- Drifft Newsletter
- Posts
- Mineral Metamorphosis: How AI Is Reshaping Tomorrow's Industry
Mineral Metamorphosis: How AI Is Reshaping Tomorrow's Industry
Navigating Uncharted Territory: Shifts in Demand to Make or Break Investment Strategies
💡What’s New: How will AI impact the mineral and mining industries?
🤔 Opinion: The question everyone’s asking - Where should I invest?
🛠️ Tools & Data: The latest opensource AI for satellites and robotics
ATTENTION!
DRIFFT is hosting a webinar this August with quant trader, Ernest Chan, on the topic “How will AI impact mining?” Grab your spot here.
💡What’s new
The past few months have seen several mineral narratives gain momentum. AI-assisted exploration of minerals is generating investments. The threat of critical mineral shortages in the US continues due to lack of domestic refining capacity and export controls from China. (This week the US House established a committee to reduce foreign dependence on critical minerals.) Meanwhile, Argentina, which has a large supply of “transition metals”, is privatizing and opening to foreign investments in a way LATAM has not seen in over 100 years.
Probably the most interesting narrative which impacts long-term US development (and therefore minerals) has been the questionable future of the US Dollar. Saudi Arabia recently inked a deal which, though it continues trade of petrol for US dollars, opens the door for BRIC countries to purchase petrol in their own currencies. This seemingly small change is likely to have significant impacts.
From a technological point of view, there’s little doubt that electricity as an infrastructure and foundation for modern society should enjoy a healthy future. And it seems all but inevitable that a group of critical minerals will rise along with electricity demand. How could it not? Production and delivery of electrical energy depends on these minerals being processed into infrastructure.
But whether or not the US has chosen the right time to transition is unclear. And now the big question–the answer to which will impact mineral markets–is whether the US can sustain and grow efforts to modernize its infrastructure.
There are plenty of interesting transition projects underway. Examples include modular nuclear reactors, widespread manufacture of electric vehicles, and infrastructure development with key funding packages we’ve all heard of by now (CHIPS act, etc). Even space technology is expected to increase overall demand for minerals.
But if recent efforts are any indicator of the future, the transition seems like it will be anything but smooth. Several points of friction remain:
The US seems to be behind in electrical and mobility infrastructure.
US vehicle manufacturers cannot compete on price with China and must resort to import duties to dissuade US buyers from adopting less-expensive (not cheaper) Chinese vehicles.
Because there is relatively little charging infrastructure, consumers aren’t purchasing EVs in numbers that make a difference (in fact, there’s at least a 12-month backlog of EVs sitting on dealer lots).
Enough cars are being sold that energy brownouts are already common in areas like California.
A sudden growth spurt in data centers to power corporate and government AI is driving massive electricity consumption.
The release and consumer uptake of opensource AI models has caused manufacturers to respond by including even more energy-hungry compute capacity (mainly GPUs) into their products.
Surprisingly, these circumstances have only generated a normal amount of short-term volatility for copper futures.

Daily copper futures

Monthly copper futures - long-term bullish trend
From what I can see, further infrastructure development in an era of high inflation and average job growth (assuming AI doesn’t eat into it) will almost certainly guarantee that either government must partner with large corporations to fund development, and/or large private utilities will emerge with unique concessions from governments.
There’s room for both.
One of the more interesting trends I am tracking is the development of cities funded by private money from wealthy investors. Especially in areas close to US mines and technology hubs, these seem like ideal points for placing mineral industry assets. This is already happening in the Middle East and in the southern US.
Related News
This new insurance risk arbitrage exchange might provide an interesting way for mining companies to hedge risks at lower prices.
Robots will impact jobs, but also mineral demand.
Do people really understand how much AI will impact economics?
🤔 Opinion
In the past few weeks, several wealthy people have asked for my views on AI and investments. Here are a few of their questions:
Will NVIDIA sustain its growth?
Where can I allocate to get a return on the AI trend?
