2025 JULY ISSUE
RACING OF THE AI
Written by ANDREW SIA
Share this article !
From the Desk of the Publisher
Everyone is talking about the Artificial Intelligence and the surge in investment may cause a “bubble” to burst. Especially when we look at the key players today, they are all the world’s richest persons in the world, and they are all in the tech-industry. We can count Elon Musk, Jeff Bezos, Larry Ellison, Bill Gates, and when they are telling us not to worry, I hope that we can be a bit skeptical and take precaution.
Looking back in recent years, we have seen the bubble bursts in many areas, such as:
Dot-com Bubble (1995-2000) – Internet companies were operating with profits attracted massive investments. Burst in 2000 when expectation met reality, as the result NASDAQ lost almost 80% of its value.
US Housing and Banking Bubble (2000-2008) – It was led by low interest’s rates and subprime lending inflated housing prices. Collapsed when borrowers defaulted leading to the 2008 global financial crisis. It caused bankruptcy of Lehman Brothers.
Commodities and Oil Bubbles (2008) – A barrel of oil hit $147 fueled by speculation in the middle of the year. Price collapsed to under $40 by year end when the recession hit.
Cryptocurrency/Bitcoin Bubble (2017-2018, 2021-2022) – Bitcoin soared to $20,000, fell to under $3,000 by 2018. It repeated to $69,000 and collapsed in 2022. It was all because of speculative trading.
Biotech and SPAC Mini Bubble – It is worth mentioning that during Covid pandemic, low interest rates sparked booms in biotech, EVs, and SPACs (special-purpose acquisition companies.) It collapsed when interest rates rose, and investor optimism cooled.
I hope that we can be smarter this time and learn from lessons. Remember, the winners will only be a handful of those tech companies, and the rest will be crushed when the bubble burst.
PS: The last I read that Nvidia chief Jensen Huang said that AI boom was “dramatically different” to the dotcom bubble as the hyperscale like Microsoft, Google, and Meta are so much richer than those bubble listings at the time. And then Mary Daly, head of the San Francisco Fed, said that an AI bubble will not be a threat to financial stability. Jeff Bezos, Amazon founder, even hailed the AI boom as a “good kind of bubble.” I suggest that we should start to find cover now.
All big-tech groups are racing to expand their huge new data centers. It began its debut with OpenAI when it released its ChatGPT in November 2022. Since then, it has been widely used and as mentioned the Big-Tech groups, Microsoft, Alphabet, Amazon and Meta plan to increase their capital expenditure to more than $300 billion this year. It has been reckoned that a total of $475 billion will be spent on data centers this year. It’s an increase of 42% on 2024.
It has been predicted that by 2050 the total of $5.2 trillion would be required. An insatiable appetite from billions of users is using sophisticated AI models which are more complex, costly, and energy-guzzling than the power grids, telecoms networks or even cloud computing system that came before them.
The dramatic increase in power demand which consume all available grid power is due to the unexpected demand which gabble all available grid power. The new AI chips has transformed data center design and the racks of computers running Nvidia’s chips consume at least 10 times as much power as regular web servers. Its processors are giving so much heat that air conditioning is not enough to cool them.
The data centers need to keep within their safe operating temperature range between 30° to 40° C. And to avoid malfunctions, more advanced cooling method is required. Cold water pipes have been installed to transfer the heat away from the equipment. The water is then circulated to the cooling towers. Evaporation is used to reduce the heat to a safe range. Then this leads to water loss, with one tower churning through 19,000 liters per minute. Then the introduction of seawater for the same purpose is being introduced.
One of the first clusters for the data center in the U.S. was developed in northern Virginia, known as the “Data Center Alley” with cheap land, tax incentives, and access to subsea cables linking North America and Europe which help to transmit internet traffic. New clusters are coming up from Atlanta, Columbus, Dallas and Phoenix, with lower energy costs and the laying of optical cables.
But close to 95% of commercially available AI computing power is operated by the U.S. and China tech groups. China’s drive to expand its AI computing infrastructure is spurred by its local governments, including the remote regions in Xinjiang and Inner Mongolia to build data centers. Through a negotiation between Washington and Beijing, Nvidia is resuming its shipments of AI chips to China. But the U.S. is still keeping the most powerful AI semiconductors from China.
