Why the AI Boom Is Happening Now
Sophisticated AI systems like ChatGPT and Copilot require two things: enormous amounts of computing power and massive datasets on which to train.
Those assets have only become available in recent years.
The rise of cloud computing
In the 2010s, cloud computing services like Amazon’s AWS (Amazon Web Services) and Microsoft’s Azure offered corporations near-unlimited computing power for a subscription fee. These cloud services are grounded in remote data centers, warehouse-size buildings that require a large and continuous supply of power to operate and cool the servers.
In the era prior to cloud computing, developers requiring massive amounts of computing power (known as “compute”) had to purchase or lease time on rare high-end machines like the Cray supercomputer. With cloud computing, developers could scale up computing power without having to own the machine.
Evolution of the Nvidia chip
Computing power also took a leap forward in recent years with the development of Nvidia’s AI chips and systems.
Intel dominated the computer chip market for years with its CPUs—central processing units that served as the foundation for computing processes.
Nvidia focused instead on GPUs—graphical processing units meant to serve the growing computer gaming market. Nvidia and AI developers discovered that GPUs could perform calculations concurrently in a way that CPUs could not. This made Nvidia chips more energy efficient and better able to handle the sophisticated computing demands of AI developers.
The growth of datasets
The expansion of cloud computing allowed for the creation of ever-larger digital datasets. YouTube videos became a global phenomenon, creating 30,000 hours of new content every 60 minutes. The music industry moved online, digitizing millions of songs. Unauthorized troves of scraped data like Books3—which contained the complete contents of more than 191,000 copyrighted books—appeared on the web. Datasets grew large enough to provide the billions of data points required to effectively train AI systems.
These public (but often legally protected) datasets were used to train generative-AI systems by Meta, Bloomberg, and others. OpenAI, the developer of ChatGPT, has never released information about the datasets used to train its systems.
The race to “own” AI
With Big Tech consolidated in the hands of a small group of global corporations—Alphabet, Microsoft, Google, Apple, Meta—there is a widespread assumption that the companies that capture the early AI market will survive and thrive, while those who do not will fail. For Big Tech companies, the race to “win AI” has become an existential competition.
Many investors, meanwhile, see AI as the next big gold rush. Those who were early investors in Microsoft and Google were handsomely rewarded, and many see a similar opportunity in the opening of the AI era. That has resulted in a massive influx of investment capital into the AI space, further propelling the win-at-all-costs frenzy.
Next: