Will AI and GPUs Save the Japanese Economy? The 'Metal Fantasy' That Turns a Blind Eye to the Reality of Electricity
Rintaro Takechi (Editor-in-Chief, Weekly Bubble Watch)
AI Market Revival Due to Exploding GPU Sales? The 'Metal Faith Economy' Where Expectations Run Ahead of Reality
'GPUs are selling. NVIDIA's market cap has surpassed Apple's. AI semiconductors are the new reserve currency.'
—I can almost hear the headlines of the Nihon Keizai Shimbun.
And, as if by default, the following sentence would be:
'We must not miss this trend. SBG's stock price is aiming for 12,000 yen.'
I see. Wonderful. GPUs sell, AI learns, and stock prices rise—it's as if we live in a country where electricity and resources are not finite.
Reality: A GPU without electricity is just a 'heavy ornament'
But reality is always full of irony. A GPU is merely a 'material' that only functions as an 'AI brain' after pouring in vast amounts of electricity and wasting energy.
Manufacturing requires high-purity silicon, rare metals, ceramics, copper, gold, and more, with each GPU requiring hundreds of kWh of electricity to produce.
And when in use, the entire data center supporting the AI consumes as much electricity daily as a large power plant. The figure of '5GW per site' mentioned in concepts like the 'Stargate Project' is far too naive.
This is a scale comparable to the total output of all four Barakah nuclear reactors built by the UAE over a decade (approx. 5.6GW), and it even exceeds the total output of all six reactors at the Fukushima Daiichi Nuclear Power Plant (approx. 4.7GW).
In short, without electricity, a GPU is nothing more than an expensive 'piece of pottery.'
The Sin of Nikkei-style Optimism: Predictive Models Without 'Electricity Costs'
When discussing the future of AI and GPUs, why do reports from Nikkei-affiliated analysts completely omit fundamental constraints like power supply capacity and electricity costs?
The following 'tacit assumptions' are spoken of with composure:
・A world where GPUs run infinitely with zero electricity costs?
・Electricity just springs up naturally without building nuclear or thermal power plants?
・Is it okay for one robot to consume 300kW? And in units of 10 billion?
This type of optimism treats infrastructure, resources, and time as 'negligible variables,' making it less like science and more like a 'cult faith.'
In fact, Peter Diamandis, a standard-bearer for the singularity movement, says this:
However, if just one of them consumes 300kW, the world would face a 'power collapse' every single day. A robot economy without energy constraints is nothing more than a product of escapism.
In the 'Japan revival stories' that the Nikkei loves so much, the 'gritty realities' like hydroelectric power plants, transmission losses, and fuel costs for power generation never seem to appear.
The 12,000 yen stock price air valve: Predictions without analysis, hype without responsibility
Is the basis for SBG's stock price rising to 12,000 yen its NAV? Arm's potential? AI hegemony?
These are all part of a fantasy economy that assumes electricity is always available.
And those who spread this fantasy with a megaphone, single-mindedly, are the 'sales analysts,' 'follow-up rating agencies,' and Nikkei-style economic optimists.
True future forecasting means using a 'power meter' rather than a 'calculator'
What is needed in the AI era is 'power generation capacity' rather than 'computing power'.
What is needed for forecasting is a power supply and demand balance sheet rather than a stock price chart.
If the power runs out, GPUs, robots, AI, and SBG will all become nothing more than scrap.
Still, readers who want to believe in 12,000 yen should read the Nikkei.
Readers who want to know the reality should read the Weekly Bubble Watch.
'It won't go bankrupt even if the stock price falls' does not apply to SBG
During the SoftBank era, it survived a 1/100th crash after the IT bubble burst. But the current SBG is on the verge of collapsing at 'any moment'.
In 2000, SoftBank Corp. (now SBG) recorded a stock price of 200,000 yen at the peak of the IT bubble, and then plummeted to the 2,000 yen range in the early 2000s. It was a sharp drop of more than 99%.
