Will There be an AI "Killer App" or Use Case?

It should come as no surprise that Nvidia is the clearest beneficiary yet of the emergence of large language models. If AI in general is destined to become a general purpose technology, one would expect infrastructure to be the starting point for commercial deployment. 

Initial infrastructure is always the starting point because without it, no technology can become a GPT. 

For electricity, power plants, transmission lines, and distribution networks were the starting point. For the internet: communication networks and internet service providers were the beginning. 

For artificial Intelligence of any sort, it is specialized hardware (GPUs, TPUs), software development tools and data centers equipped to provide LLM training and inference operations. 

Typically, there are lead industries or use cases. Electricity saw early deployment in manufacturing, transportation and lighting.

The internet lead to disintermediation (replacement of roles in the distribution function) in retailing communication and media. 

We can expect similar progressions for AI as well. Infra is the initial focus, but then lead use cases will emerge. Steam power, for example, found early application for pumping water out of mines, before other use cases opened up. 

GPT	Initial Infrastructure Focus	Industries Gradually Affected	Examples
Electricity (1870s-1910s)	Power generation and distribution networks	Manufacturing, transportation, communication, homes	Factories shifting to electric motors, streetcars replacing horse-drawn carriages, telephones powered by electricity, electric appliances.
Internal Combustion Engine (1900s-1940s)	Fuel production and refining, car manufacturing, road networks	Transportation, agriculture, construction, leisure	Rise of automobiles, tractors replacing horses, trucks revolutionizing freight transport, car tourism and road trips.
Internet (1990s-2000s)	Telecommunication infrastructure, web browsers, servers	Retail, finance, media, education, communication	E-commerce, online banking, digital media streaming, online learning, email replacing traditional mail.
Mobile Internet (2000s-2010s)	Cellular networks, smartphones, app development	Communication, navigation, entertainment, social media, healthcare	Texting and mobile calls replacing landlines, GPS navigation apps, on-demand services, social media platforms, remote patient monitoring.

Between 1712 and 1775, for example, steam engines were used to pump water out of mines. 

From 1775 to 1800, steam power was harnessed for iron and steel production. James Watt's improved steam engine, with its higher efficiency and reliability, powered bellows in ironworks, replacing inefficient waterwheels and enabling larger furnaces and increased steel output. The industrial revolution was built on it. 

From 1785 to 1830 steam power was applied in the textile industry. Steam engines replaced water wheels in textile mills, for example, leading to mass production.

From 1800 to 1850 steam power was applied in transportation, powering steamboats and locomotives.

After 1850 steam power was progressively applied in agriculture, printing, papermaking, food processing and other industries where mechanization became possible. 

The clear point is that AI investment and operations will begin with infrastructure. Only later will we develop and create commercial applications, and then likely in stages. Not every industry, function or use case will be equally compelling at first. 

As always, we will be looking for the AI equivalent of “killer use cases or apps.”