The rise of AI and how to profit from it!

I have been using ChatGPT 5 with Codex to create html website pages and write scripts for apps - see UKCGT.xyz for an example.

I am truly amazed at the 'intelligence' of the system. There were no hallucinations or serious code mangling or reversals. It just worked beautifully. What I did in a few days would have taken me months in the past.

I can see why there is so much money pouring into AI, but we must remember it is mainly doing what other people have already done before. It uses designs and algorithms that other coders have already developed.

However, I have quickly run into a snag - and that is I am running out of £money!

You see, I bought a Codex Plus subscription for approx. £17/month, but what they don't tell you that this limits you to a max. of 5 hour sessions and approx. 40 minutes of ChatGPT 'thinking time'. After that, it wanted me to upgrade to a £100/month subscription or wanted me to wait a week for my next month's subscription to start.

This started me thinking. Many businesses are starting to see the power of AI. Many businesses are reducing their staff and instead, paying AI subscriptions. Only 3 junior staff are needed where 10 were needed before, because those 3 can do the work of 10 and are good enough to check the work for bugs, etc.

So I can see the AI revolution like this:

• Businesses see the enormous benefits of AI.
• AI subscriptions are taken out and many staff are made redundant.
• These businesses train their AI model and they now rely on it.
• The AI subscription prices go up (as they were loss making before just to hook you in).
• Since the business has got rid of all it's extra staff, it has to pay the higher price for AI.
• AI companies will compete for business by lowering their prices as the performance will be good enough for most people.
• Large businesses will consider buying and installing their own AI centres rather than being held hostage by an AI company. Small businesses will continue to subscribe.
• AI companies that have efficient, lower cost data centres will succeed, other more expensive centers will go bust (high capital costs = high debt, and high running costs).
• This means that cost and efficiency is the key to success for AI providers, not necessarily speed or power. AI must be affordable.

But, which of these will win and which will lose?

Picks and Shovels

However, they will all need storage (drives) and memory (esp. fast memory) and only a few manufacturers in the world can make this.

Here is ChatGPTs take on the best 'picks and shovels' companies...

⚡ Power & Energy Infrastructure

AI chips consume 3–4x more power than standard servers. The bottleneck isn't just generating power, but managing and distributing it without failure. [1]

• GE Vernova: Supplies gas turbines and nuclear services; essentially a play on the new power generation needed for AI.
• Eaton & Schneider Electric: Leaders in power management and distribution. They make the "switchgear" and uninterruptible power supplies (UPS) that prevent AI data centres from going dark.
• Caterpillar: Provides the massive backup generators required for data centre "uptime". [1, 2, 3]

❄️ Advanced Thermal Management (Cooling) []

Modern AI GPUs generate so much heat that traditional air conditioning is no longer sufficient. The industry is rapidly shifting to Liquid Cooling. [1]

• Vertiv: Widely considered the market leader in liquid cooling and thermal management for AI racks.
• Modine Manufacturing: Specialized in heat transfer technology and liquid-to-air cooling systems.
• Johnson Controls: Provides industrial-scale HVAC and water-cooling blueprints for "Gigawatt-scale" AI factories. [1, 3]

🧠 Memory (HBM) & Storage [1]

AI models require High Bandwidth Memory (HBM) to feed data to GPUs fast enough to prevent "idling". [1, 2]

• SK Hynix: The current dominant player in HBM3 and HBM3E technology, supplying the vast majority of chips for NVIDIA’s top-tier accelerators.
• Micron & Samsung: Major competitors ramping up HBM production to meet the global shortage.
• Western Digital & Seagate: The "shovels" for the massive amounts of data storage required to train and house LLMs. [1, 2, 3, 4]

🌐 Connectivity & Fibre Optics [1]

Speed between servers is as critical as speed within the chip. AI clusters require ultra-low latency "interconnects". [1]

• Corning: The world leader in the glass and fibre optic cables that form the physical backbone of data networks.
• Broadcom: Dominates the market for high-end networking chips and switches that move data between AI server racks.
• Amphenol: Recently acquired CommScope's cable business for $10.5 billion to solidify its lead in AI-specific fibre and connectors.
• Coherent: Supplies the advanced lasers and transceivers needed for ultra-fast (800G and 1.6T) data transmission.

I have invested in many of these companies and some others. I don't need to pick a few winners in the race when I can pick the companies that they all need to buy from.

As with all investments however, we need to buy at a low price and sell at a high price. Many of these companies already have forward demand priced in. Many cannot expand quick enough to meet that demand. As more capacity comes on line (new factories, etc.) stock prices and profits may drop. For now however, I think demand is greater than availability.