Gulf states are also preparing to invest billions into AI infrastructure. The United Arab Emirates announced a huge data center cluster for OpenAI with ambition of up to 5 giga-watt of power. Saudi Arabia is also showing its intention to build AI factories.
For the time being, we are going to share our findings in the U.S. and allow us to walk you through our study:
Somewhere in the middle of nowhere, a town called Ellendale in North Dakota with population of 1,100 is building an AI factory with a price-tag of $15 billion. It is equivalent to a quarter of the state’s annual economic output. It is said that North Dakota’s windy, wintry climate helps to keep down cooling costs for the data center.
Over the past three years, leading tech firms have committed more toward AI data centers like the one in Ellendale. This is like the days when Industrial Revolution was first introduced. This is a building spree where no one knows what it would lead us to and when will the investment returns.
It is a mega-speculation that the technology will rapidly improve, transform our economy, and start to bring steady profits. Tech firms are pouring multiple billions and betting on something without knowing the outcome and how it can benefit the civilization.
Silicon Valley watchers are worried about the enthusiasm for AI which can turn into bubble, just like the “internet bubble” that went burst dramatically in the early 2000s. It was also known as the “dot-com bubble” and at that point Nasdaq Composite Index rose by about 600%. When it collapsed a mass sell-off of technology stocks occurred, causing the Nasdaq plummeted by nearly 77% in October 2002. The crash caused trillions of dollars in investment losses.
OpenAI’s Sam Altman called his data center “Stargate,” in reference to a 1994 movie talking about interstellar time-travel portal. The company executives laid out plans that would require at least $1 trillion in data-center investment. Altman lately committed the company to pay Oracle an average of around $60 billion a year for server in data centers in the coming years. Yet OpenAI is on track to take in just $13 billion in revenue from all its paying customers this year.
CoreWeave, a company started six years ago, was an obscure cryptocurrency miner with fewer than two dozen employees. It is now an AI cloud-computing company based in Livingston, New Jersey, specializes in providing cloud-bases graphics processing unit (GPU) infrastructure to artificial intelligence developers and enterprises, and also develops its own chip management software. Microsoft is its major customer, which can also represent 30% of CoreWeave’s 2026 revenue estimated at $12.1 billion. CoreWeave has been metamorphosed into a computer goliath and a market value bigger than General Motors or Target.
We have to know that today’s numbers are far larger than in those “dot-com bubble” days. We need to be very careful when we are making our investments worthwhile.
Venture-capital form Sequoia estimated that money invested in AI infrastructure in 2023 and 2024 alone requires consumers and companies to buy roughly $800 billion in AI products over the life of chips and data centers to produce a good investment return. It is believed that the useful life for most AI processors of three and five years.
Consultants at Bain & Co. estimated the wave of AI infrastructure spending will require $2 trillion in annual AI revenues by 2030. This is the combined 2024 revenue of Amazon, Apple, Alphabet, Microsoft, Meta, and Nvidia, and more than five times the size of the entire global subscription software market.
Morgan Stanley estimates that last year there was around $45 billion of revenue for AI products. This sector made money from a combination of subscription fees for chatbots such as ChatGPT and money paid to use these companies’ data centers. It is found that consumers have been quick to use AI, but most are using free versions.
OpenAI counts roughly 700 million people which represent 9% of the world’s population, as weekly users of ChatGPT as of August, up from 500 million in March. Its revenue is on track to triple over 2024. It is already said that AI can replace a large number of white-collar jobs. The saving will be enough to pay back the investment.
CoreWeave has racked up over $42 billion worth of contracts with tech companies renting its servers in the coming year. The growth also enabled the company to raise lots of debts. It has around $15 billions of debts with interest rates start above 8% on financing its deal with top notch companies including Microsoft.
CoreWeave owes more to the landlords and from its securities filing show that it has $56 billion in payments for leases on data centers and those leases usually run around 10 years. But its deal with the tech companies typically run for two to five years, and it has to make billions of dollars in payments after the tech-companies deals expire. Currently it is carrying rents of data centers where it hasn’t been able to find customers to rent the servers.
If the wave of building centers proves far more than demand, the risk of CoreWeave’s data centers could end up like those dormant fiber optical cables that ran through the U.S. in the 2000s.
As CoreWeave grows, it is going after energy-rich areas ahead of its rivals. It is spanning from New Jersey to Texas to Oregon.
