AI Supply Chains: When Compute Becomes the New Oil

Computing power and information storage is becoming increasingly important as AI use spreads across developed economies.  A generation ago, relatively few personal computers were connected to the Internet.  Today, virtually all Western consumers are online and generating terabytes of information, from files to streaming video, that need to be stored in the cloud.  Producers and consumers are using AI software to answer questions and create digital products, ranging from text to audio to video clips.  As more and more people become consumers of AI, computing power and data storage will become the new “oil” - a crucial resource.

Data Centers as the New Factor Markets

Thanks to cloud computing and AI, most of what our personal computers do is not done on the devices’ hard drives.  Rather, data centers with powerful computer servers do the work, either processing or storing large amounts of information.  Those who build and own data centers have a tremendous advantage, as they can sell portions of this processing power or storage to others.  This is done by the cloud services industry, of which major tech firms like Microsoft, Google, and Amazon currently dominate.

Markets for use of data centers have already developed, with companies signing hefty contracts for guaranteed data use.  Consumers, also, can purchase paid subscriptions to AI software and enjoy significantly greater capabilities than the freemium versions available to all.  The cloud services industry will be subject to market forces, with the cost of AI and data storage services falling as more competitors enter the industry and rising as barriers to entry limit new firms.

Oligopoly Market for Data Centers

Due to the tremendous expense of creating data centers, which requires significant resources and lots of technical expertise, the cloud services industry is an oligopoly market dominated by a small number of producers.  The “big three” of Microsoft, Google, and Amazon currently control more than 60 percent of the market.  A growing foreign rival, Alibaba of China, is the fourth largest and controls about 5 percent of the market, followed closely by IBM with 4 percent.  

Could the Oligopoly Cause Shortages?

The relatively small number of cloud computing firms could put the market at risk of AI and data storage shortages; a problem affecting any single firm could affect thousands of customers.  Firms may also have less incentive to expand, and thus lower prices, if clients have signed multi-year contracts.  If many large clients have long-term fixed contracts, firms have little to gain by lowering their prices - they won’t be able to convince customers to switch providers in the short run.  While there is some incentive to expand and attract new customers, this may increasingly be limited to smaller clients and individuals, reducing the drive for large-scale innovations.

Economics of Cloud Computing

Substitution Effect Will Lower Prices

Fortunately, cloud computing firms are responsive to the threat of substitutes, with an increase in suitable substitutes for cloud computing services driving down the cost.  For example, more advanced computer chips can increase computing power within hard drives themselves, reducing the need for cloud computing.  As the cost of computer servers decreases, more firms will be able to purchase their own data storage and data processing devices, reducing their need to outsource these services to cloud computing companies.  Many firms, especially those that handle large volumes of sensitive data, may be quick to spring for their own servers even when the cost remains greater than cloud computing subscriptions in order to keep this data secure.

Supply Shortages Will Raise Prices

However, firms that hire cloud computing services may struggle to access substitutes like independent servers due to supply shortages.  Currently, there is a shortage of GPUs, or Graphic Processing Units, which is slowing the production of advanced computers.  These complex chips are very expensive, meaning their consumption is dominated by the “big three” cloud computing companies.  Thanks to their ability to buy in bulk and negotiate directly with GPU producers like NVIDIA, cloud computing firms will be able to process AI more cost-effectively than small firms with their own individual, and more expensive, servers.

Government Policies and Geopolitical Tensions

As more companies come to rely on cloud computing firms for their AI processing and data storage, fears arise over these two digital functions becoming tools for espionage, sabotage, or manipulation.  When cloud computing is done by a foreign company, what is to prevent a foreign government from accessing and manipulating data?  Such was the fear over the recent Tik Tok drama in the United States, with many policymakers fearing that owning firm ByteDance might allow the Chinese government to access the data of American users of Tik Tok.

Geopolitical tensions are rising between the United States and China as the latter, long considered technologically inferior to the West, has made tremendous strides in computer production and AI processing.  Some Americans fear that China’s AI improvements could give it economic and military advantages, especially since current generative AI use is significantly regulated by the government.  For its part, China harbors some resentment toward the U.S. and its allies for using export controls to limit the access of Chinese firms to advanced microchips needed for AI.

Will Export Controls Play a Major Role in Limiting AI Growth?

The U.S. has placed export controls on advanced microchips to China, while China has placed export controls on rare earth minerals, necessary for the creation of those microchips, to the United States.  These dueling export controls are currently limiting the production of advanced computers on both sides of the Pacific, likely limiting AI expansion.  What remains to be seen is whether both powers choose to compromise and reduce their export controls, allowing for mutual technological growth, or attempt to strengthen those controls to “choke out” tech expansion by their rival.

Both nations will undoubtedly seek substitutes for the resources dominated by the other, with the U.S. seeking new sources of rare earth minerals and China seeking to develop domestic chip-making capabilities.  However, both nations are in a bind, as they and their allies lag significantly behind the other side in those respective areas.  It will take many years for the U.S. to reach China’s ability to process rare earth minerals, and it will take many years for Chinese companies to reach the ability to make microchips as refined as their American rivals.