Stargate Aims for $500 Billion Investment in U.S. AI-Focused Data Centers

The Stargate AI data center initiative, led by Oracle, OpenAI, and SoftBank, arguably aims to create up to $500 billion worth of new high-performance-computing infrastructure in the United States. 

While it does involve scaling up infrastructure to meet the growing demand for AI models and computational power, the initiative seems also to reflect a focus “up the stack,” even if the core effort can be described as creating high-performance-computing infrastructure at scale. 

Oracle’s participation suggests there will be an effort to optimize capabilities for enterprise applications while OpenAI’s presence obviously means a focus on apps requiring AI and capable of supporting more capabilities moving towards some form of artificial general intelligence. 

But even before that happens, Stargate aims to support compute-intensive AI apps across some key verticals that would seem to include healthcare, finance, and autonomous systems.

Remarks by President Donald Trump at the press conference announcing the venture also indicate some revision of permitting rules to allow local generation of electricity for the data centers as well. 

Though it is always possible the estimates will prove inflated, most observers expect a huge increase in HPC-capable data centers in the United States to 2030. On the other hand, existing data centers that add HPC capabilities are also likely. So “new” might better refer to “new capacity” rather than “new data centers” when looking at aggregate growth. 

Year	New U.S. HPC Data Centers
2025	50
2026	60
2027	70
2028	80
2029	90
2030	100

All that noted, a significant portion of HPC global demand will happen in the U.S. market. More than half of hyperscale data centers globally operate in the U.S. market, for example. 

 

source: Statista 

Stargate will not be the only entity aiming to build such infrastructure. Macquarie Asset Management has committed $17 billion to investments in Applied Digital and Aligned Data Centers, focusing on AI and HPC infrastructure across the Americas. 

Amazon Web Services, Google Cloud and Microsoft Azure also are racing to upgrade their HPC capabilities.

Nvidia is collaborating with various companies to build AI factories and data centers as well, including CoreWeave. 

"Leave It All on the Field"

 I found this performance of "Baba Yetu" by the choir from Stellenbosch University in the process of doing other things. It is a rendition of the Lord's Prayer in Swahili, I am told. It soars. The musical precision and commitment are something else. I am reminded of a coach telling his players to leave everything on the field, to give it all. Total commitment and effort. Passion, intensity. So it should be for every human endeavor. 

LLM Hallucinations Should Not Surprise You: that the Math Works Really Should

Despite the occasional large language model hallucination (an instance where the model generates output that is factually incorrect or nonsensical), most of us likely are more than a little shocked that the applied mathematics actually works as often as it does. 

Despite the fact that we can call generative artificial intelligence a form of “intelligence,” LLMs actually do not “think” at all. In fact, LLMs are powerful statistical and probability engines. Having been trained on human language and language relationships, they infer (probabilistically) what the next word in a sentence should be. 

That should sound almost crazy, but that is the actual process, and illustrates one important reason computing is so powerful: the machines can process fantastically-huge amounts of data very quickly. In the case of LLMs, they are essentially trained on the sum total of human knowledge as principally found on the global internet 

If the input to the LLM is ambiguous or lacks sufficient context, the model may make incorrect assumptions and generate inaccurate information. 

But, basically, an LLM is applied math, a probability engine, working with “tokens” (think of tokens as “bytes”) that are essentially letters or punctuation. Models then predict the next individual characters in a sentence or group of sentences (for text operations). 

source: Lakera  

For visual analysis, image LLMs typically deal with visual elements that include “patches” (the image is divided into small, overlapping patches and each patch is a token). Or, images are analyzed by features such as edges, shapes, colors, and textures that are treated as tokens.   

In other cases, an image is analyzed to identify and locate objects within it that are considered tokens.

For music-focused large language models, tokens represent structured, interpretable chunks of musical information such as notes, chords, rhythmic patterns, beats, pitch and time intervals. 

The point is that all LLMs are predictive and probabilistic when providing output or analyzing input. What almost should shock us is how often LLMs “get it right” when doing so.