However, SoftBank did not collapse. Instead, it shifted to bold offensives such as the acquisitions of Yahoo BB and Vodafone Japan, and achieved a revival.
But as of 2025, I am warning that if SoftBank Group's (SBG) stock price falls by just a few dozen percent, the 'risk of bankruptcy' will become a reality.
This point is not widely reported in the Japanese media, but it is already considered common knowledge in high-quality overseas financial newspapers (quality papers).
If the stock price 'falls by just 30%,' SBG will enter the red zone
The current SBG stock price is about 7,700 yen. If this falls by 30% to around 5,400 yen, the possibility of a margin call (a demand for additional collateral) being triggered for the massive collateralized loans held by SBG will become extremely high.
In fact, during the 2020 COVID shock and the 2022 tech stock crash, SBG's credit risk began to be acutely recognized in the market. In 2020 in particular, the market value of Alibaba shares and SVF assets held by SBG plummeted, leading to a situation where Masayoshi Son himself had to pledge a large amount of his own SBG shares as additional collateral.
The collateral value is strongly linked to SBG's stock price, and the structure is such that an increase in valuation losses leads directly to a funding crisis.
SBG is not like a normal company
For a general company, even if the stock price falls from 1,000 yen to 500 yen, it does not immediately affect funding. This is because the stock price is merely a market valuation, and bank loans are screened based on the company's operating cash flow and fixed assets.
But in the case of SBG, the situation is the exact opposite.
Most of its borrowings are based on leveraged finance collateralized by market-value assets such as SBG shares.
Therefore, if the stock price falls by 30-40%, the LTV (Loan to Value ratio) will drop sharply, and forced repayment or the sale of collateralized shares (loss-cutting) will become a reality.
In short, for SBG, the stock price is not the 'face of the company' but its 'lifeline'.
Why did the SoftBank era hold up?
Between 2000 and 2002, SoftBank Corp.'s stock price plummeted from 200,000 yen to the 2,000 yen range. However, the SoftBank of that time had a structure centered on operating cash flow generated from its core business, and it did not rely excessively on its own shares or the shares of its investments as collateral for loans.
On the other hand, the current SBG is completely different.
It has adopted a fund-type model that invests in unlisted startups by leveraging its equity and massive borrowings, and much of its asset valuation is composed of unrealized gains (= paper profits).
If the stock price falls, these 'unrealized gains' will vanish in an instant, and the very foundation of its valuation will collapse. In other words, the SoftBank era was supported by 'substance,' but SBG is currently barely supported by 'market value'.
If the stock price of 7,700 yen falls below 5,000 yen, it enters the 'red zone'.
According to analyses by multiple analysts and credit research agencies, the limit for SBG to maintain its current LTV is considered to be around a stock price of 5,000 yen. In other words, if the stock price falls by another 30-40%, SBG will approach 'bankruptcy conditions'.
Some readers might think, 'You should only worry if the stock price drops to 1/10th,' but SBG has already become a 'company that will collapse if it drops by 30%'.
The stock price is its collateral, the backing for its credit, and its lifeline in the corporate bond market.
If the valuation of AI-related assets (such as Arm) collapses, the possibility that the entire SBG will collapse in a chain reaction is extremely high.
Conclusion: SBG is a 'Pyramid of Market Value Economy'
SBG is a 'pyramid of market value economy' that can attract massive funds as long as its stock price is high. However, it harbors the risk that the entire structure will collapse if the 'stock price' that supports its foundation falls by 30%.
Unlike SoftBank, which endured even when its 'stock price became 1/100th' during the collapse of the IT bubble, the current SBG is a 'company that might end before its stock price becomes 1/2'.
That is the fate of the company known as SBG, and it is the very danger of this 'structurally dependent bubble economy'.
Masayoshi Son's Stock-Collateralized Loans and Margin Call Risk: Continuing to the May 2025 Latest Analysis
Rintaro Takechi
