AGI Has Arrived. Here's How I'm Investing In It.
Summary
Alex’s central thesis is that the release of GPT6 Astra signals the arrival of AGI because it can autonomously operate a computer via a user interface to complete multi-step tasks without human supervision. Alex believes this shifts the investment landscape from "instruction-based AI" to "goal-based AI," which requires significantly more compute power—estimated at 15 to 100 times more than traditional chatbots. Alex expects a massive expansion of the global AI market, projecting a 38.5% compound annual growth rate through 2034. Alex focuses on the hardware "stack" necessary to power these agents, moving beyond just GPUs to include critical components like CPUs, high-bandwidth memory (HBM), and silicon photonics.
Mentioned Stocks
Reasoning: Alex views Micron as the primary American beneficiary of the surge in memory demand. Alex notes that revenue more than tripled year-over-year and that the stock is historically cheap at a forward P/E under eight despite the structural shift in demand.
Reasoning: Alex has been a shareholder for over 10 years and views Nvidia as a core holding. Alex highlights Nvidia's new custom 'Vera' CPUs and their strategic $2 billion investments in laser suppliers like Lumentum and Coherent to control the full AI hardware stack.
Reasoning: Alex argues that AI agents require massive CPU power to run background tools and software. Alex notes that AMD's EPYC CPUs are integral to Microsoft's new agentic AI data center instances, giving the company high pricing power in a constrained market.
Reasoning: Alex describes Cerebras as a pure-play bet on inference speed due to its massive wafer-scale chips. Alex highlights a $25.4 billion backlog and notes that while the company currently loses money, it holds $8.6 billion in cash and provides hardware five times faster than Nvidia's B200.
Reasoning: Alex points to the 2.5 million Grace CPUs already shipping that use ARM architecture. Alex also highlights the $2 billion in demand for ARM's new 'AGI CPU' and doubling data center royalties, though Alex notes that manufacturing costs may impact future margins.
Reasoning: Alex is 'pounding the table' on Lumentum because they provide the light sources for optical transceivers. Alex notes that their revenues for co-packaged optics doubled year-over-year and are expected to double again.
Reasoning: Alex views Coherent as a broad leader in lasers and transceivers for data centers. Alex emphasizes that Nvidia's $2 billion commitment to the company is a strong signal of its essential role in the AI infrastructure stack.
Reasoning: Alex highlights Krato's specialized copper cables used inside AI racks, noting that this product line is expected to triple by 2026. Alex is bullish on the company's 115% year-over-year revenue growth and strong management guidance.
Reasoning: Alex states that Astera Labs makes critical controllers for shared memory and has seen revenues double. However, because the stock has recently undergone a massive rally, Alex is keeping it on a watch list for a better entry point.
Reasoning: Alex notes that Tower Semiconductor is a key foundry for silicon photonics, which are chips that convert electrical signals to light for data transfer. Alex points out that the company has already signed $1.3 billion in photonics contracts for 2027.
Reasoning: Alex identifies SK Hynix as the world leader in high bandwidth memory (HBM), which is critical because AI agents require memory to stay active for days. Alex states that the stock is currently undervalued, trading at a forward P/E of under eight.