Somewhere in the last year, "AI skills" quietly moved from a nice-to-have on a job description to an assumption. Nearly every AI-powered tool is built on the cloud, so the two skill sets have stopped being separate tracks and started being one. If you are a cloud person watching the AI wave and wondering whether you are behind, here is the reframe I want to offer: you are not behind. You are standing on the exact ground everyone else is scrambling to reach. The trick is knowing which "AI skills" actually count, because most of what gets hyped is not the part that matters.
Why "AI skills" and "cloud skills" merged
Every AI product a company actually ships runs somewhere. The model is served on GPUs in a data center, the app around it runs on cloud compute, the data sits in cloud storage, the whole thing is wired together with cloud networking and secured with cloud IAM. AI is not a separate universe floating above infrastructure. It is a workload that runs on infrastructure, and that infrastructure is the cloud.
That is why "AI skills are expected" and "everything is built on the cloud" are the same sentence. The moment AI became something companies deploy rather than demo, the people who understand deployment, cost, security, and scale, cloud people, became central to it. The AI skill the market actually needs is not "can you write a clever prompt." It is "can you run this thing in production without it falling over or bankrupting us," and that is a cloud skill wearing a new label.
The AI skills that actually count (and the ones that do not)
Here is the part worth being clear-eyed about, because the hype points you at the wrong things.
Over-hyped and quickly fading:
- Prompt tricks. The elaborate prompt-crafting skill is real but shrinking in value as models get better at understanding intent. It was always scaffolding.
- Knowing this week's model rankings. Fun, low-value, stale in a month.
Durable and in demand, and mostly cloud skills already:
- Running AI in production. Serving a model, scaling inference, handling the traffic. This is just "running a workload," which you already know. The AI part is a detail on top of infrastructure fundamentals.
- The cost of AI. Token spend, GPU economics, inference cost, what things cost at scale. This is FinOps pointed at a new resource, and it is one of the most in-demand skills right now precisely because AI spend is exploding and few people can manage it. If you can already read a cloud bill, you are most of the way there.
- Securing AI. What can this agent do, what data can it reach, what is the blast radius if it misbehaves. This is least-privilege and blast-radius thinking, unchanged, applied to a new kind of workload.
- Data plumbing. Getting the right data to the model reliably, storage, pipelines, retrieval. Cloud data engineering with an AI destination.
Notice the pattern: the durable AI skills are the cloud skills you may already have, aimed at a new target. The scaffolding skills (prompts, model trivia) are the ones fading. The market is quietly re-pricing exactly the fundamentals cloud people already hold.
How to actually add "AI" to your cloud skills
You do not need to become an ML researcher. You need to point your existing skills at AI workloads deliberately. Concretely:
- Deploy one AI thing end to end. Serve a model or wire an app to a model API, on your own cloud, and take it to something that runs. The goal is not the model, it is understanding the workload shape: how it scales, where it costs, how it fails.
- Learn how AI costs money. Tokens, GPU-hours, inference pricing, what makes an AI bill spike. This is the single most leverageable skill because so few people have it and every company now needs it.
- Apply your security instincts to it. Take one AI agent or integration and ask your normal questions: what can it access, what happens if it is compromised, where is the blast radius. You already know how to do this.
- Skip the prompt-guru rabbit hole. Learn enough to be effective, then stop. That is not where the durable value is.
That is a few weekends of pointing skills you already have at a new kind of workload, not starting over.
The mindset shift
The thing I want a cloud person reading this to walk away with: "AI skills are now expected" is not a threat to you, it is a re-pricing of your existing skills upward. The industry spent a year acting like AI was a separate discipline that only ML specialists could touch. Then it turned out that shipping AI is mostly an infrastructure problem, running it, costing it, securing it, scaling it, and those are your problems. The people who feel behind are often the ones who never realized the ground moved toward them.
The take
AI skills are expected now because nearly every AI tool is built on the cloud, which means the two skill sets have merged into one, and cloud engineers are standing on exactly the ground the market is racing toward. The AI skills that actually count, running it in production, managing its cost, securing it, feeding it data, are cloud skills pointed at a new workload, not a new discipline. Deploy one AI thing end to end, learn how it costs money, apply your security instincts, and skip the prompt hype. You are not behind. You are early, and you did not even have to move.
If you are a cloud person, what is the AI skill you keep meaning to pick up, and is it actually a cloud skill in disguise? For most of us the honest answer is AI cost management, which is just FinOps we already know, aimed somewhere new.
Top comments (0)