IS YOUR AI STRATEGY BUILT ON SAND

The price you're paying for AI is not the real price. The tide is coming.

In the early 2000s, an entire generation watched as media became free and accessible by all by torrenting. Music, film, software—available on demand, at no cost, apparently indefinitely. It looked like a permanent shift. It wasn't. Rights enforcement arrived. Regulation followed. But it paved the way for the likes of Netflix, Spotify, Adobe CC. The era of free ended. The era of fair pricing began.

Most people working in that industry didn't participate in the free-for-all. They watched it happen and wondered how long it could last.

AI is in exactly that phase now. And the same question applies.

𝗧𝗵𝗲 𝗽𝗿𝗶𝗰𝗶𝗻𝗴 𝗴𝗮𝗽 

The vast majority of AI users pay nothing at all. Of the remainder, most pay a nominal amount—Claude, ChatGPT, Gemini professional subscriptions sit around $20/month. That price was not set to reflect cost. It was set to acquire users.

The economics underneath are a different conversation entirely.

OpenAI spent $9 billion to generate $3.7 billion in revenue in 2024. Anthropic's gross profit margin was negative 94% the same year. Perplexity—the AI-powered search tool increasingly used as a Google alternative—spent 164% of its revenue on the compute costs of running the service alone.

Every free tier and every $20 subscription is being subsidised by venture capital, sovereign wealth funds, and strategic investment from the platforms themselves. The pricing is not economics. It is land-grab.

Honest pricing—accounting for true infrastructure cost plus the kind of sustainable margin that makes a software business viable long term—sits somewhere between $180 and $220 per user. Per month. Per platform.

For context: I currently run between four and eight AI platforms simultaneously, experimenting across tools as the landscape shifts. I pay Adobe Creative Cloud and MaxonOne combined for less than that honest per-platform figure. I might be willing to pay full price for one AI platform that genuinely earns it. At $180-220 per platform per month, running four to eight simultaneously—as many professionals currently do—is not a sustainable personal or business expense at true cost. The market will consolidate around that reality whether the industry admits it yet or not.

That gap—between what the market is currently charging and what the market actually needs to charge—is the sand your strategy is sitting on.

𝗧𝘄𝗼 𝘁𝗶𝗱𝗲𝘀 𝗮𝗿𝗲 𝗮𝗽𝗽𝗿𝗼𝗮𝗰𝗵𝗶𝗻𝗴 𝗳𝗿𝗼𝗺 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗱𝗶𝗿𝗲𝗰𝘁𝗶𝗼𝗻𝘀

The first is the maturing tide. Regulation is arriving—the EU AI Act is gaining teeth, IP litigation is landing, data sovereignty requirements are tightening. Copyright settlements are already being reached: Anthropic settled a class action brought by authors for $1.5 billion earlier this year. The era of training on everything and pricing at nothing is ending. The industry is moving, reluctantly and unevenly, toward honest economics.

The second is the bubble tide. States and tech giants are pumping capital into data centre infrastructure faster than any previous investment cycle in history. McKinsey estimates between $3.7 and $7.9 trillion in data centre investment is required by 2030—just to build the infrastructure, before a single query is run. The four largest hyperscalers spent $413 billion on AI infrastructure in 2025 alone. Guidance suggests $600-700 billion in 2026.

Each dollar committed to that infrastructure creates pressure to generate returns. Which requires more users. Which requires keeping prices artificially low to acquire them. Which requires more investment to cover the losses. This is not a growth loop. It is a pressure loop. And pressure loops do not resolve gently.

𝗦𝗮𝗻𝗱𝗰𝗮𝘀𝘁𝗹𝗲𝘀 𝗮𝘁 𝘁𝗵𝗲 𝘄𝗮𝘁𝗲𝗿𝗹𝗶𝗻𝗲

The dot-com crash did not kill the internet. It killed the companies burning capital on the assumption that market position today would translate into profit tomorrow—without a credible path between the two. What survived was genuine utility, defensible infrastructure, and businesses that understood what they were actually selling.

The AI correction, when it comes, will follow the same shape. Consolidation. Honest pricing. And then—for the tools that deliver real returns at real prices—a sustainable market. The Netflix moment is coming. The question is which platforms earn the right to be Netflix, and which were always the free version waiting to be displaced.

For anyone building AI into business strategy, workforce planning, or technology procurement right now: the tools you are evaluating are priced at acquisition cost, not operating cost. The platforms you are committing to may not exist in their current form in three years. The capabilities you are treating as stable are shifting monthly.

Building permanent structural decisions on current AI pricing and current AI capability is building a sandcastle at the waterline.

The tide doesn't care how impressive the architecture is. It doesn't arrive with a warning. And it is coming from both directions.

Sources: McKinsey Global Institute data centre investment projections, JLL 2026 Global Data Centre Outlook, Anthropic and OpenAI financial disclosures, TechCrunch, wheresyoured.at analysis of AI company economics