What drives markets – and puts them at risk
This year, the AI industry experienced a wave of investment: global spending on AI is estimated at around USD 1.5 trillion, representing a 35% increase over 2024. OpenAI, which is not yet publicly traded, is valued at approximately $500 billion. Nvidia has a market capitalisation of around USD 4.5 trillion. Leading industry representatives such as Sam Altman (OpenAI), Jeff Bezos (Amazon) and David Solomon (Goldman Sachs) are talking about a bubble. If this bursts, it is likely to lead to a market shake-up that would primarily affect unprofitable, smaller or highly indebted providers.
Valuation and market behaviour: between growth and the risk of a bubble
However, compared to the dot-com bubble in 2000, there is still ‘room for manoeuvre’: the “Mag7” trade at an average P/E ratio of around 26.8 (the ‘dot-com Top7’ reached a value of around 52). With an expected P/E ratio of around 28 for 2026, Nvidia is well below the valuation level of a leading network equipment supplier at the time of the dot-com boom (P/E ratio: around 100). The difference today is that the Magnificent Seven are financing their record investments from strong cash flows and remain profitable.
Cycles and systemic risks: when capital flows reinforce themselves
Large technology and AI companies have built up tightly knit capital and supply networks. One example is OpenAI, which has received tens of billions of dollars in total funding from Microsoft, SoftBank and investors from Saudi Arabia, amounting to around USD 65 to 85 billion. OpenAI purchases infrastructure from Oracle (keyword ‘Stargate’, around USD 300 billion) and hardware from Nvidia. Nvidia, in turn, is investing up to USD 100 billion in OpenAI – partly in connection with the development of its own systems – and is financing new cloud providers that use graphics processors as collateral for loans. At the same time, OpenAI is investing up to $300 billion in a US chip manufacturer and is expected to gain access to up to ten per cent of the company.
The result is circular dependencies: revenues and valuations are increasingly interdependent, which experts refer to as an ‘indefinite money glitch’. A historical parallel can be seen in the dot-com bubble, when large telecommunications equipment suppliers attracted attention with similar financing and investment structures and later failed.
Most recently, OpenAI awarded a manufacturing contract worth around USD 500 billion to one of the leading semiconductor manufacturers. In total, the orders placed thus amount to around USD 1.5 trillion. Data centre capacity is expected to increase from around 2 gigawatts in 2025 to around 28 gigawatts in 2030. According to studies, hyperscalers could control around 60 per cent of global data centre capacity by 2030. This would correspond to a volume of around 220 gigawatts (today it is around 80 gigawatts), without taking into account the additional capacity planned by OpenAI. Sam Altman even outlines a volume of around 250 gigawatts for the year 2033.
Revenue momentum and growth prospects: Focus on OpenAI
OpenAI's revenue is estimated at around USD 13 billion for 2025, which is roughly three times the previous year's figure. Around 70 per cent of this will come from subscriptions (around five per cent of the approximately 800 million weekly users are paying subscribers), with a further 20 to 25 per cent coming from corporate customers via the application programming interface (API). In order to become profitable by around 2030, revenue would have to increase roughly tenfold to between USD 125 and 200 billion.
In the first half of 2025, technology investment in the US exceeded one per cent of gross domestic product – a higher level than at the time of the dot-com bubble. At the same time, there is a significant gap between investment in AI infrastructure and actual end-user revenues: around USD 1.2 trillion globally compared to current AI revenues of around USD 100 billion.
According to studies by Bain and McKinsey, annual AI revenues would have to rise to more than two trillion US dollars by 2030 for current investments to pay off economically. Even then, a gap of around 800 billion US dollars would remain. Two trillion US dollars is roughly equivalent to the gross domestic product of Italy or Canada. The global software sector currently generates around USD 700 billion.
The CEO of a leading semiconductor manufacturer compares the current expansion of AI infrastructure with historic basic innovations such as the railway and the internet, which form a critical foundation for the future global economy.
Energy as a potential bottleneck
According to estimates, AI data centres will consume between 20 and 50 terawatt hours of electricity this year. This corresponds to around five to twelve per cent of the total energy consumption of data centres in 2024, which was around 415 terawatt hours, and around 0.5 per cent of global electricity demand. A data centre with a capacity of one gigawatt requires around one terawatt hour of energy for 1,000 operating hours. AI is expected to double electricity consumption by 2026.
