CloudRoute · Cornerstone guides
The CloudRoute blog
The in-depth, neutral reference behind CloudRoute — 36 cornerstone guides across five topics: how AWS credits really work, building GenAI on Bedrock, migration and DevOps playbooks, and the levers that cut an AWS bill.
Each topic opens into our full library of 341+ reference pages.
AWS Credits & Funding
How the credit programs actually work — Activate, Portfolio, partner-filed, POC — and how to stack them to the maximum at every stage.
- AWS Activate tiers — Builders vs Founders vs PortfolioAWS Activate tiers compared (2026): Builders ($1K) and Founders ($5K) are self-serve; Portfolio ($100K) is gated. Amounts, eligibility, validity, how to climb.
- Stacking AWS credits to the maximum — the 2026 referenceAWS credits stack additively, not arbitrarily. The 2026 guide to combining Activate Portfolio, Build for Startups, Bedrock/GenAI POC, MAP and accelerator credits — the rules, the additive-workload requirement, the order to file in, realistic ceilings ($150K typical, $300K+ for AI), and the mistakes that downgrade the secondary tracks to $0.
- How to get AWS credits without a VC (2026)You don't need a VC for AWS credits. The 2026 guide for bootstrapped and angel-funded startups: self-serve $1K–$5K, the partner-attested path to ~$25K, the exceptions above it, what disqualifies you, and how a partner attestation replaces the VC vouch.
- The complete guide to partner-filed AWS credits (2026)The definitive 2026 reference on partner-filed AWS credits: what partner-filed means, how the ACE / APN mechanic works, why it unlocks the $100K Activate Portfolio tier the self-serve form cannot, what the partner submits, what AWS reviewers check, timelines, and who should use it.
- How to get $100K in AWS credits (2026)The honest 2026 how-to for $100K in AWS credits: who qualifies, the four routes, the exact application AWS reads, the use-case that gets approved, and how to reach $150K.
Generative AI & Bedrock
Building, choosing, and cost-tuning generative AI on AWS — Bedrock models, RAG, agents, and the FinOps of inference.
- AWS credits for generative-AI startups — the full 2026 funding mapThe 2026 guide to AWS credits for generative-AI startups: Activate Portfolio, the Bedrock POC track, the $300K GenAI tier, and the up-to-$1M Accelerator — who qualifies and how.
- The Amazon Bedrock cost-optimization guide (2026)The definitive FinOps-for-GenAI playbook for Amazon Bedrock: model right-sizing, intelligent routing, prompt caching, batch inference, provisioned throughput break-even, output-token discipline, RAG vs long context, distillation, and embeddings — each with mechanism, real numbers, and impact ranking.
- Bedrock POC credits — the full 2026 unlock guideThe definitive 2026 guide to AWS Bedrock POC funding: what the program is, the $10K–$50K range, eligibility, what counts as a fundable POC, the partner-filed mechanic, how it stacks with Activate, the timeline, and turning a funded POC into production.
- Choosing a Bedrock model — the full 2026 frameworkThe 2026 framework for choosing a Bedrock model: map task to family (Claude, Nova, Llama, Mistral, embeddings), weigh quality/cost/latency, measure with evals, and route by tier. Decision matrix by use case.
- Choosing a vector database on AWS — the 2026 decision guideOpenSearch vs Aurora pgvector vs Pinecone vs Redis vs Neptune vs S3 Vectors — on cost (the OCU trap), latency, scale, filtering, ops, and Bedrock integration.
- Choosing an LLM provider in 2026 — the full decision frameworkThe 2026 framework for choosing an LLM provider: the seven axes that matter, honest verdicts on Bedrock, OpenAI, Azure and Vertex, and a decision table by scenario.
- Running Claude on Amazon Bedrock in production (2026)The definitive 2026 guide to running Claude on Amazon Bedrock in production: Opus/Sonnet/Haiku model selection, the Converse API, prompt caching and batch for cost, Guardrails, provisioned throughput, evals, and the cost math at scale.
- Cutting SageMaker inference cost — the full 2026 playbookThe 2026 guide to cutting Amazon SageMaker inference cost: endpoint types, scale-to-zero, multi-model endpoints, Inferentia, Savings Plans, and when to move to Bedrock — every lever ranked.
- Enterprise AI assistant — the 2026 build-vs-buy guideThe neutral build-vs-buy guide for an internal enterprise AI assistant on AWS: Amazon Q Business, Microsoft 365 Copilot, and Glean vs custom on Bedrock. Six evaluation criteria, a weighted scorecard, and when custom-on-Bedrock wins.
- GenAI FinOps on AWS — the full 2026 disciplineGenAI FinOps on AWS: token unit economics (cost per request/user/feature), inference-profile attribution, the cost levers, budgets, chargeback, and credit-funded builds.
- GPU vs Trainium vs Inferentia — the 2026 AWS AI compute cost guideGPUs (P5/H100/H200) vs AWS Trainium vs Inferentia vs Bedrock in 2026: price-performance, the Neuron SDK porting catch, break-even math, and when each wins for training vs serving.
- HIPAA-compliant generative AI on AWS — the 2026 guideHIPAA-compliant generative AI on AWS (2026): the Bedrock BAA, the no-training guarantee, Guardrails PHI redaction, the reference architecture, and what not to do.
- How to build generative AI on AWS — the 2026 playbookThe 2026 build guide to generative AI on AWS: the 4 reference architectures, Bedrock vs SageMaker vs self-host, model selection, cost control, and POC to production.
