The AI industry is spending like it is 1999. What a small business should take from that
The numbers coming out of the AI industry stopped sounding like software and started sounding like national infrastructure. The useful question for a small business is not whether the spending is wise, but what it means for the tools you have quietly become dependent on.
What is being reported
As of 27 July 2026, the Wall Street Journal reported — summarised by Financial Express and others — that Nvidia is in talks to provide a financial backstop of roughly $250 billion to help OpenAI lease a very large data centre. Around the same time, OpenAI's own enterprise update said its APIs process more than 15 billion tokens per minute and enterprise now exceeds 40% of its revenue.
Treat the exact figures as reported rather than settled — deals of this kind change. The direction is what matters: staggering sums going into compute, and the providers of your AI tools locked into spending that has to be recovered from customers eventually.
Why a boom this size affects your small business
You are not investing in any of this. But you probably rely on it, and booms create three specific risks downstream.
| Risk | What it looks like for you |
|---|---|
| Price volatility | Today's price war (covered in our AI pricing analysis) is real, but subsidised. Free and cheap tiers exist partly to capture the market. They can tighten once the market is captured |
| Product churn | Fast-moving vendors deprecate features, rename products, and change terms. What you built on last year may not exist next year |
| Concentration | A handful of companies now underpin most AI tools. A problem at one propagates widely — see the recent cloud outages |
What history rhymes with
The late-1990s internet build-out created enormous, genuinely useful infrastructure — and also a great deal of capital destroyed by companies that assumed the growth curve was permanent. Both things were true at once. The businesses that came through best were not the ones that bet biggest; they were the ones that used the cheap infrastructure while it lasted without becoming dependent on any single provider surviving.
That is the posture to copy. Use the subsidised AI tools now, because they are genuinely cheap and capable. Just do not architect your operations so that a price change or a shutdown becomes an emergency.
What to actually do
- Keep AI swappable. Prompts and logic separate from the specific model, an evaluation set of real tasks ready to run against any provider. Adopting or dropping a model should be a config change, not a rebuild.
- Prefer monthly billing. In a volatile market, annual prepay locks you into today's price and today's vendor. The discount rarely covers the lost flexibility.
- Know your manual fallback. For any workflow with an AI step, write down what happens when the tool is unavailable or unaffordable. If there is no fallback, that is a single point of failure.
- Do not over-invest in one vendor's ecosystem. Deep integration with one provider's proprietary features is convenient until their strategy changes.
- Watch for the tightening. When free tiers shrink or usage caps appear, that is the subsidy ending. Have your alternative measured before you need it.
What not to conclude
- This is not a reason to avoid AI. The tools are useful and cheap right now. Use them, following our framework for choosing AI tools.
- It is not a prediction of collapse. Nobody credible knows how this resolves. The point is resilience, not forecasting.
- It is not urgent this week. If your AI use is a chat subscription for drafting and research, none of this changes your day. It matters when AI becomes load-bearing in a workflow.
The honest summary
Enormous sums are being spent to make AI cheaper and more capable, and you get to benefit without spending any of it. The only trap is mistaking a subsidised, fast-changing market for a stable utility. Use the tools, keep your data and logic portable, and make sure no single vendor's decision can take your business down with it.
No company paid for placement in this article. Verify current prices and terms with each provider before buying.