In regions such as Texas and Tennessee, where new large-scale facilities are being built, the first bottlenecks are already being reported. By 2030, data centres could account for two to four per cent of global electricity consumption, mainly due to the increasing use of AI. This would represent a 20- to 40-fold increase compared to 2025.
According to calculations by the International Energy Agency (IEA), the global electricity demand of data centres is expected to rise to around 945 terawatt hours by 2030, which is roughly equivalent to Japan's current consumption.
The growing deficit carries the risk of price spikes and could limit computing power, creating a potential growth factor for the industry. A leading US technology company is currently building a new data centre in Louisiana for around USD 29 billion that is as large as Manhattan. Analyses show that electricity prices are rising significantly in some cases in the vicinity of large data centres – by up to 38 per cent in some regions of the US in the first half of the year.
In the United States, there is already discussion about prioritising AI data centres in the event of energy shortages – with corresponding risks for households and industry. It remains questionable whether grid expansion can keep pace with the rapid development of AI. Rising energy prices are considered likely. Countries with advanced electrification infrastructure could gain strategic advantages in the further development of AI technology as a result.
Disruption in the software sector: automation as an accelerator and risk
Around 71 per cent of companies already use artificial intelligence in software development, with almost half using untested open-source models. This significantly increases security and compliance risks such as malicious code, data theft and licence violations. Traditional software providers are losing market share, while AI-powered tools are setting new standards in development and application.
A study by the Massachusetts Institute of Technology shows that 95 per cent of the companies surveyed have not yet seen a positive return on investment from their AI activities. At the same time, OpenAI is seen as a potential threat by many software companies, as AI agents are increasingly automating core workflows.
In online retail, OpenAI is currently introducing ‘one-click shopping’ and integrating this function directly into its own platform. Collaborations with several large retail and cloud providers illustrate the speed with which market entry is taking place.
At a major technology conference in San Francisco, there was recently a consensus that the software sector is facing a profound disruption that poses significant challenges for established market players. At the same time, record sums are flowing into young AI software companies – a wave of automated solutions is growing rapidly.
Outlook: Between euphoria and caution
This year, relaxed financing conditions – lower interest rates, high credit availability and low risk premiums – created an environment with abundant capital, favourable financing and a high appetite for risk. This has stabilised AI stocks, although there are repeated warnings of a bubble forming. Central banks and large investment houses do not expect a significant reversal of this trend in the short term, so valuations could continue to rise for the time being. The much-discussed correction in September 2025 did not materialise, but the debate about a possible bubble continues. An abrupt end to the upswing is not currently foreseeable.
However, the massive investments in AI infrastructure are showing signs of overheating. Leading technology companies themselves admit that overinvestment is possible, but the risk of falling behind in key developments outweighs this. OpenAI's large-volume orders also show signs of a period of overinvestment. Financing commitments are unclear in some cases and estimates vary widely. Corrections therefore remain possible and are likely, as has already happened several times in the past two years. Regardless of this, AI technology is likely to profoundly change the economy and everyday life, comparable to the dot-com phase that marked the breakthrough of the internet. At present, however, this development is still in its infancy.
So far, public perception has been shaped primarily by software innovations, such as AI chatbots and search models. In the next stage of development, physical AI applications such as autonomous vehicles will reach market maturity and be scaled up. Their areas of application are increasingly expanding from US cities to European metropolises. Chinese manufacturer Momenta is planning to produce level 4 automated vehicles in collaboration with German automotive companies.
Humanoid robots are then likely to trigger the next wave of innovation. At the same time, advances are expected in DNA analysis, drug development, simulation technology and scientific applications. In addition, the topic of self-evolving AI is gaining in importance: Google recently published studies on language models that improve their capabilities independently. This is an area in which major technology providers are currently making significant investments.
Long-term efficiency potential through AI
The use of AI opens up significant efficiency gains in numerous industries, from design and software processes to industrial automation. In the long term, this could result in structural changes that have a positive impact on productivity, profit margins and cash flows. At the macroeconomic level, technological progress driven by AI is likely to have deflationary effects.
Legal information
Marketing ad - All information published here is for your information only and does not constitute investment advice or any other recommendation. The statements contained in this video reflect the current assessment of DJE Kapital AG. These may change at any time without prior notice. All statements made have been made with care in accordance with the state of knowledge at the time of preparation. However, no guarantee and no liability can be assumed for the accuracy and completeness. The video and its content are protected by copyright. Long-term experience and awards do not guarantee investment success. Securities are subject to market-related price fluctuations that may not be offset by the active management of the asset manager. Please note our risk warnings.