Big tech has committed hundreds of billions of dollars to AI, fuelling debate over whether markets are witnessing a bubble or a fundamental shift in the global economy. Joining us today is Dr. Ryan Lemand, Founder and CEO of NeoVision Group. Ryan, welcome to Wall Street to Mena.
Thank you for having me, Rachel.
Big tech has poured hundreds of billions into AI infrastructure. Some strategists call this bubble territory. Others say the AI trade still has room to run. Are we in a genuine bubble or is this a structural repricing of the economy?
Having lived through the year 2000 internet bubble, I can confirm that we are in a bubble today. From a valuation perspective, I do not think there are two analysts who disagree that we are in overvaluation territory. The second way we evaluate bubbles is when companies are overvalued with no clear path to profitability. Let me give a concrete example. I currently pay $200 a month for my AI subscription, but I consume roughly $5,000 worth of tokens each month. Since the beginning of this year, I have consumed 1.3 billion tokens. That difference of $4,800 every month is being financed by VCs and hyperscalers like Microsoft and Alphabet. This is unsustainable. As long as AI companies cannot charge me $5,000 a month or lower the cost per token to match — and that is not close to happening — we are still in the first generation of this technology. Just like 2000, we were in generation one and most of those companies are gone. We will evolve into more profitable, more efficient companies.
You have been using machine learning for portfolio optimisation and NLP for client engagement. Where is AI genuinely changing how wealth managers allocate capital?
It is the speed and access to information. Today this is a quantum leap compared to what we had before. When we receive a complex client inquiry, we used to need an analyst to work on it, produce a note, and send it. Today that is almost instantaneous thanks to AI. But we still need that analyst. All this talk about job destruction is completely incorrect in my view. It is very similar to the Industrial Revolution — everyone said machines would destroy jobs, but the reverse happened. Job growth followed, because to operate those machines you needed the specialists who used to produce things manually. The analyst today has to upgrade their skills to use AI to produce information and analysis for clients. The big difference is speed and access to information — that is what has completely changed since 2023.
How are the US and China's AI approaches differing and what are the downstream implications?
There is a big difference. The US has 5,500 data centres, of which 20 to 30% are not yet operational because there is not enough energy to supply them. The US approach is centralised data access through massive data centres. There are two problems: energy access and technology rollover — every six months a new chip comes out and these massive centres have to upgrade regularly. China's approach is completely different — they are pursuing small language models rather than large ones. China has just 500 data centres. They focus on a decentralised approach where each unit has its own mini data centre based on a small language model. China has also worked aggressively on energy. Gigawatt energy production in China is going vertically up — energy expansion there is much higher than anything we have seen in history. Whereas US energy production is essentially flat. Even Microsoft is now switching to Chinese models for many applications — because they are less expensive, less energy hungry, and not reliant on the six-month chip rollover cycle. Which one will win? It is not yet decided. The medium term will tell us.
You mentioned 5,500 US data centres struggling with energy. With roughly a fifth of global oil supply transiting through the Strait of Hormuz, what is your base case if the US-Iran ceasefire talks collapse again?
The impact is mainly on pricing today. Energy prices and refined products are very high. Even during the recent de-escalation when oil prices dropped significantly, refined products were still priced at near $100 a barrel equivalent. Here in Paris, many cars are struggling to find diesel — you have to go 20 to 30km outside the city. For data centres, which are extremely price sensitive to electricity, higher electricity prices directly raise the cost per token. Instead of $5,000 per month in my case, it becomes $6,000 or $7,000. There is also the helium issue — helium is essential to manufacturing chips, and with refined products and gas blocked through the Strait of Hormuz from Qatar and other GCC countries, access to helium is much more difficult and expensive. The pricing of hardware is going up directly because of the Strait of Hormuz.
If the Strait reopens immediately — has the impact already happened?
The energy shock has already happened and it will feed into inflation over the medium term. If we compare this to the second largest energy shock in history — the 1970s — that impacted 6 to 7% of international oil exports. Today it is impacting 21%. Even if the Strait of Hormuz fully reopens today, logistics experts estimate it will take roughly three months for traffic to normalise. Then many refineries in the GCC that have shut down will take another three months to fully operate. And refineries in China and Russia — many of which were damaged by the Ukraine war — are also not at full capacity. The problem is not just oil exports but refined products like gasoline, diesel, and helium — all of which will take months to come back. This is why Treasury rates are above 5% on longer maturities. Bond vigilantes are telling the Fed and other central banks: you are mistaken to freeze rates. We are heading toward a higher inflationary environment and a rising interest rate environment.
Doctor Ryan, thank you so much for joining us today.
Thank you, Rachel.