- AI image generation on AWS — the 2026 prompting & production guideThe definitive 2026 guide to image generation on Amazon Bedrock: Nova Canvas vs Titan vs Stable Image compared, prompting (structure, negatives, seeds, cfgScale), editing, batch + S3/CloudFront delivery, cost per image, and watermarking/safety/licensing.
- AI inference cost optimization on AWS — the full 2026 guideThe definitive 2026 guide to cheap AI inference on AWS: Bedrock vs SageMaker vs self-hosted EC2/Inferentia cost models, the levers per option (caching, batch, routing, Spot, Graviton), quantization & distillation, and the build-vs-buy break-even.
- The ML lifecycle on AWS — the full 2026 MLOps pathThe definitive 2026 MLOps guide: the ML lifecycle on AWS end to end — data, training, evaluation, deployment, monitoring, retraining on a SageMaker Pipelines + Model Registry backbone.
- Porting PyTorch to AWS Neuron — the real 2026 engineering guideThe engineering guide to porting PyTorch to AWS Neuron in 2026: the compiler + runtime, torch-neuronx / transformers-neuronx / Optimum Neuron, the trace/compile step, effort tiers, blockers, distributed training, serving on Inferentia, and the cost payoff.
- Tiered model routing on Bedrock (2026)The definitive 2026 guide to tiered model routing on Amazon Bedrock: route each request to the cheapest model that meets the quality bar. Bedrock Intelligent Prompt Routing vs custom routers (classifier, cascade, confidence-based), the architecture, quality gates and fallback, measuring savings with a counterfactual, pitfalls, and the worked math.
Migrate to AWS
Moving off Heroku, GCP, Azure, on-prem, Oracle, or SAP — the strategy, the cutover, and the MAP funding that pays for it.
- How a MAP-funded AWS migration actually worksThe operational walkthrough of a MAP-funded AWS migration: the timeline from discovery to optimize, who does what (you vs partner vs AWS), how the funding flows to net out near $0, a realistic worked example, and how to start.
- AWS MAP funding — the definitive 2026 referenceHow AWS Migration Acceleration Program (MAP) funding works in 2026: the three phases (Assess, Mobilize, Migrate & Modernize), exactly what AWS funds at each, how funding scales with workload size and committed spend, eligibility, the partner-filed APN unlock, and MAP vs Activate.
- The AWS migration playbook — discovery to optimizedThe definitive 2026 AWS migration playbook: portfolio discovery, the 7-R disposition, the TCO/business case, the landing zone, wave planning, cutover mechanics (MGN/DMS, test→cutover→validate→rollback), post-migration optimization, change management, and how MAP funds it all.
- The 7 Rs of cloud migration, explainedThe 7 Rs of cloud migration — Retire, Retain, Rehost, Relocate, Repurchase, Replatform, Refactor: what each means, its cost/effort/risk, and when to choose it (2026).
Cost Optimization & FinOps
The levers that actually cut an AWS bill — commitments, right-sizing, storage, egress — and how AWS funds the work for you.
- Building a FinOps operating model on AWSThe definitive guide to running FinOps on AWS: the Inform-Optimize-Operate framework, who owns cost (engineering, finance, FinOps lead) with a RACI, the recurring rituals, unit economics, showback vs chargeback, crawl-walk-run maturity, KPIs, and FinOps-as-a-service.
- How AWS funds your cost-optimization workAWS funds partner-led Well-Architected cost reviews — and a review can unlock remediation credits. The 2026 guide to the cost pillar, the funding programs, and how to start.
- The Well-Architected cost pillar, in practiceThe 2026 guide to the AWS Well-Architected cost-optimization pillar: five design principles, four practice areas, the HRIs a review surfaces, and how a partner-led WAFR makes the fixes AWS-funded.
- Cloud egress & data-transfer costs, explained (2026)The definitive 2026 guide to AWS data-transfer costs: internet egress, cross-AZ, inter-region and NAT charges — every per-GB price, why it balloons, and the free fixes.
- How to reduce your AWS bill — the 2026 FinOps playbookEvery lever that moves an AWS bill, ranked by impact and effort: Savings Plans, right-sizing, Spot, Graviton, storage tiering, the data-transfer/NAT trap, idle cleanup, scheduling, and governance — each with its mechanism and typical savings, plus when a funded audit pays off.
DevOps & Platform
Terraform, EKS, CI/CD, landing zones, observability, and reliability — what good looks like on AWS, and when to bring in help.
- Do you actually need Kubernetes? (AWS, 2026)Most AWS teams do not need Kubernetes. The honest 2026 guide: the 4 signals that justify EKS, the simpler paths (ECS/Fargate, App Runner, Lambda), the true cost (the people), the resume-driven trap, and a decision framework by stage.
- The production-ready EKS checklist (2026)The definitive Amazon EKS go-live checklist (2026): networking, IP exhaustion, IRSA vs Pod Identity, Karpenter, security, cost, DR, and upgrades as actionable checks.
- Infrastructure-as-code best practices on AWS (2026)The definitive 2026 IaC reference for AWS: choosing Terraform, OpenTofu, CDK, CloudFormation, or Pulumi; module design; remote state + locking; multi-account structure; CI/CD plan/apply gates; policy-as-code; testing; secrets; and migrating off ClickOps.
- Terraform vs OpenTofu in 2026 — the definitive guideThe neutral, current 2026 guide to Terraform vs OpenTofu: the BSL license change and the fork, feature parity, state encryption, the registry, momentum, migration effort, which to choose by org type, and the AWS angle (both work).