Everything in technology is supposed to get cheaper every year. We think that is ending, for storage and for AI, and that small firms and new graduates will feel it first. This is opinion, not fact. Here is what we are seeing, what we expect, and where we could be wrong.
In short
Storage has stopped getting cheaper, and we do not think it is a blip.
Unlimited AI for a flat fee is on the way out. The best tools are heading for a meter.
Graduates are being hired for AI skills that many employers will then ration. That gap worries us most.
1. Storage stops getting cheaper
For decades the cost of storing a terabyte fell every year, so nobody had to think about it. Keep everything, for ever, in three places. Why not?
In the past year that has reversed. Hard drive and flash prices have risen sharply, because the data centres being built for AI are buying everything the factories can make. Several cloud storage companies have put their prices up this year, and Microsoft raised its business prices in July.
Our prediction. Most analysts expect prices to settle from 2027, once new factories open. We are less relaxed. We think AI's appetite will grow as fast as the new supply does, and that storage will keep getting dearer for years, possibly steeply.
Where we could be wrong. The factories may arrive faster than the demand. It has happened before.
What to do either way. Know how much you store and what it costs each month. Stop keeping everything by default. Move finished work to an archive you own, with a backup, where the cost is paid once.
2. AI goes on a meter
The AI tools most of us use were launched at a flat monthly price that did not cover what they cost to run. That was the offer that got everybody hooked.
This year the meter arrived, and look at who is holding it. Uber is reported to have used its whole year's AI budget by April and now caps each employee at about US$1,500 a month per coding tool. Apple is reported to have cut its engineers to US$200 of AI a day. Tesla is reported to have set a weekly limit that needs a manager's sign-off, and every division at Microsoft is reported to have a budget. Accenture funds a set amount of AI for each employee and charges anything above it back to the project, although it says that is not a hard cap. When firms of that size start counting, the free lunch is finished.
Our prediction. The very best AI becomes something only large companies can afford to use freely. Everyone else gets a ration, or a cheaper tier.
Where we could be wrong. The price of a given amount of AI has fallen a long way, and smaller models that run on an ordinary computer improve every few months. The future may be two tiers: cheap and good enough for most, costly and best for a few.
What to do either way. Do not build a way of working that only holds up with unlimited AI. Know which jobs truly need it. Keep the skills to do the work without it.
3. The graduates caught in between
More than a third of entry-level job adverts in the United States now ask for AI skills, according to the graduate employers' body NACE, up from about one in eight the autumn before. Very few employers train for them.
Put that next to prediction two and there are two gaps, not one.
The first is the graduate who learned to work with AI at university and arrives at a firm where it is rationed, restricted or switched off. They were hired for a skill they are not allowed to use.
The second is quieter. It is the graduate who leaned on AI all the way through and never built the skill underneath. The junior tasks that used to teach it have gone to the machine, so nobody is teaching it at work either.
A studio needs juniors who can do the thing and can also judge the machine's version of it. That takes deliberate training, and it will not happen by accident.
What to do. When you hire, ask to see work done both ways. Give new starters real tasks before AI-assisted ones. Write your house rules for AI down. Ours are here.
Why an IT firm is saying this
Because all three lead to the same advice. Own your data, know what it costs, and keep the skills in the building. This week's Microsoft outage was a small rehearsal for the first of those. We wrote about it here.
Want to know what you are storing?
We can tell you what you hold, where it lives, what it costs today and what owning a copy would look like. No padlocks, no scare stories. Talk to us, or read about backup and disaster recovery.