AI Workload Design Scorecard
Turn a vague AI idea into a workload and architecture checklist before choosing a model or provider.
Compute College Tools
Learn by estimating AI compute costs.
Free calculators for estimating GPU-hour cost, training-run budgets, and model serving cost from your own inputs. Every calculator shows its assumptions and links to the lesson that explains the math.
Estimates are planning tools, not quotes.
AI Engineering labs
Learn by testing the system
Start with browser-local learning tools for workload design, context budgets, prompt evaluation, retrieval quality, agent permissions, and accepted outcomes. No account, model API, or market quote is required.
Context Budget Calculator
See how instructions, retrieved material, and reserved output consume a context window.
Prompt Evaluation Worksheet
Compare accepted results, retries, and review time across prompt versions.
RAG Retrieval Exercise
Explore precision, recall, and the token cost of adding retrieved chunks.
AI Workload Planner
Estimate token cost and cost per successful request from traffic and success rate.
Agent Permission Designer
Classify tool actions by risk, reversibility, approval, and automation.
Cost per Successful Task Calculator
Include failed attempts and review in the cost of producing an accepted result.
Estimate cost
Estimate cost from your own inputs
Three calculators for sizing a budget from scratch. Each runs in your browser and updates as you type.
GPU-Hour Cost Calculator
Estimate AI compute cost from GPU price, runtime, utilization, and overhead.
AI Training Cost Calculator
Estimate a training-run budget using GPU-hours and operating assumptions.
Model Serving Cost Calculator
Estimate recurring inference cost from usage and capacity needs.
Decide between options
Pick between approaches before you spend
Two side-by-side decision tools for the most common AI compute trade-offs. Each one produces a break-even and a recommendation.
API vs Self-Hosted Calculator
Compare API serving cost to a self-hosted GPU cluster, with break-even output volume and utilization sensitivity.
Reserved vs On-Demand Calculator
Compare a GPU reservation to pure on-demand; see break-even GPU-hours and stranded capacity.
Convert infrastructure
Translate physical infrastructure into compute
When a number lands in megawatts, racks, or square feet, this is where you turn it into a GPU range.
Power-to-Compute Translator
Translate an announced megawatt number into a GPU-capacity range using PUE, rack density, and utilization.
Decision guides
Compare your options before you estimate
These lessons walk through the trade-offs behind each number — access terms, utilization, and how to read a quote.
On-demand vs reserved vs spot GPU pricing
Compare access terms, savings, and interruption risk.
How to compare GPU cloud quotes
Compare rate, reliability, network, and completed-workload cost.
How to estimate monthly AI compute burn
Turn training and serving usage into a recurring budget.
What is GPU utilization?
Why paid capacity can cost more when it sits idle.
Compare quotes
Compare two or more provider quotes
GPU cloud quotes are not just rates. The checklist below names the eleven fields to record on every offer so two quotes can be read side by side without confusion. Use it as the column header for your own comparison sheet.
GPU Cloud Quote Comparison Checklist
Eleven fields to record on every quote: GPU generation, effective cost, commitment, region, capacity, networking, storage and egress, SLA, interruption, flexibility, and quote source date.
How to compare GPU cloud quotes
The narrative method behind the checklist.
Start with the basics
New to compute cost? Learn the units first
Compute College explains GPU-hours, utilization, training cost, and serving cost in plain English before you reach for a calculator.