Where is the Best Place for Data Centers supporting AI? What are the Implications of DeepSeek R1 on AI-focused Data Centers?
Our assessment on these questions, and our investment ideas in this space
“In today’s hyper-connected world, data centers have become the nerve centers of our digital economy. Hyperscale tech giants like Microsoft and global investment powerhouses like BlackRock are pooling their vast resources and pouring billions into data center construction, betting that these facilities will underpin the next wave of technological innovation. With artificial intelligence evolving at a dizzying pace, the stakes have never been higher.” - Forbes, source link
We are at the dawn of a new technological space race - in Artificial Intelligence (AI). While a thorough review and replication is currently in progress, the introduction of China’s announced new open-source based AI software DeepSeek R1 may very well be the “Sputnik” moment for this new space race of AI.
This new freely-available software from China was developed supposedly for less than USD 6 Million in just a few months by China, with a relatively much smaller number of Graphics Processing Units (GPUs) than the U.S.-based AI industry even thought possible. These GPUs are semiconductor chips focused on supporting AI applications, manufactured by Nvidia, and restricted for access by and sales to China by the U.S. - as such, China embarked on overcoming these sanctions and restrictions by innovating through software a way to minimize all the hardware and infrastructure generally thought necessary to support AI applications, at extremely low costs (estimated to be a factor of 30x less expensive).
Here is a a video explaining the technology below and another video by one of our Most Admired Advisors (MAAs) Sean Foo elaborating on the economic implications:
Even tech legend Marc Andreessen, whose Netscape Navigator enabled the internet revolution, is reaching back decades to find what he believes is the most appropriate historical parallel—the day in 1957 when the Soviet Union beat the U.S. into space with the launch of the first satellite into orbit. “DeepSeek-R1 is AI’s Sputnik moment,” - source link
Link here for the take by one of our Most Admired Advisors (MAAs) Louis-Vincent Gave, exploring these issues:
Was only $6 Million used to develop DeepSeek R1 or did it take much more extensive funding and in particular a larger number of GPUs?
Is there a built-in bias in DeepSeek R1 on alignment with China’s CCP views?
Is this a positive confidence boost to China’s entrepreneurs and innovators?
What are the current effects and what are the implications to China’s equities market and other asset classes?
“If you take a step back what made the US market exceptional, what set the US apart relative to say China or Japan or France or any other major markets was the presence of these moat building mega cap tech stocks. Take that away. And all of a sudden maybe the US doesn't look quite as exceptional as it used to.” - Louis-Vincent Gave - source link
As we await for a thorough review of what has been innovated, and assuming that the new software can be replicated and confirm the results presented, we view this development as helping to progress and scale AI applications widely and cheaply. Robust scalable data centers focused on AI will still be required to support the massive growth of AI applications in all industry sectors.
Where is Best to Setup and Operate AI-focused Data Centers?
Where in the world do we have these necessary requirements and business friendly environment that can foster the setup and operations of the data centers required for supporting AI applications?
A place with lots of energy resources, including not only renewable energies but also oil and gas and even nuclear energy, able to meet the power requirements of AI-focused data centers?
A place with lots of water or with the potential for infrastructure able to support adiabatic or air-cooled systems to provide the cooling systems required on AI-focused data centers?
A place with a pro-business environment, characterized by business-friendly and innovation-promoting regulations necessary to build and operate AI-focused data centers?
A place with lots of land and open spaces conducive to setting up AI-focused data centers?
A place with lots of low-to-reasonably costing labor, or the potential to recruit, educate and train the required labor, in a business-friendly environment?
