A two-year-old startup just raised $500 million in a single funding round. They don't build chatbots. They don't build language models. They build the networking hardware that connects AI chips inside massive data centres.
They're the motorway network that makes all of this work at scale. And they were oversubscribed.
Here's what matters from this week.
The money is going into the plumbing
Next Hop AI closed $500 million in Series B funding, pushing their valuation to $4.2 billion. They were founded two years ago. Their job is building the switching systems that let AI models actually run inside data centres.
Meanwhile, Alphabet, Amazon, Meta and Microsoft are expected to spend around $650 billion on AI data centres in 2026 alone.
You're obviously a long way from a data centre. But every time a major infrastructure player raises this kind of money, it means the tools that sit on top of that backbone get faster, cheaper and more reliable for everyone downstream. Including you.
The businesses already building AI into their operations will be miles ahead when the next generation of tools drops on top of this infrastructure. Running a business in 2026 without AI in your operations is a bit like running one in 2010 without a website. You can survive for a while. But the gap is widening every month.
The question isn't whether you'll eventually get there. It's how far behind you'll be before you start.
NVIDIA's big week tells you where this is all going
NVIDIA's GTC 2026 conference kicked off this week in San Jose. It's become the most watched annual event in AI, and it's not just a chip conference. It's the week where the world's most powerful AI company lays out what's next for AI capability.
The announcements are covering chips, software, AI agent platforms and major enterprise partnerships. Eli Lilly launched what NVIDIA is calling the most powerful AI factory wholly owned by a pharmaceutical company, built to accelerate drug discovery.
What's happening in pharma is genuinely wild. Smart labs that can run experiments without a human involved, analyse results, make decisions and run new tests. If that level of automation is possible in drug discovery, think about what's possible in simpler businesses with simpler processes.
The pattern I've seen over the years working alongside senior leaders is this: the businesses that get ahead aren't the ones who move fastest when the technology is shiny and new. They're the ones who have built the internal habits and the adoption mentality to absorb new tools when they arrive. That's a leadership challenge, not a technology one.
OpenAI is already retiring its own models
As of yesterday, OpenAI retired GPT 5.1 entirely. Thinking and pro variants, all gone. Existing conversations have been migrated to newer versions automatically.
This is the era of disposable AI versions. Models are improving so fast that the tools you're using today will be replaced within months. That's not a reason to wait. It's the opposite.
Alongside the retirement, OpenAI rolled out a new feature called Skills for business and enterprise users. Skills are essentially repeatable AI workflows. You give it instructions and files, your SOPs for how a thing gets done, and it follows the process every time.
Think of it like this: if someone in your team does a task the same way repeatedly, you can now build that into a skill. The AI follows the process, and you get consistency without the bottleneck.
This is where things start getting practical for small businesses. Not the headline-grabbing stuff, but the quiet automation of the repetitive work that currently lives inside somebody's head or somebody's inbox.
Will