercise. It's an operational discipline with real cost and reliability consequences.

## [The Real ROI of Cloud Cost Optimization: A Defensible Model](/content/blog/cloud-optimization-roi/index.html)

Most cloud optimization ROI math is wrong because it ignores the baseline, undercounts engineering time, and overstates savings. Here's the model that survives a finance review.

## [Spot Without Bidding: What Smooth Pricing Means for Real Workloads](/content/blog/spot-smooth-pricing/index.html)

On November 14, AWS quietly took the bidding model out of Spot. Prices now move slowly with capacity, capped at On-Demand. The runbook for running Spot in production just got shorter, and a different one got longer.

## [Why We Started Xosphere: The Case for Intelligent Spot Orchestration](/content/blog/why-we-started-xosphere/index.html)

An honest look at why cost dashboards, right-sizing reports, and DIY Spot management kept failing the same way every month — and the operating model we built to replace them.