I recently met with Greg Crennan, Founder and CEO of Coastal Journal, to get a different perspective on what the financial markets are signalling about this pending Generative AI bubble.

Not all is going well with Generative AI.

Since 2023, Sam Altman and Jensen Huang have been touting the need to invest in compute to support the $2 trillion already spent on LLMs and accelerate their growth to almost $3 trillion in data centres. This move has led to massive debt among hyperscalers like NVIDIA, CoreWeave, and Oracle.

AI promised it would automate jobs - Goldman Sachs predicted 300 million full-time jobs would disappear. – in 2025 there have been numerous reports that the job losses or decline in job-entry hires were not the result of AI but rather an inflation surge, driven by pandemic supply and demand imbalances – marked the start of the Federal Reserve rate hikes and the cost of capital that had exploded overnight — which led to drastic cost cutting measures including firing juniors and limiting new hires. Jing Hu and I wrote about this recently.

2025 had also proven to be less disruptive than was previously expected. Two-thirds of respondents in a McKinsey report said they have not yet begun scaling AI across the enterprise. Curiosity with agents was just that, with 62% indicating they were still experimenting with the technology.

The tech is not working. LLMs are great for pattern recognition and next-word prediction, but they are rife with errors.

There are countless examples of AI doing the entire job, only to have a human step in to remediate the outcomes and right the ship. People have become super “prompters,” getting the exact output they intended from countless prompts. Did they save time? Perhaps, but was this the level of prompt understanding that users envisioned? Certainly no.

And what is the ROI from the outcome that includes a human in this loop? For AI to be valuable, I read that it has to replace high-wage workers who can spot and fix those errors. After all, that is not the goal of automation.

What we’re also seeing are other behaviours — signals that perhaps speculation about this bubble is legitimate afterall.

* The massive investment in data centers and the ensuing debt among hyperscalers

* A circular investment within AI tech that is confusing investments for revenue

* The spurious chip inventory levels in NVIDIA remain high.

The latter three are the areas I spoke with Greg Crennan about. The numbers don’t lie, no matter how Big Tech hypes their performance.

Finally, Greg gave me some early insights about Google’s $20 billion deal with Apple.

Enjoy!

About Greg Crennan

Chief Market Strategist | Founder, The Coastal Journal

Macroeconomics | Forensic Accounting | Market Liquidity

As Chief Market Strategist at Golden Coast Consultants, I identify market price divergences from economic reality, prioritizing capital preservation over narrative momentum. My work includes early calls on gold (126%) and silver (165%) as core assets during the fiat debasement cycle, which have been top performers over the past five years heading into 2026.

My approach is based on Austrian economics, business-ownership principles, and forensic accounting, avoiding technical speculation or headline-driven narratives. I analyze how liquidity, balance sheets, incentives, and accounting distort prices and how these distortions are resolved. This has earned me the nickname “The Punisher” for applying math and fundamentals where belief systems often prevail.

I also founded The Coastal Journal, an independent financial research publication on Substack, which has grown rapidly through organic readership.

Thanks for reading System Malfunction! This post is free to consume and to share. Please let me know how I’m doing!

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit systemmalfunction.substack.com/subscribe