I recently read an article in the Wall Street Journal titled “A Way for Energy Investors to Ride the AI Boom.”
The gist of their thesis is that AI data centers are power-hungry monsters that need increasing amounts of energy.
While this thesis makes obvious sense to us here at Equifund – especially if we see oil prices continue to climb higher…
We think the more interesting question to ask is the opposite – how can AI investors ride the prospective oil and gas boom?
More specifically, how are oil and gas companies utilizing large language models to crunch their massive amounts of geological data sets to drive efficiencies?
That’s the topic of today’s issue of Private Capital Insider.
Let’s get into it,
-Jake Hoffberg
How AI is Transforming Oil and Gas
For those who are new to Private Capital Insider, we’ve been tracking two major market narratives – the somewhat paradoxical “most predicted recession of all time” Doom narrative and the “Just Add AI!” Greed narrative.
On one side of the equation, we’ve got what appears to be a solid bull case for commodities and other real assets – especially if inflation continues, a market crash wipes out financial assets, or some combination of both.
On the other side, we’ve got the infinite hype machine, pushing a somewhat messianic view of AI’s impact on increasing productivity and reducing costs amidst a growing labor crisis – but also, it could destroy humanity in a Terminator / Matrix style event.
But every coin has three sides – heads, tails, and the edge that separates them.
So how do we combine both of these narratives into a single “super stock” thesis?
Simple: look for commodity producers that have the ability to dramatically increase free cash flow, thanks to a combination of higher commodity prices, increased productivity, and lower net costs.
And if we had to pick the two commodities that seem to be most sensitive to inflation – and also happen to be performing this year – it’s Oil and Gold (both of which have outperformed the S&P 500 year-to-date).
The problem with growth is that it has limits, and at the heart of those limits is the oil price.
When demand rises broadly in the economy, the oil price rises, and more of consumers’ money is spent at the pump.
Too much demand, and the rising price means the rest of the economy loses out.
The same goes for other commodities, to a lesser extent, with manufacturers surveyed for the latest S&P survey of purchasing managers reporting sharply higher input costs in March and raising their own prices at the fastest rate in almost a year.
The danger for investors is that recent rises in inflation-sensitive oil, gasoline, copper and gold … are a signal that economic growth will lead to a second round of inflation, forcing investors to flip back to worrying about rising prices again.
So how does this connect to AI-powered oil and gas stocks?
It all comes down to whether or not you think we’re looking at a soon-to-be realized supply side squeeze, similar to what we saw in the 1970s (please refer to our April 6th Weekend Edition for more on this).
Here’s why…
Energy companies rely heavily on data and analytics for innovation, and are built upon increasingly nuanced and complicated processes…
Especially as we see the explosion of sensors – and other “internet of things” (IoT) enabled devices – being utilized throughout the supply chain.
According to a new research report from the IoT analyst firm Berg Insight, the installed base of wireless devices featuring cellular, satellite, or “Low-Power Wide-Area” connectivity in the oil and gas industry is forecasted to grow at a compound annual growth rate of 19.3%, from 7.8 million units at the end of 2023 to 18.8 million connected devices by 2028.
This growth is primarily driven by remote monitoring of assets, such as industrial equipment, tanks, and pipeline infrastructure, in the midstream and downstream segment.
But now the question becomes, “What do you do with all that data?”
Enter: The AI Magic of Machine Learning and Large Language Models
We’ve kind of had an emergence of machine learning and artificial intelligence in oil and gas in the operation side of oil and gas and it kind of it kicked off in around 2005
One of the Independents then – Hess – was doing development, and they had an approach that involved Lean Sigma. But they also introduced machine learning algorithms, primarily multi regression modeling, to optimize completions.
The reason for that was of course the expansion into the unconventional reservoirs in North America, which coincided around that time it kind of exploded, which was enabled by advanced drilling techniques (horizontal drilling) and the advancement and completion techniques over long laterals.
That was kind of what kicked the whole ball rolling.
The result of that kind of approach in technology terms was we generated a huge amount of data – which previously we had but it took an awful longer time to generate that kind of data.
So that’s kind of what kicked off the whole machine learning and AI effort in the oil and gas operational side.
It had existed to some extent in some of the seismic and signal processing side, but in operations that’s where it really kicked off.
Drilling now in unconventionals can just be a few days versus what used to be several weeks months and and sometimes even longer
It also is less capital intensive than the traditional deepwater offshore operations, and generates obviously faster returns, which is kind of what we’re all looking for.
The data sets are massive and the unconventional reservoir is actually where we focus on most of the efforts that we do modeling for today at ConocoPhillips.
To summarize – oil and gas companies have been experimenting with what is now considered “artificial intelligence” for nearly a decade.
Why? Because oil and gas is an extremely, mathematically intensive sector – especially the upstream segment.
And somewhat ironically, there’s a major challenge with integrating artificial intelligence models into the oil and gas industry – classically trained engineers aren’t computer scientists that can build AI models.
Naturally, this means there’s no shortage of skepticism from the engineers who’ve had success using classical engineering techniques taught for the last 100+ years.
But despite this resistance, all of the top 20 global oil and gas producers – be they state-owned entities or public-listed ones – have a clear AI strategy for their upstream (i.e., exploration and production), downstream (i.e., processing and refining) and, where applicable, midstream (i.e., pipeline and logistics) businesses.
