Listen to the audio version of this article (generated by AI).
Hello, Reader.
Three founders of the world’s leading AI companies walk into a bar…
One says, “We’re in a new era of rapidly improving AI.”
Another says, “We’re only at the beginning.”
The third says, “AI is already superhuman at many things.”
It sounds like the setup for a bad joke. But here’s the punchline: The people building the technology believe AI’s biggest transformation is still ahead of us.
OpenAI CEO Sam Altman, outgoing Google DeepMind CEO Demis Hassabis, and Tesla Corp. (TSLA) CEO Elon Musk have all been discussing the possibility of a dramatic leap in AI capabilities (bar not pictured):
- In the Relentless podcast last month, Altman said “I’ve been waiting for this my whole life.”
- Musk recently echoed the same view on X.
- And Hassabis published a Substack essay outlining how the world may be approaching a major new AI breakthrough.
At the heart of their vision is a powerful feedback loop: When AI becomes better, it can design even better AI models itself, and so on indefinitely.
Well, that loop is already beginning.
In June, Anthropic CEO Dario Amodei said that Claude is increasingly helping engineers write code, test ideas, and improve AI models – thereby accelerating the pace that new systems can be developed.
And this feedback loop could accelerate dramatically with the arrival of artificial general intelligence (AGI). This is when AI becomes as smart as humans.
Hassabis’s essay, mentioned above, explains why it’s so important:
This is a pivotal moment in human history. Artificial General Intelligence (AGI), a system that exhibits all the cognitive capabilities the brain has, is probably only a few short years away.
The magnitude of this technology’s impact will be unprecedented, perhaps 10x of the Industrial Revolution at 10x the speed.
But AGI won’t be just another tool. It will likely become a system that contributes to its own advancement.
For companies, that could mean surging demand for the infrastructure needed to power AI.
And there are already signs that this transformation is underway. OpenAI recently made a breakthrough that offers a glimpse of what’s ahead.
So, in today’s Smart Money, I’ll look at the signs that AI is moving toward AGI – and which companies are helping make that future possible.
AI Is Learning to Learn
OpenAI recently reported that its GPT-5.6 Sol model could improve its reasoning over time.
But first, a little context.
ARC-AGI-3 is a benchmark designed to test the type of intelligence that matters for AGI: adaptation.
Unlike a traditional test where an AI answers questions based on what it already knows, ARC-AGI-3 puts it in unfamiliar environments and assesses whether it can explore unfamiliar environments, learn from experience, build internal models, and plan actions.
In other words, it tests whether AI can learn how to solve new problems – something much closer to how humans learn.
Two weeks ago, the company found that its GPT-5.6 Sol model was much smarter than the benchmark originally suggested. By changing two system settings, the model was able to preserve parts of its earlier thought process during long tasks, similar to how we use notes. It was then able to build on that earlier reasoning to improve its performance.
That may sound like a small technical improvement. But it points to something much bigger…
AI systems are becoming better at adapting over time and demonstrating capabilities that could bring us closer to AGI.
OpenAI’s recent breakthrough suggests the next AI leap may come from systems that can remember, reason, and improve over time.
To turn it back to Hassabis:
AGI has the potential to be the ultimate tool for advancing science and medicine, and to drive enormous productivity gains and economic growth.
If AI reaches human-level intelligence, the biggest investment opportunity may not be the companies building AI models — but the companies supplying the chips, energy, memory, and materials needed to scale that intelligence.
Hassabis mentions a “precious window before AGI arrives.” For investors, that window could be an opportunity to get ahead of the infrastructure boom needed to power the next era of superintelligence.
Intelligence may be digital, but scaling it is a physical resource problem. The world cannot reach AGI simply by writing better software. It needs enough physical resources to run that software.
And that starts with the chips themselves.
Here’s the Punchline…
One company I’ve got my eye on is PDF Solutions Inc. (PDFS).
PDF Solutions helps semiconductor companies produce more usable chips from every manufacturing run. In simple terms, it helps chipmakers make more good chips.
PDF’s software identifies defects, improves manufacturing yields, and helps chipmakers reduce costly failures. Those capabilities become even more valuable when AI chips are in short supply and every usable chip counts.
If AI advances toward AGI, it could dramatically increase the amount of computing the world needs. As manufacturers race to increase chip production, companies like PDF could benefit from the growing pressure to maximize chip production.
The world’s leading AI builders all believe AI is entering a new phase where it learns, reasons, and improves more rapidly. If they’re right, every leap in AI capability will require even more physical infrastructure – making the suppliers of chips, memory, power, and data centers some of the biggest winners.
That is why, in my new free Market Shock presentation, I reveal 15 companies across the raw materials, energy, and memory spaces that I am watching very closely.
These companies, like PDF, are the picks-and-shovels providers behind the AI buildout. They may not be household names, but they could play an essential role in supplying the physical infrastructure required to power the next phase of AI.
And that, folks, is no joke.
Click here to access these companies – ticker symbols and all – for free.
Regards,
Eric Fry
