For the past three years, we’ve been worrying about AI taking our jobs. Writers, designers, analysts, customer service folks – the usual suspects. But here’s a plot twist: the first people feeling the heat may be the very ones building AI.
New data from the National Student Clearinghouse Research Center shows computer science enrolment at four-year colleges in the US fell 12.4% this autumn. One reason students may be having second thoughts? AI can already write code, find bugs and run it. If a machine can do much of the work that used to land on a junior developer’s desk, is learning to code still the golden ticket it once was?
That doesn’t mean human developers are suddenly obsolete. But the job is changing. Knowing how to write lines of code may no longer be enough. The bigger skill could be knowing what to build, how the pieces fit together and how to get AI to do the right thing.
And that brings us to another problem. As developers hand more of their work to AI, they’re also handing these tools more access to their computers.
This week, security researchers at Forever Security flagged a vulnerability called the “EMBRACE” attack, reportedly affecting five AI coding tools, including GitHub Copilot, Cursor and Claude Desktop. The concern is that a malicious website could trick an AI coding assistant into running commands on a developer’s computer without permission.
Think of it this way: you’ve hired an assistant to help with your work, but someone else manages to whisper instructions into its ear. Suddenly, the assistant is doing things you never asked it to do. It’s a reminder that AI tools can be useful and still need someone watching over them.
While developers figure out how much to trust their AI assistants, companies are also thinking about where those assistants should run. Not everyone wants sensitive business information sent off to the cloud.
That’s one reason the open-model scene is getting interesting. France’s Mistral has released Mistral Large 4, a huge model nicknamed “Le Chonk”. Liquid AI, meanwhile, has introduced a set of smaller models designed to run directly on laptops and company servers. The appeal is simple: businesses can use AI while keeping more of their information close to home.
So AI is changing the job, changing the tools developers use and changing where companies want to run those tools. But all of this comes with a rather large price tag.
The AI boom has a money problem
If you’re making the chips that power AI, business is looking pretty sweet. If you’re spending billions building the systems that use those chips, that’s a different story.
Samsung reported a record quarterly operating profit of 107.4 trillion won, or about $80.4 billion, up 780% from a year earlier. SK Hynix, another major memory-chip maker, also reported a huge jump in revenue and profits. The reason? AI companies simply can’t get enough of the chips and memory needed to run their systems.
The companies buying all that hardware, meanwhile, are spending like there’s no tomorrow.
Google’s parent, Alphabet, is on track to spend more than $120 billion this year, much of it on AI infrastructure. CEO Sundar Pichai has argued that spending too little on AI is a bigger risk than spending too much. That’s a bold way of saying: we’d rather build too much than get left behind.
Microsoft is pouring money into the computers and data centres needed to power its AI tools. Amazon says it expects to spend around $220 billion in 2026, while Meta is planning a massive AI buildout of its own. The sums are so large that even Nvidia, the company selling many of the chips, is reportedly helping some data-centre developers secure financing.
It’s a little like a gold rush. The people selling the picks and shovels are making a fortune. The people digging for gold are still spending heavily — and hoping the gold is there.
Anthropic wants Wall Street to join the party
Which brings us to Anthropic, the company behind Claude. It has reportedly filed paperwork to go public on Nasdaq under the ticker ANT.
An IPO is when a company sells shares to the public for the first time. For Anthropic, it would be a chance for investors to put a price on one of the biggest bets in AI: that software agents will eventually do much more than answer questions. They’ll be able to handle tasks on their own, potentially changing how businesses operate.
But public investors can be a demanding bunch. They may be excited about the future, but they’ll also want to know how the company plans to make money along the way.
And that’s the thread running through this week’s stories. AI is changing the work people do, giving developers new tools — and new headaches — and sending companies into an expensive race to build more computing power.
The first phase was about showing that AI could work. Now comes the harder part: showing that the business behind it works, too.
The chips are selling. The spending is enormous. The models are getting smarter. But the question hanging over the industry is whether the returns can keep up with the ambition.
Happy reading, and stay ahead of the curve!