Huge Losses, Failed Robots, & My Bet for 2026.

TLDR;

Welcome back! Today, we take a look at a long and exciting list of news, ranging from new deals by OpenAI to how Microsoft revealed OpenAI’s losses (they are huge), why AI robots are still far from being remotely useful, a concerning trend amongst Hyperscalers’ huge AI bets, and more.

We’ll conclude with my newest stock purchase, a company I believe is outrageously undervalued and potentially a huge stock winner in 2026.

TSMC/ASML-level bottleneck in the entire AI supply chain,

all 2026 revenue pre-sold…

Yet deeply ignored by markets amidst what could soon be one of the biggest supply squeezes in history.

Enjoy!

THEWHITEBOX

The First Space GPU

Two weeks ago, we covered the announcement that a startup supported by NVIDIA wanted to deploy data centers in space, named StarCloud.

At the time of the announcement, many pundits even ridiculed the idea, claiming that space would not serve as an effective heat dissipator. This is problematic because GPUs generate a lot of heat and thus require extensive cooling, originally with fans and nowadays with newer generations, such as Blackwell, utilizing liquid cooling.

But a few days ago, they launched the first NVIDIA H100 to space, as you can see in this video, so we will soon know who was correct.

TheWhiteBox’s takeaway:

The jury is out on this one. But if this company is correct and GPUs can be deployed in space, it would be amazing for several reasons:

Infinite and free energy from the sun

No grid interconnect permits

No neighborhood associations up your ass asking you to stop pushing energy prices upward or wasting water

No transformer and turbine lead time bottlenecks

Thus, this is one of two things: revolutionary or an embarrassing money-grab.

ENERGY

US Solar Panel Manufacturing Rise?

A video released by publicly traded company T1 Energy shows footage of their solar panel manufacturing facility in Dallas, which has reached 14 MW-per-day production capacity, putting the total output of this factory at 5 GW per year.

TheWhiteBox’s company:

In theory, this is unrelated to AI. Still, our Sunday analysis highlighted the importance of solar PV to the future of this technology, given its significantly shorter lead times (1.5 years, more than four times shorter than those of nuclear power plants).

And although markets have priced in Trump’s disenchantment with green energies, I believe most of his hatred toward them is focused on wind, and the appeal of solar to become a way to increase manufacturing onshoring, added to Elon Musk’s known continuous love for solar PV, means I believe we’re going to start seeing a narrative shift toward solar from the Administration.

And to make matters better, T1’s video clearly proves that there’s a future for US-manufactured, solar PV energy in the US; they just need to seize the opportunity.

RESEARCH

Meta’s Continual Learning Proposal

Continual learning seems to be the hottest thing in AI research right now, especially after Thinking Machines Labs (TML) tackled this problem (and proposed a solution we covered).

The problem…

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