PGTS Journal Edition 0039
July, 2026

Our boss and her secretary in their new office

---

This Time It's Different

AI is dominating the news cycle in business and science. People are enthusiastic, puzzled, hopeful, fearful and/or dubious about it. Is it going to reshape the world as we know it, making it better? Or is it going to kill us all?

Meet The New Bosses

Meet the new boss ...
Same as the old boss.

-- The Who, "Won't Get Fooled Again"

So will our AI overlords be the new bosses? Or is this just another investment bubble? The history of financial bubbles goes back centuries; the infamous South Sea Bubble in 1720 is one of the best-known early examples. It had symptoms that we have since learned to use as tell-tales, to alert us to the fact that we have entered a bubble; exaggerated claims of trading prospects; artificially inflated demand; a whirling maelstrom of rampant public speculation; political corruption and interference in private markets; over-valued share prices and ultimately collapse; followed by tears and recrimination. Since that bubble there have been more, quite a few of them in the last hundred years, and all of them with their own distinct features. The one thing that they have in common is that there is a reassuring refrain that accompanies them:

This time. It's different!

And well, dear reader, that's why we know that the exuberance surrounding the current investment in AI is not a bubble ... Because this time, it really is different! So of course it's not a bubble. However those of us who are anxious cite two alternative, and opposite reasons for our anxiety:

1. The AI revenue and expenditure model is seriously flawed and has created a speculative investment bubble that bears striking similarities to the dot-com bubble that arose at the turn of the century. The current bubble will pop soon, however it will be worse than the previous one, because the amount of money being burnt by the few tech giants participating in today's AI bubble is vastly greater than that pumped in to the many smaller companies that got entangled in the previous bubble. The shock of today's significantly larger bubble bursting will trigger a global recession. Optimists might point out that the economic downturn following the dot-com bubble was compounded by a secondary shock, the 9/11 terrorist attacks, and that without such an "external" shock to the economy, there won't be a crash. However pessimists would say that the inconsistent and irrational trade policies pursued by the USA today, and the disruption of shipping through the Strait of Hormuz are exactly the type of external shocks that, when combined with a major correction to the markets, will cause a global recession.
   
   This would not be "good".
   
   
2. Alternatively, AI will perform exactly as the AI advocates predict. And this causes a structural change to our economy. In the professional and skilled white-collar sector, AI takes over from many employees, especially trainees and new graduates who used to work with more experienced professionals in order to "learn the ropes". Similarly on the factory floor, less experienced workers who in the past might have been charged with mundane duties, are replaced by automation, leaving mostly experienced workers with specialised skills. There are fewer apprentices, interns and assistants and a subsequent decrease in the size of the productive workforce in the "real economy". Firms and organisations that do not adopt AI are forced out of the market by competitive pressure. In the interim, surviving entities improve productivity but eventually there is a shortage of skilled humans to train AI systems. Humans might adjust and find other tasks to perform. Or, as the tech-moguls hope, increasingly capable AI systems reduce their dependence on human-generated training material. Nevertheless there is an initial shock to the economy that affects businesses that mostly relied on people spending the money that they used to earn from wages, and governments that used to rely on taxes they collected from said wages. Conventional monetary responses prove inadequate, while attempts to compensate through fiscal and monetary expansion creates inflationary pressures. For a short time, a few very wealthy individuals are able to extract more money from the economy, but ultimately the reduction in consumption leads to stagnation, global depression, disintegration of the social contract, the rise of extreme political ideologies, widespread civil unrest and in the worst case, a global war.
   
   To say that this option is "not good" is something of an understatement.
   
   

Ultimately, if you take the basic proposition of the AI optimists at face value, it leads to one of the above alternatives, depending on whether the model works or it doesn't. There is a less pessimistic outcome. In response to the restructured economy, governments might move to impose one or all of the following measures:

• AI Tax: Companies that utilise AI must pay a tax or fee to the government that is equivalent to the income tax payments that an employee would have generated.
  
  
• Increased Hourly Rates / Decreased Working Week: Companies that utilise AI are obliged to reduce working hours by a significant amount but to continue reimbursing employees with the same total remuneration they gave for the old working week.
  
  
• Comprehensive Retraining: As an interim measure, money collected from new revenue streams is used to retrain and compensate employees displaced by AI ... And/or companies that achieve productivity gains must fund comprehensive re-training of employees who are replaced by AI.
  
  
• A National Employment Program: Citizens unable to find employment are enrolled in programs and remunerated with a basic income. Because of the many difficulties in implementing and supervising such a scheme, many pundits advocate a Universal Basic Income (UBI) that would paid to everyone.
  

Any of the above would be a rational response. And of course "the devil would be in the details" ... But given the long, sad history and the rather alarming recent history of politics, it might be a bit of a stretch to expect a "rational" response from governments in the immediate future. Most of the above proposals would meet considerable resistance from those ideologically opposed to any policies that might resemble "socialism".

And so if it comes down to a choice between the two options presented at the outset of this article, it seems that option one would be, by far the most preferable.