We have to learn from the previous bubble-bursts and the way of the AI boom is like the days of the gold rush, which happened recently in the beginning of 2000. Towns that have data centers set up are going to plan housing for serve the workers. We have to keep our fingers cross this time and furthermore we don’t know what AI would lead us for the work that are available for our younger generations.
That leaves us in deep thoughts for other continents, and Europe is not leaving behind and they have declared the plans for five AI giga-factories to develop and train complex AI models.
China’s DeepSeek has made its hyper-efficient AI model which is running on its own data to keep it away from the U.S. for political reason. It was only founded in 2023 and backed by the hedge found High-Flyer. It was using the Nivida H800 chips, and not the high-end ones. It spent approximately $5-6 million compared to GPT-4, which was over $100 million and required only a fraction of the computation power.
DeepSeek achieved the success with advanced technical innovation, resourceful hardware strategy, research-first, commercial-later strategy, a young and skilled talent pool, and an open-source mindset shattered the myth that the cutting-edge AI requirement. Through architectural creativity, optimization, and strategic resources, this small and lean startup has disrupted the AI industry and redefined how powerful models can be developed.
Written before the press:
While I was collecting writeups and information about artificial intelligence, I came across the piece from Thomas Friedman in his column in the New York Times. He was using the title of, “Why the U.S. and China Should Collaborate, Not Compete, on A.I.”
He mentioned that it is both a question of geopolitical and geoeconomic question. I can agree to that. But for the two greatest powers on earth to collaborate on a shared level of trust that guarantees it can always remain aligned with human flourishing and planetary stability is a bit naïve and unrealistic. He is saying that if this is not going to happen, the world will become fractured where every nation will build their own wall-offed AI and guarded by incompatible standards and mutual suspicion.
We have to know that YouTube can’t be used in China. WhatsApp, Facebook, ChatGPT, just to name a few, can’t be used in China. We are a bit skeptical when we have to use WeChat and TikTok for instance. There is the lack of trust.
We have to be realistic when we are exploring this question. First of all, the values and systems are totally different. Our U.S. system is based on democracy, rule of law, and openness; China’s governess is centralized, authoritarian, and party controlled. This leads to fundamentally different approaches to rights, freedom, and responsibilities.
Our difference in geopolitics is the U.S. committed to alliances and freedom of navigation in international waters. China continues its effort to claim its sovereignty in the South China Sea and over Taiwan.
Our difference in geoeconomics is the U.S. pushing for transparent, open market, and market-driven rules. Whilst China is still pursuing state-led industrial policy and subsidies while pushing their “Made in China 2025” and “Belt and Road Initiatives.” They use it as its national strategy.
The differences in values and strategic ambitions are so fundamental that it is not easy to bring the two together. Any collaboration will likely to be seen as—limited, fragile, and transactional.
I don’t mind seeing that the two greatest powers go their separate way.
Last and not the least, I wrote a prose for this with a very heavy heart. I am asking you to take precaution with your investment and don’t fall into the trap when the bubble burst in 2000.
Human – A Bunch of Living Creatures
It is actually not such a big deal
When we look at the remnants on our planet
Many of the things still can’t be comprehended
Such as the structures of the pyramids which can be found in different continents
And the remains on the mountains of Machu Picchu
Not to forget to mention those giant statues on Easter Island
I found the artificial intelligence both fascinating and scary
We have found it too easy to use and too powerful in giving any solutions
I am afraid that it will take away our jobs and even our power in making research
Soon college education isn’t that critical and all we have to do is to ask the chatbot
It is taking away our earning power and make us lead a life without any thinking
Soon we will become so lay back that we will return our life to the Garden of Eden
Artificial intelligence will drive robots to do all kinds of tasks for us
It will take away our job and build houses for us to live
When we look at pyramids and city walls seamlessly using stones we don’t even know how to lift them
Let alone to have any skill in solving any problems
We all become a bunch of useless creatures roaming the globe without purpose
The leaders and politicians are running the earth and enslave people without heart except for their own pleasure
Then someone from above will descend and look at us with the question how we could have ended in such an incredible situation
Because of our human’s greed, we have ended up running our civilization to the ground again
To burn us to ground or to drown us like the fable told in Noah’s day
This experimental field has proved once again a big failure
Everything is already so hopeless and helpless
Even the creator will have to give us up ultimately
Written on September 4, 2025