Colin’s Note: For newer Bleeding Edge subscribers, I’ve often said that the artificial intelligence boom will occur in three phases: hardware, software, and finally, everywhere.

We’re in the deep end of the hardware phase right now as chipmakers are still flying high after soaring to prominence in 2023… and tech companies are building out more data centers to meet the rising demand of AI.

But we’ve got our eye now on an overlooked industry as we wade into the software phase that’s tailor-made for an AI-powered boost – forecasting.

Right now, humans and their computer systems pore over massive data sets to make weather forecasts… but spotting patterns and poring over data is what AI models excel at. It’s what they’re made for.

I get into all the details in today’s Bleeding Edge… Just click below to watch or read on for the transcript.

Bleeding Edge subscribers, hopefully, guys are doing well.

Look, we’ve been going full speed on artificial intelligence (AI) since I joined in June 2023. The blueprint for hardware, software, everywhere has been playing out almost maybe too perfectly.

Broadly diversified semiconductor ETFs are up 75% just over the past year. And if you look at the individual stock performances of companies like Nvidia, AMD, and Supermicro, it’s been even better than that.

The truth is the hardware cycle shows little sign of slowing down. Although, you have to remember that Wall Street expectations are higher now, and the stock prices reflect that.

Remember that it’s the large corporations and, in some cases, sovereign governments that are buying the high-end AI supercomputers, not your average consumer.

So when your favorite permabear Macro Pundit is out there talking about the looming debt crisis, inflation, consumer confidence… That’s not who’s buying semiconductors and driving this explosive growth that we’ve seen.

Instead, it’s Microsoft, Amazon, Google, and other buyers that are buying these semiconductors. And at last check, these companies have a combined hundreds of billions of dollars in cash in the bank… and a Federal Trade Commission and a government that tries to block every acquisition these companies attempt to make.

So in many ways, larger technology stocks have nothing better to spend their money on than computer chips from Nvidia. The software phase of AI is taking shape as well.

Large incumbents like Microsoft have layered its AI Copilot onto 30-year-old software and are squeezing more dollars out of their corporate clients.

As the semiconductors continue to be installed inside of these data centers, expect more software and services to arrive. But other than Microsoft and other large enterprise software providers, what AI software companies will do well in the age of AI?

One area that is flying under the radar just a little bit and that it’s tailor-made for AI is the forecasting industry.

Forecasting, or predicting, relies on large data sets that humans – along with computer systems – pore over. In the past this industry used buzzwords like “big data” or “analytics,” but the reality is AI will help forecasting take a large leap forward in the coming years.

Take, for example, Google’s Flood Hub. The online tool has been available for about six years now but more recently has been able to expand its reach thanks to AI.

Floods are among the most common natural disasters, impacting around 20% of the world’s population. Google has been trying to predict when floods might occur, so it can alert the public before the natural disaster occurs.

With the use of AI, Google can predict floods up to seven days in advance. That includes not only populated areas of the world like here in the United States but also remote regions in Africa. Similar techniques are being applied to hurricanes and other natural disasters… hopefully, potentially saving millions of lives.

One interesting study about forecasting just released by UC Berkeley out here in California found that AI could predict future events better than the average human.

Using a custom-tuned model from OpenAI, the team of scientists at the university could predict future events with an average accuracy of 71.5%.

One subject that the AI model struggled to forecast was finance, in part because models like OpenAI’s GPT-4 liked to hedge answers for safety. So I guess my job is safe from AI, at least for now.

One final note about these forecasting models… they’re really expensive to run. Google is using satellites and data processing power that is more or less subsidized by Google search profits.

The study conducted by the university professors cost nearly $1 per query. So every time they asked a computer system a question, it cost them a dollar.

That’s a sign that the software phase of AI has hardly begun. The cost to conduct experiments and launch products will drop significantly in the coming years as more hardware is delivered to these large data centers.

Believe it or not – and to use a baseball reference – we’re still in the early innings of the AI software revolution. We’ll certainly be here to cover it for you on The Bleeding Edge. That was the newsletter for today. Hopefully, you guys have a great day. I’ll see you again soon.