Are we (the retail share market) being asked to fund AI because they are short of capital but must expand to survive even though they have not defined a way to be profitable until at least 2028? What use are AI data centres if only a few can afford the subscription?
 |
| Recent and Future IPOs |
AI is truly remarkable. It can increase productivity dramatically. I can make new discoveries faster. It can be used to remind us to pick up the laundry or it can almost self-drive our car and pick up the kids from school (not).
BUT - are these companies profitable? Where will the GigaWatts of extra power come? By the time these large AI data centres are operational, will new 2027/2028 technology developments give us both faster and cheaper computing power than now?
Why is it even legal to IPO a very large company onto the stock market when it isn't even profitable and has no credible plan?
Do they really expect companies to pay them for an Agent which sends billions of requests when that company could invest in their own AI data server and not be dependent on a 3rd party?
Let's look at the long-term expectations: new custom AI chips (so what about the energy-hungry, extremely expensive nVidia chips we just bought?) Higher utilisation - only if the price is right. Smaller models - these will run on a companies own server. Tiered products and agents - i.e. more expensive (but there are many different AI models which can and will compete on price).
At the end of the day, this is a commodity business which is not scalable. If you want twice the customers you need twice the servers. It's not like software where you can sell the same software x100000 times. They could sell Agents, but I don't see companies wanting to pay for cloud-based Agents on a subscription basis with no overall control on usage or costs.
Is it not valid to ask AI - 'please design a faster, better and more energy efficient AI engine which uses cheaper hardware and train it using existing AI data and inferences.' AI model efficiency will improve and will become cheaper but not using the very expensive hardware we have just purchased.
If OpenAI, Anthropic, Google, Meta and xAI all continue investing aggressively—whether or not they are immediately profitable—they still need to buy:
- GPUs and AI accelerators
- high-bandwidth memory (HBM)
- networking equipment
- racks and power distribution
- liquid cooling systems
- backup power generation
- transformers and switchgear
- optical interconnects
- SSDs and enterprise storage
This aligns with my investment themes around companies such as Vertiv, GE Vernova, Siemens Energy, Caterpillar, memory manufacturers (Micron and Korea) and infrastructure suppliers. In other words, the AI developers are competing intensely with uncertain margins, while many infrastructure suppliers can benefit regardless of which AI platform ultimately wins. My favourite Picks and shovels suppliers include:
- Vertiv
- Eaton
- Schneider Electric
- GE Vernova
- ABB
- Arista Networks
- Amphenol
- NVIDIA
- SK Hynix, Samsung
- Micron
- Dell, HP, Lenovo, Super Micro Computer
- Teradyne
- Advantest
- Tokyo Electron
- Siemens Energy
ETF Index Trackers
Cap-weighted World Index trackers will become even more US-weighted as AI IPOs become absorbed into them.
This means your All World ETF becomes even more dependent on Tech, AI and the $USD/£GBP rate.
The $USD is high right now, but I don't think it will last. This all points to UK investors taking a large risk.
 |
| USD forecast as at June 2026 |
What to do?
I am not buying any more Global/All World ETFs (and may convert some to ex-USA ETFs).
I am also top-slicing some of my top USA performers and buying these instead:
- EXU1 - XTrackers ex-USA (distributing)
- BNKE - STOXX Euro Banks
- IUKD - FTSE dividend companies
- KORI/CSKR - Korea ETFs
- Various individual USA companies (with stop losses set)
The main reason is that AI seems to only make economic sense for certain jobs (coding, customer front-end, cyber, graphic/data analysis, etc.). For a lot of the retail market, it is just too expensive (until the cost of compute can come down by a large factor).
The USA have already banned Mythos - it is only a matter of time before the other AI engines catch up and are also hog-tied too.
The U.S. government ordered Anthropic to take down its new AI models, Claude Fable 5 and Mythos 5, just three days after their public launch on June 12, 2026. Citing national security and export control powers, the Commerce Department mandated the suspension of access for all foreign nationals, leading to a global shutdown. The move followed reports of a "jailbreak" bypassing safety guardrails, potentially turning the models into unrestricted cyber tools. White House adviser David Sacks claimed Anthropic refused to fix the issue, while Anthropic disputed the jailbreak's severity and criticized the government's opaque process. Amazon CEO Andy Jassy reportedly alerted officials to the security risks. This aggressive action sets a significant precedent for government intervention in commercially deployed AI, amidst an ongoing conflict between Anthropic and the Trump administration.
I am not saying that AI will fail, just that it is currently overpriced and they seem to be running out of (our) money (esp. if the Fed interest rate does not come down). It was bad enough when Tesla 'driverless' cars came up, but now that many people start to believe that we can put 1 million humans on Mars - well, I call that as mad as Tulip Mania, the South Sea Bubble or believing a President that has declared that peace with Iran is imminent - 38 times..
Including the period before the ceasefire, Trump has done it at least 38 times. That’s the number of times he’s said directly — in social media posts, public appearances and phone calls with the media — that a deal was nigh or claimed Iran was desperate to cut one.
My £140k deposit T212 account is up over £50k to date (starting from 10K and adding to it over 2 years), so I want to preserve that growth as much as possible.
I believe an AI bubble is coming, I just don't know when or how big and I want to be prepared.
Some areas I do see AI as being immediately profitable is in bioscience and robotics. Unfortunately, it is hard to pick the 'winners' in these two areas.
No comments:
Post a Comment