Why does AI infrastructure investment matter for small business owners?

When major investment goes into the infrastructure that powers AI tools, the tools that sit on top of that infrastructure get faster, cheaper, and more reliable. That means the AI tools available to small business owners will continue to improve significantly over the next few years. Owners who are already building AI habits now will be better placed to benefit when those improvements arrive.

What is the OpenAI Skills feature and how can small businesses use it?

Skills are repeatable AI workflows that tell an AI tool how to perform a specific task in a consistent way. Think of them as SOPs for AI. If someone in your business does a task the same way every time, that process can be documented and given to an AI tool as a skill. The tool can then detect when that skill is needed and execute it automatically, without being prompted. It is available in beta for business and enterprise users.

Is it too late for small business owners to start using AI?

No. But the gap between businesses that have built genuine AI habits and those still waiting is widening every month. The more useful starting point is not whether it is too late, but which task in your business follows a consistent enough sequence that an AI tool could begin to handle it. Starting with one specific, repeatable process is more valuable than waiting until you feel ready to overhaul everything at once.

Why do AI models keep changing and should I be worried about it?

AI models are updated and replaced regularly because the technology is improving quickly. OpenAI retired GPT-4.5 this week and migrated users to newer versions. This is not a problem. New models are better. But it does mean that if you tried an AI tool months ago and found it inconsistent or frustrating, that specific version may no longer exist. The tool available today is likely meaningfully better than the one you tested.

What is NVIDIA GTC and why does it matter to business owners?

NVIDIA GTC is an annual conference where NVIDIA, the most powerful company in AI hardware, announces its roadmap for chips, software, and AI platforms. Every AI tool a business owner uses runs on top of NVIDIA's infrastructure. What gets announced at GTC shapes what AI tools will be capable of in the next twelve to twenty-four months. It is the clearest signal available for where AI capability is heading.

What This Week's AI News Actually Means for Small Business Owners

Three big AI stories dropped this week. A two-year-old startup raised $500 million. NVIDIA kicked off the most watched event in the AI industry. And OpenAI quietly retired one of its flagship models and replaced it with something new.

None of these stories were written for you. They were written for investors, developers, and enterprise technology buyers. But each one has a direct implication for how you run your business in the next twelve to twenty-four months.

Here is what actually matters.

The $500 Million Bet on AI Infrastructure

Next Hop AI, a two-year-old startup that builds the networking hardware inside AI data centres, just closed a $500 million Series B round. It is now valued at $4.2 billion. The round was oversubscribed. Two years old.

They make the switching systems that connect AI chips inside massive data centres. Think of it as the motorway network that allows AI models to run at the speed and scale they need. Alphabet, Amazon, Meta, and Microsoft are expected to spend around $650 billion on AI data centre infrastructure in 2026 alone. Next Hop is positioning itself to capture a slice of the networking market that sits underneath all of it.

The reason this matters to you has nothing to do with hardware.

The money is no longer flowing only into the chatbots and language models you interact with directly. It is going into the backbone that makes those tools faster, cheaper, and more reliable. Every time a major infrastructure player raises this kind of capital and pours it into the plumbing, the tools that sit on top of that plumbing get better for everyone downstream.

The business owners who are already building AI into their daily operations right now will be significantly ahead when the next generation of tools arrives on top of this infrastructure. Not marginally ahead. Significantly ahead. Because they will already know how to use the tools, already have the habits, and already understand how AI fits inside their specific business.

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 between you and the businesses that have built these systems properly is widening every week.

The question is not whether you will eventually get there. It is how far behind you will fall before you do.

NVIDIA's Annual Conference and What It Signals

NVIDIA GTC 2026 is running this week in San Jose. It has become the most watched annual event in the AI industry. Jensen Huang delivered the keynote, and the live announcements are covering chips, software, AI agent platforms, and major enterprise partnerships.

Two things stood out.

