All Articles

How AI Started A Global Race For GPUs

July 27, 2023
x min read

The words artificial intelligence had at best a vague meaning in the cultural lexicon at the beginning of this year. And then, in an instant, OpenAI released ChatGPT and took the globe by storm.

While it was hardly the first ever artificial intelligence project to be launch, it was the first one that experienced rapid consumer usage, rocketing to a staggering 100M individual users virtually overnight.

Suddenly, the AI movement was happening. Everyone was talking about AI and everyone was trying to fund it.

Generative AI startups and projects seemingly spawned from every angle, with large scale funding rounds and nearly limitless demand from venture capitalists and angel investors.

But, there was one problem:

who would supply the computing requirements?

The Big Picture:

  • Artificial intelligence has a co-dependent relationship with computing. In order to build and train Large Language Models, you need GPU power and compute to do so
  • Currently, there is a supply and demand issue around finding GPUs that have a reasonable cost

The Landscape:

  • Nvida (NASDAQ: NVDA) has been the long standing dominant force of GPU and cloud computing, with their H100
  • Advanced Micro Device (NASDAQ: AMD) is the next leader, with their MI300X

How Does It Work:

  • Simply put, every AI startup is building or training technology that requires high levels of data inputs and processing power
  • GPUs, which differ from CPUs, are able to handle the high computing requirements of machine learning, data analysis and engineering capabilities required for AI
  • The more data you can give your Large Language Model or generative AI project, the more machine learning can do to build a better end product. This requires enormous computing resources

Market Reaction:

  • NVDA is up +211% year to date
  • AMD is up +72%

Venture Capital Reaction:

  • The demand is for GPUs is so high, many VC’s are deploying capital directly into owning “clusters” of GPUs
  • One such instance is Daniel Gross and Nat Friedman, who put $70M into their own computing cluster to be used exclusively for their portfolio companies
  • Cloud computing company Coreweave (private) raised $200M at a $2B valuation at the end of May, with participation from NVDA
  • Many in the startup world have been outspoken about how dislocated supply and demand is for GPUs
  • “There’s a full blown run on GPU compute on a level I think people do not comprehend right now. Holy cow.” - Suhail Doshi, founder of Playground AI

Discover what’s next. The world’s biggest ideas, disruptive trends, most exciting early stage companies and groundbreaking entrepreneurs.

By clicking Subscribe you're confirming that you agree with our Privacy Policy.
Thanks for subscribing!
Keep an eye out for a welcome email shortly.
Oops! Something went wrong while submitting the form.