Dying AI Jobs, Klarna's Mistake, & More

TLDR;

Welcome to your weekly Premium Insight Rundown. This week, we take a look at the fast progress we are seeing in humanoid robotics, concerning ‘bubbly’ trends we are starting to see in the data center business, the rapid fall into obsolescence of the hottest AI job, and the first great defeat of an AI early adopter, among other news.

Enjoy!

GPU EXPORTS

Malaysian GPU Imports Skyrocket

US GPU imports to Malaysia have exploded in the last quarters, with March growing to $1.96 billion, a 3,400% year-over-year increase, putting the total value for the year at almost $7 billion.

This wouldn’t be news considering Malaysia is rapidly developing a data center industry, but its close ties to China suggest a much ‘darker’ reality we have already commented on in this newsletter in the past: Malaysia is being used by Chinese companies to smuggle GPUs to avoid GPU export controls set by the US on companies from that country.

TheWhiteBox’s takeaway:

In my humble view, the fact that this is happening is hardly debatable. Chinese companies are becoming very witty in finding ways to circumvent export controls, which makes me believe this is simply a lost cause.

What the US is doing, far from crippling China, is simply preventing NVIDIA and AMD from making more money while also incentivizing China to develop its own chip design and manufacturing capabilities.

You can’t stop China from competing; you just can’t, and the data from places like Malaysia, Taiwan, and Singapore prove this point.

AI BUBBLE

The Growing Trend of Securitized Debt in Data Centers

An article by The Information has illustrated a growing trend in the data center business to securitize future cash-flows to be sold to investors in return for cash.

In simpler terms, data center companies are going to investors and offering their future profits plus interest in return for immediate cash to fund their ongoing operations and, more crucially, their expansions.

It’s a very similar trend to the one we observed back in 2008 with securitized mortgages, known as MBS (Mortgage Backed Securities) and CDOs (Collateralized Debt Obligations), which played a central role in the great crash of America’s Housing Market (and that of the world soon afterward).

Is this AI’s new reality?

TheWhiteBox’s takeaway:

Before we jump to conclusions and claim the AI Bubble is alive and well (that warrants a more detailed analysis), this doesn’t look like the same situation we saw back in 2008 because the debtors are totally different.

Back in 2008, the underlying risk was huge because the people paying back those mortgages were very risky tenants, as most of those loans were subprime. In this case, the loan tenants are companies like the Hyperscalers (Amazon, Google, Microsoft, Meta) or Neoclouds like Coreweave, companies with strong liquidity, especially the former.

At the bubble's peak in 2008, banks were giving out NINJA loans (No Income, No Job, No Assets), meaning people receiving the loan could not pay it back in a thousand years, so defaults were…

Enjoyed the introduction? Subscribe or sign in to continue with this edition.