First, NVIDIA announced a multi-year strategic partnership with Thinking Machines Lab to deploy at least one gigawatt of its next generation Vera Rubin system. Second, Eli Lilly launched what NVIDIA is calling the most powerful AI factory wholly owned by a pharmaceutical company, built to accelerate drug discovery.

The pharmaceutical example is worth sitting with for a moment. The smart labs being built in that industry can run experiments without human involvement, analyse results, make decisions, and run new tests in sequence. Some of the predictions around what AI can do for new drug discovery are extraordinary, including the potential elimination of certain diseases within a generation or two.

If that level of automated experimentation is possible inside one of the most complex industries in the world, the question worth asking is what could something similar look like inside a simpler business, running simpler processes.

GTC is not really a chip conference. It is the week where the world's most powerful AI company sets out what comes next for AI capability. Every tool you use in your business runs on top of what NVIDIA is announcing this week.

When the industry talks about the era of AI factories, it is not just describing pharmaceutical companies or hyperscalers. It is describing a shift where every serious business will have AI systems running in the background, doing work that currently lives inside someone's head or inbox.

The businesses that typically get ahead in these moments are not the ones who move the fastest when the technology is shiny and new. They are the ones who have already built the internal habits and have an adoption mentality that lets them absorb new tools when they arrive. That is partly a technology question. But it is also a leadership question. How do you prepare the people in your business to use these tools well? How do you shift your own thinking about what AI can take off your plate?

Those are not technical problems. They are human ones.

OpenAI Retired a Model and Rolled Out Something Worth Knowing About

As of this week, OpenAI retired GPT-4.5, including its thinking and pro variants. All closed down. Existing conversations migrated automatically to newer versions depending on the plan. It happened quietly, mid-week, while most people were focused on other things.

This is worth noting not because of the specific model, but because of what it signals. The models your team learned to use six months ago have been quietly replaced. The version someone in your business got comfortable with, built their prompts around, or used to draft their client reports, that version may no longer exist.

This is not a problem. The new versions are better. But it does mean that if you tried AI once, found it frustrating or inconsistent, and filed it under things to revisit later, the tool you tried is not the same tool that exists today. That is worth knowing.

Alongside the model retirement, OpenAI also rolled out a new feature called Skills in beta for business and enterprise users.

Here is the clearest way to describe what Skills are.

If someone on your team does a task the same way every single time, they have a skill. That skill can be documented. Once it is documented, it becomes an SOP. Once it is an SOP, it can be given to an AI tool. And once the AI tool has it, the tool can execute that task repeatably and accurately, without needing to be told how to do it each time.

What makes Skills different from simply giving an AI a prompt is that the tool can detect when a skill is required without being asked. It identifies the context, pulls in the relevant skill automatically, and executes accordingly. You do not have to tell it to work that way. It just does.

Think about what that means practically. The process your operations manager follows every Monday to compile the weekly report. The way your team qualifies a new enquiry before it gets to you. The sequence your admin uses to onboard a new client. Each of those is a skill. Each of those can be documented, handed to an AI tool, and run without someone having to think through the steps each time.

Skills can also be shared across an entire workspace, which means one person building the process once benefits everyone using the same tool.

Claude, the AI assistant from Anthropic, has a version of this feature too. Most of the leading models are moving in this direction. It is the natural next step from prompting an AI to ask it to do something, toward building AI that already knows how your business works and operates accordingly.

What to Actually Do With This

Three stories, one theme. The infrastructure underneath AI is being built at scale and speed. The tools sitting on top of that infrastructure are improving faster than most people are tracking. And the features arriving now, like Skills, are starting to look less like clever assistants and more like systems that can carry real operational weight.

The gap between businesses that have built genuine AI habits and those that are still watching and waiting is not closing. It is widening.

If you want to start somewhere practical, pick one task in your business that follows a consistent sequence. Something that gets done the same way most of the time. Document how it gets done. Then test whether an AI tool can run it using that documentation. That single experiment will teach you more than an hour of reading about AI ever could.

If you want to go deeper on how to actually build this into your business without it becoming another project that stalls, get in touch and we can have a conversation about where it makes sense to start.