How can I plan for the long-term when businesses keep getting upended?
It’s true that there is more risk in an era with accelerated change. My opinion is that a lot of the risk is due to the fact that older investors with the bulk of discretionary investment funds do not understand new technology. Just try explaining AI to your grandfather (who likely has 1000x the cash reserves you do). You will quickly find he probably has no clue. Case in point: Warren Buffet didn’t invest in Apple until it had already achieved over 90,000% returns since its founding!
It’s my opinion that investors are in a new regime of investing, one where we’ll be forced to become nimble. A period of high, sustained inflation and continual business upheaval should be expected. In an era of accelerated risk, I think it’s more important than ever to education oneself on finance and the systems used to transact money and investments. It will not be enough to trot along with the index-fund herd.
The herd investment paradigm has concentrated wealth into a small group of managers–Vanguard and Black Rock–and has given way to meddling. At the same time, market capitalizations have concentrated into a small number of stocks (e.g., AAPL, NVDA, MSFT, GOOG). The index-fund approach now seems to resemble more of an individual stock approach. Moreover, when investors run portfolio optimization algorithms, I wonder if they’re weighting near-term volatility profiles to account for the macro changes in the investing environment. Financial legal jargon is constantly repeating that past returns are no indication of the future. And that also means the volatility profiles of yesteryear are not guaranteed to be stable. But a lot of people seem to be making investment decisions as if they were.
My criteria for investments have not changed. And they never will. From a business point of view, an investment makes money. If it doesn’t make money, it’s not an investment!
Today, when I tell people I am an AI engineer focused on mineral markets, they are puzzled. I like minerals because most technology people underestimate the difficulty in changing physical reality. There is little friction in cyber space. But everything slows down when it touches the real world. So while AI will grind away at SaaS business models, my expectation is for longer-term sustained growth in mineral commodities. We’ll see.
🛠️ Tools and Data
Incredible new tools for building the future with geoAI and satellites.
Sen2Nbar: https://github.com/ESDS-Leipzig/sen2nbar Nadir BRDF Adjusted Reflectance (NBAR) for Sentinel-2 in Python | ![]() |
SSL Data Curation: https://github.com/facebookresearch/ssl-data-curation Automatic Data Curation for Self-Supervised Learning: A Clustering-Based Approach | ![]() |
Spatio Temporal Asset Catalogue: https://stacspec.org/en a common language to describe geospatial information, so it can more easily be worked with, indexed, and discovered. | ![]() |
Use neural nets for snapshot compressive imaging, useful for hyperspectral satellite systems. |
Ames Stereo Pipeline : https://github.com/NeoGeographyToolkit/StereoPipeline Automated geodesy and stereogrammetry tools for processing stereo images captured from satellites, robotic rovers, aerial cameras, and historical images, with and without accurate camera pose information to produce digital terrain models, ortho-projected images, 3D models, and bundle-adjusted networks of cameras. | ![]() |
ICESat2VegR: https://github.com/carlos-alberto-silva/ICESat2VegR Functions for downloading, reading, visualizing, processing and exporting NASA’s ICESat-2 ATL03 (Global Geolocated Photon Data) and ATL08 (Land and Vegetation Height) products for Land and Vegetation Applications in R environment. | ![]() |
RoboFleet: https://github.com/ut-amrl/robofleet A system that enables ROS-based robot-to-robot communication, as well as visualization of robot data via a web-based frontend. Robots will run a client application to exchange data with a central server. | ![]() |
AI models, datasets, and tools for real-world robotics in PyTorch. | ![]() |
OpenVLA: https://github.com/openvla/openvla A scalable codebase for training and fine-tuning vision-language-action models (VLAs) for generalist robotic manipulation | ![]() |
A framework for training any-to-any multimodal foundation models. Scalable. Open-sourced. Across tens of modalities and tasks. | ![]() |
Thanks for reading! Want me to look into a particular topic? Email your suggestions and and I will dig.