How Business Angels Profit from AI Beyond Startups
In short, data is no longer free — and that’s creating space for new aggregators, from telcos to marketplaces to social platforms, to turn data licensing into a revenue stream.
Even technology has its fashion moments. In Q1 2025, AI startups owned 58% of all global VC funding — that’s 15 times more than the once-trendy crypto sector. But before you run off to back the next AI image generator for Amazon sellers, pause. Most of that capital isn’t going to shiny apps — it’s flowing straight into companies building foundational AI models. And those companies, in turn, spend their investor money on… compute.
Let’s break that down. The bulk of global Q1 2025 investment ($40 billion) went to OpenAI. Another $3.5 billion to Anthropic. Unless you’re a sovereign wealth fund, you’re likely not getting in on those rounds. So where can you invest to ride the AI boom? Here are three less obvious ways.
Investment Map: AI related sectors. Source: Blank
Data centers: the real estate of the AI era
The infrastructure behind AI is overheating — literally. Commercial data centers are struggling to keep up with demand, and the gap is only widening. By 2030, 8% of electricity in AI-leader countries will be consumed by data centers, with 20% of their capacity devoted purely to AI processing.
The global data center market is currently worth $210 billion and is expected to grow 8–10% CAGR through 2028. While the U.S., UK, Germany, and China dominate, new facilities are rising in India, Brazil, Mexico, and Saudi Arabia. Considering it takes 4–5 years to launch a new center, this is prime time for early investment.
Biggest risk? Energy. So much so that in the U.S., the main barrier to building new data centers is just getting access to the grid. Cooling systems (especially chillers) are the biggest energy hogs. That’s why innovative methods like adiabatic cooling — basically spraying water mist to cool down servers — are hot in hot countries. Oil-rich regions are also experimenting with repurposing flare gas to power data centers.
Smart software for smarter infrastructure
Raw server space isn’t enough. What matters now is efficient use of that space.
That’s where DCIM ( Data Center Infrastructure Management) comes in. This market hit $6.6 billion in 2020 and has been growing 20% annually. Once reliant on RFID tags and handheld readers, DCIM is now evolving into DMaaS (Data Center Management as a Service) powered by IoT and cloud analytics. By optimizing airflow, predicting energy spikes, and automating resource allocation, these systems can unlock 10–30% more capacity from existing centers.
Modular prefab data centers are also gaining traction — especially in locations where it’s hard to build full-scale facilities. Think of them as the Airbnb of server space: plug-and-play infrastructure ready to scale AI needs fast.
The metal behind the machines: copper
AI isn’t just consuming teraflops — it’s devouring raw materials too. Chief among them? Copper.
Training large AI models demands massive data centers, which in turn need vast amounts of wiring and cooling systems — all of which rely heavily on copper. By 2030, AI could push copper demand up by 1–5 million tons per year. BHP forecasts that AI and data centers will add 3.4 million tons of annual copper demand by 2050. That’s a 72% surge from 2021 levels, driven not just by AI, but also by the parallel boom in renewable energy — another copper-hungry sector.
Today, data centers make up less than 1% of global copper demand. But that’s expected to climb to 6–7% by 2030–2040.
From free to fee: the business of training data
AI training data is becoming big business. As models scale, access to clean, high-quality, labeled data becomes a bottleneck — especially as platforms restrict what can be scraped.
Forget scraping Wikipedia. Companies are now licensing data — and it’s not cheap: $1–2 per image, $2–4 per short video, $100–300 per hour-long video. The training data market is currently worth $2.8 billion and is growing at nearly 28% annually. By 2029, it could hit $9.5 billion.
Legal battles are already shaping this space. The New York Times is suing OpenAI for using its archives. Getty Images is going after Stable Diffusion. But others are cashing in: Reddit struck a $60M deal to license its API to Google, Shutterstock signed a six-year deal with OpenAI, and StackOverflow is monetizing its code database.
In short, data is no longer free — and that’s creating space for new aggregators, from telcos to marketplaces to social platforms, to turn data licensing into a revenue stream.
Final Thought
AI’s future is notoriously hard to predict. Generative models exploded faster than anyone expected. The next breakthrough could come from anywhere. But while you wait for the next big thing to be invented, there’s a quieter — and maybe more profitable way to play the AI game: by investing in the tools, energy, copper, cooling, and data that make AI possible in the first place.