While places like the United States and Canada may have lots of land and energy resources, they are many challenges with regards to reasonably-costing labor and a pro-business environment, especially given complex, extensive and confusing regulations which in some cases promote regulatory capture:
“Access to clean energy is limited due to government-induced barriers, including inefficient permitting, weaponized litigation, high interest rates, and insecure supply chains. These are failures of the government, not the market.” - source link
For those that are hopeful that a new U.S. Administration, with a new advisory committee called the Department of Government Efficiency (DOGE) will reverse the trend for the U.S. on unsustainable U.S. government spending, budgets and regulations:
“The upshot is that any recommendation from the DOGE Commission, if it ever sees daylight, will land on the desk of a bureaucrat whose budget it targets. Their first response will be a memo—ten times longer than the original proposal—arguing why it’s absurd, harmful, and impractical. Then it becomes mired in legal challenges, with cases bouncing between courts for years. The legal back-and-forth is endless, and while the White House may occasionally win, each legal battle is an isolated case and requires enormous effort.” - source link
So, is there a place meeting the above requirements for AI-focused data centers? Where do we think this place is?
Argentina
“Why are the titans of Silicon Valley cozying up to a radical anarcho-capitalist notorious for wielding a bureaucracy-slashing chainsaw? Mostly, it’s his promise of a low-regulation artificial intelligence (AI) hub.” - source link
Argentina holds the potential for being the global hub of AI-focused data centers:
“Argentina with its newfound potential to enable quick data center buildouts concurrently with cheap and reliable electricity from diverse generation resources. A business-friendly environment will enable it to finally capitalize on its high solar and wind potential, and abundance of natural gas from the Vaca Muerta shale patch, while also attracting emerging technologies like small modular reactors. Argentina is also blessed with world-class copper and lithium. Data centers are particularly copper-intensive and increasingly reliant on lithium-based batteries to replace diesel generators.” - source link
We think Argentina has greater potential to be the global hub for AI-focused data centers, even bigger and better than nearby Chile which has recently announced plans to be the digital infrastructure hub of South America.
Argentina has announced a bold initiative to position itself as a global leader in artificial intelligence (AI) and sustainable energy through its strategic nuclear energy plan, led by small modular nuclear reactors (SMRs) to support the energy demands of AI systems, which continue to grow as models like OpenAI's GPT-4o require massive computational resources, but also to grow as models like DeepSeek R1 (and other similar software and models) scale massively to support growth in AI applications globally and across all industries.
While AI systems based on OpenAI consume significant amounts of power during training and operation, AI systems based on DeepSeek R1 will also consume significant amounts of power as AI applications become increasingly widespread and scaled. Complex queries will cost thousands of dollars in energy. Argentina's focus on energy development, including natural gas and nuclear power, aims to address this challenge by providing scalable and sustainable energy applications and associated solutions.
Argentina’s natural gas resources bring great potential to the AI-focused data centers vision:
On the cooling of AI-focused data centers, Argentina has it all - lots of water, lots of space, and even cold climate areas like in Patagonia which can also help to meet the cooling requirements and to reduce cooling costs as well.
And Argentina can leverage and expand from its already existing experienced labor in the data centers industry, given the many existing data centers throughout the country currently - source link. Map of where the current data centers are in Argentina:
So the idea is to have all the AI-based applications do their work at the data centers in Argentina - not for Argentina to necessarily send energy and power abroad - and for the subsea Internet cables to provide the communications with AI-developing hubs globally:
A Shift to a More Regulatory Efficient Environment
With the new Presidential administration in place led by President Javier Milei, Argentina has made great strides on reviewing and optimizing the regulatory environment:
“From December 10, 2023, when Milei assumed the presidency, to December 7, 2024, there were 672 regulatory reforms. On average, that means that during his presidency, Milei has been issuing 1.84 deregulations per day, counting weekends. Out of the total amount of reforms, 331 eliminated regulations and 341 modified existing regulations .. Their deregulations are increasing economic freedom, reducing opportunities for corruption, creating greater transparency in government, helping to formalize the informal sector, stimulating growth, and setting an example for countries around the world to follow.” - source link
Vision
Our vision is for Argentina to become a world-class leader in AI-focused Data Centers - powered by the natural gas being developed now and the small modular nuclear reactor strategic program - and most likely placed in the areas near the natural gas development and in particular near the coastal areas of Argentina and near the cooler Patagonia areas of Argentina.
Investing in the AI-focused Data Center vision for Argentina
How are we investing now or positioning to invest in this space? Let’s delve in …