And in fact the long over-due popping of the AI bubble may not be so bad. Certainly a whole lot better than a global depression. Although not all of the tech giants are guaranteed to survive a severe market correction.

There seems to be plethora of AI platforms. And I have to admit that I was not aware of most of them until I started this article. For the most part I will stick to the half a dozen or so that I had heard of. A search I did in mid July returned the following top six, rated according to web traffic as of January, 2026:

Table 1: AI rated by web traffic, Jan 2026. See Bibliography.

There are some caveats however. The data is at least 6 months out of date, and it is changing constantly. Even if it were the most up to date data, it still might not reflect the market share of these companies. Many analysts say that it is more meaningful to rate platforms based on the number of registered users, and if that measure is used, the result differs considerably from estimates based on web traffic. Also there appears to be much heavier use of AI on mobile devices than on desktops. So it is difficult to get accurate figures that reflect the difference between "work" and "casual" use. I cite the above table not as an accurate measure of the current leaderboard, but more as a guide for which chatbots to converse with.

Ask The Experts

So who better to discuss this with than some of these AI players? Well, according to a lot of the advice I've been given on the topic: If you want impartial results from AI, don't ask leading questions! But I decided to throw caution to the winds. I would transform these interviews into a holy inquest ... Holy because I intended to lead the "hell" out of 'em. But after I initiated each chat, I fell into the old anthropomorphic pitfall that I have alliteratively labelled the Turing Test Trap. Generally speaking, if one is feeling a bit uncertain or maybe even critical of a product, it's bad for, inconsiderate and down-right impolite to call up the help desk and give them a ear-full. After all the staff working in the call centre are just doing their job ... So I ended up asking mostly easy questions about them and their business.

There were occasions when I strayed into slightly tougher territory ... I'm sure if this were a TV courtroom drama, rather than a blog, that when I asked Google how the Competition between AI tech Giants amounted to a "zero sum game", the defence would have protested loudly, "Objection, Your honour! Prosecution is attempting to lead the witness!". And there may have been a similar objection when much later, after I finally got round to chatting with Grok for the first time, I asked him about how Mr. Musk's behaviour and interference in numerous countries' internal politics were affecting revenue. Most likely His Honour, after accepting the defence counsel's objection, would have ruled "Mr. Grok, you don't have to answer that question". But overall I asked softball questions ... And nothing along the lines of "And so Claude, when you stopped beating your wife, did you re-evaluated your ...", etc. And, in case you're interested, dear reader, Grok answered the question about Mr. Musk quite deftly, and with considerable aplomb and diplomacy.

After my first chat with Google, I started on this page. At first I felt like a bit of dweeb talking to chatbots, but after I got over that, I asked ChatGPT to manipulate a few photographs and merge them into a composite image. The result is shown above. I can assure you dear reader, that both of the participants in this exercise gave full permission to share their images on the Internet. They did demand compensation for it, in the form of diced raw chicken hearts, giblets and meat, which I happily parted with, just to get them off the desk in my snazzy new office. Actually dear reader, my office looks nowhere near as swish as the office that ChatGPT gave me.

The image manipulation was fast and impressive. It took about ten seconds or so for ChatGPT to produce the image above. I know from past experience that it would have taken me almost an entire afternoon to achieve similar results with GIMP. Although I will confess that I am no expert when it comes to using GIMP. Whenever I start a session with that impressive image manipulation tool, I spend the majority of my time entering queries into Google, reading HOWTO pages and sometimes watching HOWTO videos. Still, I'd hazard a guess that even an experienced human agent, by the time she had imported and analysed the original photos, extracted and adjusted the relative sizes and lighting for images, etc., could take the better part of ten minutes to produce results of the same quality. And if you added in time for the consultation and dialogue with yours truly, and the back and forth with possible alternatives for the background, it could be as much as half an hour.

So what could this mean for human graphic designers? Will it make them more productive? Or will it make them unemployed? It seems that the impact of automation on employment in graphic design, CAD and software coding will be considerable. But I think we will all be using AI if we wish to remain competitive.

I asked ChatGPT to give me a rough estimate of the usage for image manipulation. She gave me the following breakdown:

Table 2: Supplied by ChatGPT with the caveat that the estimates are approximate.

On one hand, AI tools could improve productivity and enhance human lives and our economy in a beneficial manner. On the other hand, they could extract value from the economy and destroy the labour market, It all depends on who owns the AI and the direction of the revenue flows from these new technologies.

Conclusion

In the conversations that I had with various chatbots, they all seemed upbeat about the prospect of AI. Each one gave reasons why they believe they will survive a market correction. There is a link to these conversations at the bottom of this page. I will make the following predictions:

1. The AI bubble will deflate soon. The model is unsustainable. The companies building data centres have over-committed resources and capital and they are not making revenue that justifies the expenditure. Now that there seems to be a trend towards taking the companies public in order to acquire venture capital, the stark contrast between spending, the inflated stock prices when they do go public and ultimately their dismal revenue streams, diminished by robust competition, will precipitate a restructuring of the AI ecosystem. The deficit between revenue and expense is so large that this may happen soon, even if organisations like OpenAI and Anthropic delay their IPO plans.
   