Here’s a few highlights worth noting:
- AI Market Growth in Oil and Gas: According to a report by Future Market Insights, AI in the oil and gas market is estimated to grow from $3.5 billion in 2024 to $13 billion by 2034, with a 14.1% compound annual growth during the forecast period.
This growth is driven by the need to optimize production, improve safety, and reduce environmental impact.
- Strategic AI Investments: According to Forbes, more than 92% of oil and gas companies are investing in AI technologies, with their capital expenditure on AI projected to hit $2.38 billion by the end of 2023, rising to $4.21 billion by the end of 2028.
- AI Accelerating Fossil Fuel Extraction: Global Witness highlights that major oil and gas companies like Shell, TotalEnergies, and ExxonMobil are investing in AI to speed up the extraction of oil and gas.
Shell announced the use of generative AI to hasten oil and gas exploration, and BP’s venture capital arm invested in an AI company to unlock critical data for its operations.
- AI Importance and ROI: An IBM report states that 56% of oil and gas executives surveyed consider AI important to the success of their organizations, with this number expected to increase to 84% in three years. AI investments have generated an average 32% return on investment in the past year.
- AWS AI Innovations for Oil and Gas: Amazon Web Services announced five generative AI innovations for the oil and gas industry, aiming to enhance employee productivity and transform businesses. This includes services like Amazon Bedrock and Amazon Titan Embeddings, which are designed to help oil and gas organizations build new generative AI applications.
- Databricks’ Data Intelligence Platform for Energy: Databricks launched the Data Intelligence Platform for Energy, which enables companies to better forecast load, predict outages, and balance supply and demand. The platform is being used by industry leaders, such as Shell and TotalEnergies, to optimize energy infrastructure and mitigate volatility.
So if you’re on the “Just Add AI!” bandwagon and are looking for some creative ways to get exposure to AI that doesn’t include buying more NVIDIA stock…
There’s a real case to be made for why asset heavy commodity producers – like oil and gas – make for a compelling investment.
Instead of being forced to accept nosebleed valuations based on the potential future value of the intellectual property…
Valuations in AI-powered commodity producers should be reflected by the actual financial results – both in the volume of commodities produced, and the profit margin of commodities sold.
Looking for an AI-powered oil and gas play?
If so, you might be interested in checking out Pytheas Energy – an upstream oil and gas producer that is currently raising capital on the Equifund Crowdfunding Platform.
Disclaimer: Equifund does not make any buy/sell recommendations, provide individualized investment advice, or otherwise “endorse” any specific investment opportunity – including ones listed on the Equifund Crowdfunding Platform. Please do not make any investment decisions based solely on the information published in this article.
Here’s some quick backstory on how the company was founded:
CEO Josh Zuker and COO Geoff Brandt have owned and operated a management consulting firm for 10 years, presiding over a number of various projects.
In 2020, Josh was introduced to an operating oil field in the Permian Basin, which was in danger of being shut down due to regulatory concerns, and he was asked to execute a turnaround.
While re-engineering that asset’s operations, Josh and Geoff met Hal Matheson, who was instrumental in raising initial investor funds.
Living through a “crash course” in the oil and gas industry during that successful experience, the team secured the initial ~$1.5 million in initial funds to begin operations of what would eventually become Pytheas Energy.
According to Zuker,
Oil and gas is an extremely inefficient business – especially in the smaller, “Craft Oil” side of the business.
A lot of the properties we have either already invested in – or plan to invest in – have equipment that is 15-20+ years old.
And it’s not like the original drillers installed top of the line equipment here, either.
This means these wells need to be manually turned on and off each day – or are, at best, set up on a pool timer to handle automating the on/off switch and intermittent pumping times.
This, among other things, means a lot of these assets have been mismanaged – especially when they’re run by small, independent companies with poor cash flow management and poor asset management principles.
Coming from a technology background, I immediately saw the opportunity to improve operating margins on these wells by using already available technology solutions – like sensors and remote monitoring – to determine when the best times of day to pump oil are.
If more of these wells could be automated, it would mean we wouldn’t need as many “boots on the ground” to manage the assets in place.
Other businesses I’ve looked at have something like 25 full-time employees at $25/hour, seven days a week. With simple technology upgrades, I realized I could run the same asset with 5-6 people.
While Zuker was already bullish on the long-recognized opportunity to modernize operations within the oil and gas sector, the release of ChatGPT represented an even bigger opportunity to implement machine learning and artificial intelligence.
Generally speaking, oil and gas wells generate an enormous amount of data that, often times, must be manually transformed and interpreted by highly skilled labor.
However, by leveraging large, language models, Zuker believed he could significantly improve operating margins through data-driven asset management.
Once ChatGPT came out, that was when I saw the opportunity with large language models. I wanted all the wells to talk to each other and learn from each other.
If I can find a flow pattern of oil progression in the ground – with wells spaced two acres apart – I should be able to see the migration of oil across the entire play.
What makes our assets more valuable than the guy next door? It’s because we have more data.
I think the improved data is the whole purpose. We’re going to be getting actual data – not just relying on some guy saying “hey, what was historical production? Four barrels per day? Let’s write down three.”
That’s what’s been going on out there for a long time. I think the value of the data increases the value of the asset tremendously because you can rely on it.
Think about it this way…
If you had two similar assets for sale, which one would you buy?
One with human-reported data? Or data reported from sensors on the well and other technology?
That’s why we’re excited about how AI can fundamentally transform how these assets are valued.
To learn more about the company’s capital raise – as well as the investment thesis behind the hidden opportunity in better management of low-volume “stripper wells” – go here now.
