💰 SoftBank completes $41 billion investment in OpenAI, deepening bet on AI ↗
SoftBank says it’s fully funded a roughly $41B commitment into OpenAI - putting it at around an 11% ownership stake. That’s a loud, unmissable vote of confidence.
The money reportedly came in tranches, with a large final chunk to close it out. Translation: Son is still treating AI like the main plot, not a side quest.
🏗️ Brookfield to start cloud business amid AI frenzy, The Information reports ↗
Brookfield is reportedly launching a cloud business (Radiant) to lease chips inside data centers directly to AI developers. Not just “we own buildings” - more like “we’ll sell you the engines too.”
It’s tied to a new multi-billion AI fund and data center projects across multiple countries. Clean logic: if compute is the new oil, everyone wants to be the pipeline.
🧠 ByteDance to spend about $14 billion in Nvidia chips in 2026, SCMP reports ↗
ByteDance is being linked to a massive Nvidia shopping list - around 100B yuan (about $14.3B). It’s framed as conditional, though, depending on what hardware can be sold into China.
Even with that caveat, the direction is clear: the biggest AI players are turning into chip black holes, and the rest of the market has to orbit around them.
🕵️ Proton claims Google trains AI on Google Photos albums: does it? ↗
Proton’s team suggested Google’s image-gen edge comes from scanning users’ Google Photos libraries. The claim took off, but the evidence in the public back-and-forth looks thin - or at least not displayed.
The piece points to Google’s stated position that it doesn’t train generative AI models outside of Photos using your personal data, while also noting Photos analyzes images for features and safety. So the suspicion is relatable, and the missing ingredient is still proof.
🚗 Google AI Giving Wrong Information On US EV Tax Credit ↗
Google’s AI Overviews got called out for telling people they could still get a $7,500 EV tax credit, even though the credit had ended. That’s not a harmless typo - that’s money.
A check across a batch of EVs reportedly found the AI was wrong on most of them while sounding extremely sure of itself, which is the worst possible posture. Confidently outdated is a dangerous flavor.
🛡️ BigBear.ai Finalizes $250M Acquisition of Ask Sage ↗
BigBear.ai says it has completed a $250M cash acquisition of Ask Sage, leaning hard into “secure, regulated genAI” for government and other locked-down environments.
Ask Sage is described as serving large numbers of users across many government teams, and the combined pitch is governance-first AI - less razzle-dazzle, more “this won’t get you fired,” which tends to win procurement.
FAQ
What SoftBank’s $41 billion OpenAI investment means
SoftBank says it has fully funded a roughly $41B commitment into OpenAI, giving it around an 11% ownership stake. The funding reportedly came in tranches, with a large final portion completing the commitment. In practical terms, it signals SoftBank is treating AI investment as a core strategic priority, not a tentative side bet. It also underlines how concentrated major AI ownership and capital can become.
How Brookfield’s new “Radiant” cloud business differs from typical cloud services
Brookfield is reportedly launching a cloud business called Radiant to lease chips inside data centers directly to AI developers. Instead of only owning or operating data center real estate, the pitch is closer to selling access to the compute “engines” as well. It’s tied to a new multi-billion AI fund and data center projects across multiple countries. The wager is straightforward: demand for compute keeps climbing.
Why ByteDance is planning to spend about $14 billion on Nvidia chips in 2026
Reports link ByteDance to a potential Nvidia chip spend of around 100 billion yuan (about $14.3B) in 2026. The figure is described as conditional, depending on what hardware can be sold into China. Even with that caveat, it reflects how major AI players can become enormous sources of chip demand. That demand can shape pricing, strain supply, and decide who gets capacity first.
Whether Google is training generative AI on your Google Photos albums
A claim circulated that Google’s image-generation advantage comes from scanning users’ Google Photos libraries, but the publicly shown evidence in that back-and-forth appears thin. The reporting notes Google’s stated position that it doesn’t train generative AI models outside of Photos using your personal data. At the same time, Google Photos does analyze images for features and safety. The gap people point to is proof of model-training use, not basic analysis.
Why Google AI Overviews got the EV tax credit wrong - and how to avoid bad info
The example described involved Google AI Overviews stating people could still get a $7,500 EV tax credit even though the credit had ended. That kind of error is especially risky because it sounds confident while being outdated. A common approach is to treat AI summaries as a starting point, then verify with official government guidance and current eligibility pages. For money-related decisions, double-checking dates and rules is essential.
What BigBear.ai buying Ask Sage means for “secure” government AI use
BigBear.ai says it finalized a $250M cash acquisition of Ask Sage, aiming at “secure, regulated genAI” for government and other controlled environments. Ask Sage is described as serving large numbers of users across many government teams, and the combined pitch emphasizes governance-first deployment. In many procurement settings, reliability, compliance, and auditability matter more than flashy demos. This deal signals continued consolidation around regulated AI products and services.