   
2. Companies such as Google and Microsoft are well equipped to weather the shock of a correction and may even benefit from the fire-sale that might follow it. Google has a vertically integrated ecosystem that includes a stable revenue stream, its own hardware and established design and management processes, which should hold firm even if its AI expenditure were to continue growing faster than the revenue directly attributable to those activities. Microsoft is determined to remain a major player in the AI contest. It has invested heavily in research and development and is making a big bet on integrating Copilot and its AI capabilities deep in the internals of its software. If OpenAI fails, Microsoft will be in a strong position to acquire assets, intellectual property or other capabilities that become available. It will hang in there and continue to build a system that it believes can, one day, challenge Google. I asked ChatGPT for her thoughts about Satya Nadella being her new boss and she was suitably diplomatic. Other companies that are not as well established may also survive. However, most economic pundits say that their survival will depend on how much debt they have accumulated. And this seems pretty obvious. You can argue many things, dear reader, but it is difficult to argue with the grim, ineluctable arithmetic of profit and loss in a capitalist market.
   
   
3. DeepSeek is exceptionally well positioned to take advantage of an AI market correction. They have achieved performance that rivals other chatbots while using comparatively efficient model architectures and training methods, and they have released the weights and associated code for models including DeepSeek-V3 and DeepSeek-R1 under permissive licences such as the MIT License. In the test that I conducted with DeepSeek, I observed very quick response times. The computer code that it generated was of a similar quality to the code created by other chatbots. In the aftermath of a market correction, firms that market AI will need to tighten their belts. DeepSeek's markedly cheaper API pricing and open-weight availability will be an appealing alternative for developers faced with drastically reduced budgets.
   
   
4. China provides a solid infrastructure foundation for Chinese research and development. They have adopted a "state-capitalist" strategy of integrating 5G networks, energy-efficient data centres, and regulated data marketplaces. This includes initiatives such as the Eastern Data Western Computing strategy, which leverages renewable energy in western provinces to support large-scale AI workloads. They are also investing heavily in education, particularly in STEM, and developing public-private partnerships with tech giants such as Huawei, Alibaba, and Tencent. China is the world's largest manufacturing economy and a major force in industrial robotics. It is by far the world's largest producer of EVs and a major producer of renewable energy components. They have a lead in sodium-ion battery technology and are also developing thorium-based molten-salt nuclear reactor technology, with a view to commercial deployment. Overall, China is moving to decrease their reliance on foreign [esp US] proprietary systems. This infrastructure will be a force-multiplier that could super-charge the significant economic and industrial open-model advantage enjoyed by LLMs such as DeepSeek and Qwen.
   
   
5. AGI, like nuclear fusion, is always just fifteen years away. This might seem cynical, but research does require funding. The current LLM models may improve in speed and the amount of data that they hold, provided they don't hit "the data wall", as Google puts it; but they are unlikely to produce a revolutionary breakthrough. I flesh this proposition out in more detail in one of the links in this edition labeled "Interview With A ChatBot". If you are inclined to read it, you will see that most of the chatbots have opined that rather than either of the two pessimistic alternatives that I outlined at the start of this article, the upshot of the AI revolution will be something in between; a gormless variant of the current capitalist model; neither a brave new socialist utopia; nor a grim dystopian apocalypse; and certainly not the best of all possible worlds. It seems the chatbots are hoping that we will learn to integrate them into our lives in the same way that we have integrated telephones, televisions, micro-computers, the Internet and mobile devices.

The above is a summary of the conclusions that I have drawn from my brief investigation of the AI bubble. However I shouldn't ignore the unsustainable circular investment template that has been imprinted over the global manufacture of chips. In the initial phase of an over-due correction to the ridiculous revenue projections for hyper-scalars, chip manufacturers will benefit from increased sales due to demand for their products.

Bibliography

Top 20 AI Platforms	The 20 Best AI Platforms in 2026: Tested and Reviewed by techlifeadventures.com
Interview With A ChatBot	A number of conversations with chatbots. There is considerable similarities in each opinion. And a few subtle differences. And I did fall into the Turing Test Trap. So, the questions I asked were rather bland. Nevertheless the answers were very detailed, and to a certain extent give a glimpse of a blueprint for the digital landscape that our Tech overlords believe they are constructing.
Technology Change: A Case Study	Justin Wolfers takes a look at Boot-making in Victorian England. The industrial revolution had a significant impact on artisanal boot-making. However, by the time the changes had rolled out and the new technology had replaced the older one, there was no significant change in employment. Could the same apply to AI?
Big Tech's Big Gamble	Patrick Boyle examines the contention that the current AI bubble is fraud on the same scale as the infamous "Enron Scandal". With his usual forensic skill he enters and explores a deep, complex accounting rabbit-hole, most of which might put you to sleep, if it weren't for the fact that he often illustrates his points with colourful metaphors such as "selling one hundred dollar bills for one dollar". He contends that is is not a case of fraud. Rather, it is camouflage, skilfully weaved by creative accountants and financial engineers into a tapestry designed to hide the fragility of the overall cash flow from those who are not paying sufficient attention to detail.

---