Why the Economics of GenAI Skews for the Tech Giants (And Why That’s Good for You)

When ChatGPT first took the world by storm, it seemed like this would usher an era of new entrants upending the tech giants. Nearly two years on, however, we are seeing that the leaders in cloud, search, and enterprise software are becoming even more influential when it comes to GenAI. 

This is due to two main factors that favor economies of scale for GenAI investments:

• Compute: First, the biggest players are investing billions scooping up as many of the GPUs being produced as possible as well as developing their own specialized chips. For everyone else the most reliable way to access GPU compute is often via cloud providers. 
• Data: Second, given the same data, open and close-sourced models converge on performance. Another way to put it is that data is the real differentiator. From an Associated Press article, “more data — acquired and ingested at costs only tech giants can afford, and increasingly subject to copyright disputes and lawsuits — will continue to drive improvements.”

So while you might find thousands of open and close source models available in Hugging Face, chances are that enterprises and users will still rely on the tech giants to provide the underlying backbone to many GenAI services given their better access to compute and greater volumes of data. 

If anything, however, this is proving to be a huge benefit to small to midsize businesses; Pricing for AWS’s premiere generative AI service, Amazon Bedrock, starts at $0.0003 per thousand input or output tokens (as of April 2024) and even less than a penny for multi-modal workflows. 

Pricing for Amazon Bedrock generative AI services often come in at hundredths of a penny per thousand tokens. 

At this price, any size business can immediately start integrating world class generative AI capabilities into new or existing products and workflows, taking advantage of a mature API ecosystem, rather than spending months training a model that likely will not surpass the performance (accuracy and time/compute required to get responses) of any of the larger models available. 

In our work with Fortune 500s as well as small to midsize businesses (SMBs), we have been able to vastly accelerate their GenAI initiatives by focusing not on building from scratch, but taking advantage of the best microservices available to build more reliable and secure applications. 

If you’d like to see what we can do for your business, reach out to us to schedule a free, 60-minute strategy consultation where we work with your product and technology teams to understand the right AI use cases for your business with your tech stack.