Intelligence "Too Cheap to Meter"

 

 

Feel free to agree or disagree with anything in this video, but "abundance" has been a Hallmark feature of computing, bandwidth, the internet and digital content. Decades ago, pondering the future of nuclear energy, a widely-heard phrase was "energy too cheap to meter." In other words, the cost of billing would be more than the cost of the product sold.

So now you might start to hear about "intelligence too cheap to meter." That might, or might not, happen, just as nuclear power never produced electricity so abundantly, at such low costs, that it ever made sense not to meter its use. 

On the other hand, most important computing innovations do involve the creation of scenarios where the use of computing cycles, inferences and storage become so affordable we no longer really worry about the cost of using resources to create inferences. 

I've always referred to this as "near zero pricing," where the constraint of using the resources was low enough that they ceased to be hurdles or obstacles to doing things with them. 

AI does not actually have to become so affordable it's cost is effectively "zero." It only has to be affordable enough that nobody really worries about using it. 

If Grocer Profits Rose About 1.5% Is that Evidence of "Price Gouging" at the Grocery Store?

Some people seem to believe that grocery prices exhibited evidence of price gouging because of the Covid pandemic. 

Price gouging represents what people consider “unfairly high” or “excessive” increases of the prices of goods, services, or commodities in relation to the normal market price or the cost of providing those goods or services. 

Such potential behavior typically occurs during emergencies or periods of market disruption when buyers have limited choices and are considered vulnerable. And the key is the notion of “unfair” or “excessive” increases. 

Economists always predict price hikes when supply is constricted or demand inflates, especially unexpectedly.   

Studies do show an increase in profit rates during the Covid supply chain problems, but some of us would also note that supply and demand also could explain all the increases. 

There arguably was more demand for groceries as consumers were confined to their homes and restaurants were shuttered for “on premises” dining. Supply chain disruptions caused shortages--and shortages tend to create price pressures. 

With plants shuttered and workers at home, production dropped. Transportation networks also faced slowdowns and materials shortages which, in turn, cascaded through the rest of the value chain. Measures to protect workers also slowed productivity, since time, cost and effort had to be shifted to “sanitizing” premises rather than producing. 

One Federal Trade Commission report suggests grocery retailer profit margins rose from about 5.6 percent to as much as seven percent.  

Other studies suggest profit increases because of Covid on the order of about 1.5 percent between 2019 and 2020, with margins declining afterwards. 

“Demand for many retail grocery products unexpectedly spiked upward during the pandemic, just as the supply chain was struggling with a series of input, labor, and transportation challenges,” the FTC report notes. 

Shipping demand increased while supply decreased, for example. 

The point is that one need not assume any “price gouging” to explain temporary profit increases at a time when demand grows and supply decreases. That basic dynamic would lead to higher prices, and often higher profits in any market similarly affected. 

Event	Good/Service	Imbalance Type	Key Consequence(s)
2020-2023 COVID-19 Pandemic	Toilet paper, hand sanitizer, computer chips, lumber	Excess demand, supply chain disruptions	Shortages, delayed construction projects
1970s Oil Crisis	Gasoline	Inadequate supply	Long lines at gas stations, fuel rationing, higher energy costs
2011 Japanese Earthquake and Tsunami	Semiconductor chips, automobiles	Supply chain disruptions	Production delays, higher prices for electronics and cars
Hurricane Katrina (2005)	Gasoline, food, building materials	Excess demand, supply chain disruptions	Shortages, price increases, delays in rebuilding efforts
2022 Russian Invasion of Ukraine	Wheat, sunflower oil, natural gas	Inadequate supply	Global food price inflation, energy crisis in Europe

And while there is no automatic and linear relationship between prices and profits, imbalances in supply or demand (less supply; more demand) almost always lead to higher prices. And temporary supply-demand imbalances arguably often lead to temporary profit gains. 

Situation	Demand	Supply	Price	Impact on Supplier Profits
Shortage of Housing	High demand for housing in desirable areas with limited construction	Limited supply of available homes	Increased home prices and rents	Increased profits for landlords, home sellers, and real estate developers
Concert Ticket Scarcity	High demand for tickets to a popular concert or event	Limited number of tickets available	Increased ticket prices, potential for scalping	Increased profits for the concert organizers, artists, and potentially scalpers
Collectible Item Craze	Sudden surge in demand for a rare collectible item (e.g., limited-edition sneakers, rare trading cards)	Limited supply of the collectible item	Significantly increased prices	Increased profits for collectors who already own the item and resellers
Natural Disaster Disruption	Increased demand for essential goods (e.g., batteries, flashlights, bottled water) after a natural disaster	Limited supply due to transportation disruptions and supply chain issues	Increased prices for essential goods	Increased profits for suppliers who can still provide essential goods, potentially even with unethical price gauging
Technological Innovation	High demand for a new, innovative tech product (e.g., a revolutionary smartphone)	Limited initial production capacity due to manufacturing constraints	High initial prices and potential for waiting lists	Increased profits for the tech company, despite potentially limited initial supply

Are Consumers More Comforable with AI Than Providers Believe?

As driven as suppliers might be to use artificial intelligence, consumer and user reactions are more complicated. As always, the usefulness of the innovation has to be grasped to be embraced. 

Consumers might be more comfortable with generative AI, for example, than providers expect. Students appear to be widely using it for purposes of writing essays, for example. But most consumers arguably already experience--and “use”--various forms of AI more passively, as when editing photos, using speech-to-text or searching for products to buy.  

Title	Date	Publisher	Key Findings
"2024 Consumer Study: Revolutionize Retail with AI Everywhere"	2024	IBM	Explores how AI can enhance retail experiences, particularly in inventory management and demand forecasting. Highlights the necessity of real-time data integration for building intelligent supply chains that meet specific customer needs.
"Consumer Perception and Use of Generative AI"	2024	Parks Associates	Quantifies consumer familiarity with and usage of generative AI applications. Indicates that these applications are often consumers' first direct interactions with AI, prompting new discussions about AI's capabilities and limitations.
"Consumers Know More About AI Than Business Leaders Think"	2024	Boston Consulting Group	Reveals that consumers possess a higher level of knowledge and excitement about AI than business leaders anticipate. Suggests that businesses should not underestimate consumer awareness and should engage more transparently regarding AI implementations.
"Consumers Open to AI in Marketing, But Data Privacy Matters"	2024	CDP.com	Reports that 81% of consumers are receptive to AI being used in marketing for personalized recommendations, provided that data privacy concerns are adequately addressed. Emphasizes the importance of balancing personalization with privacy.
"What Drives the Acceptance of AI Technology?: The Role of Expectations and Experiences"	2023	arXiv.org	Investigates factors influencing AI acceptance, finding that both direct experiences with AI and prior ICT experiences significantly impact acceptance intentions. Highlights the importance of managing user experiences to foster realistic expectations of AI technology.
"Consumer Acceptance of the Use of Artificial Intelligence in Online Shopping: Evidence from Hungary"	2022	arXiv.org	Examines consumer acceptance of AI in online retail, identifying trust and perceived usefulness as key factors. Suggests that enhancing content quality and automation can improve consumer attitudes toward AI-powered webshops.
"AI Is Ruining Your Laptops Now"	2024	Lifewire	Discusses consumer skepticism towards AI features in laptops, noting that such additions are often seen as unnecessary and resource-consuming, potentially deterring purchases. Highlights a disconnect between tech companies' promotion of AI and actual consumer preferences.
"Artificial Intelligence Marketing"	2024	Wikipedia	Describes how AI enables hyper-personalized advertisements by analyzing consumer data and patterns. Notes that while AI-driven personalization can enhance customer engagement, it also raises concerns about data privacy and the potential for intrusive marketing practices.