Power Limits, Mega Deals, the Rise of Grokipedia, and AI Tools!
This Week in AI Newsletter 11/3/2025
Thanks for tuning into today’s This Week in AI Daily Newsletter!
Today we are covering everything from OpenAI’s $38B AWS partnership and the launch of Grokipedia to the AI musician who just broke into the Billboard Charts.
Also… a standout AI tool worth trying this week: Wispr Flow.
Grokipedia vs Wikipedia
Last week Elon released Grokipedia.com, “an open sourced, comprehensive collection of all knowledge.” This is a direct competitor or Wikipedia, which has an estimated 11 billion page views per month. While Grokipedia is currently lacking with only 885k articles on platform compared to Wikipedia’s 7.1M, its superiority in search and thoroughness is already become apparent. AI like Grokipedia updates information instantly and allows you to use Grok to dive deeper into topics. Learn more here. Watch All-In clip here.
Funding News: Polygraf AI
Based in Austin, Texas, Polygraf AI develops secure, on-premise Small Language Model (SLM) systems that help enterprises and government agencies detect AI-driven risks like deepfakes and data leaks. Their explainable, auditable AI runs locally on minimal hardware—such as standard servers, laptops, or edge devices—providing a more private, transparent, and compliant alternative to cloud-based black-box AI solutions Polygraf AI has raised $9.5 million led by Allegis Capital in seed funding to expand its secure AI solutions for enterprise, defense, and intelligence sectors. More here.
Polygraph was recently selected as Top 10 Cybersecurity Startups at TechCrunch’s 2025 Battlefield 200
The AI Infrastructure Constraint is Power, not GPUs.
The AI infrastructure constraint is not GPU count — it is power. In a widely shared clip from the BG2 podcast, Microsoft CEO Satya Nadella said that “The biggest issue we’re having is not compute, but power…this may lead to chips sitting in inventory.”
Companies like Microsoft and Meta have each acquired millions of NVIDIA GPUs. As Satya noted, the constraint is no longer chip supply — it’s the growing challenge of securing enough power to operate them. More here.
OpenAI x AWS: $38B Strategic Partnership
This morning, OpenAI and AWS (Amazon Web Services) announced a $38B strategic partnership to “run and scale ChatGPT inference, training and agentic AI workloads.” - Andy Jassy
The agreement will continue to grow over the next 7 years and will give OpenAI access to hundreds of thousands of Nvidia GPUs.
This marks OpenAI’s 4th biggest deal to date, behind:
$500B Stargate deal
$100B Nvidia deal
$100B AMD deal
More here
Microsoft x IREN: $9.7B AI Cloud Deal
Microsoft makes a $9.7B, 5-year deal with IREN for AI cloud capacity. The deal gives Microsoft access to compute infrastructure to build with Nvidia’s GB300 GPUs. They will be deployed in Childress, Texas through 2026; with a planned 750 megawatts of capacity.
IREN CEO Daniel Roberts says the deal will generate $1.94B in annualized revenue and take ~10% of the company’s total capacity.
IREN, based in Australia, provides renewable-powered data center and compute infrastructure
More here.
Robots & LLMs
Turns out LLMs alone cannot yet power robots.
Andon Labs tested this by asking a robot vacuum, controlled by an LLM, to complete tasks transporting butter between rooms which could be a general delivery task in a household setting. It’s clear that humans still complete 55% more tasks than their robot counterparts based on their “butter-bench” leaderboard.
Takeaway: The constraint with robot systems currently is not the orchestrator (the robot vacuum), but the orchestrator (the intelligence or LLM).
More here
ChatGPT is NOT Stopping Medical and Legal Advice!
The internet has been talking about how ChatGPT is going to step away from medical, legal, and financial advice. There is some truth in this but, not to the extreme initially claimed. If you read their updated usage policies that simply claim that you can not use ChatGPT to give advice that requires a license, like legal or medical advice without a licensed professional… this is very different than what some were claiming online. More here.
Google Removes Gemma AI Model
Google removed its Gemma AI model from AI Studio following claims that it generated false statements about U.S. Senator Marsha Blackburn.
Gemma is Google’s family of open, lightweight AI models designed for developers to integrate into their own applications.
Blackburn said Gemma falsely claimed she had been accused of sexual misconduct during a 1987 state senate campaign — a claim she called entirely fabricated.
Google said Gemma was never meant for consumer use and that it’s “seen reports of non-developers trying to use Gemma in AI Studio and ask it factual questions.”
The company will remove Gemma from AI Studio but continue offering access via API.
More here..
AI Music: Xania Monet
Xania Monet’s “How Was I Supposed to Know” Debuts at the No. 30 spot on Adult R&B Airplay. Xania Monet is the product of poet Telisha Jones, her song’s arrival on the chart is the first known time an AI-based act has earned a spot on a Billboard radio chart. This prompted artists such as Kehlani to express disgust saying that “This is so beyond out of our control.”
This is not the only instance of trending AI music, as 50 Cent’s “Many Men” AI cover has amassed 2.8 million views on YouTube.
Here are some more AI tracks — Check them out!
Billie Jean — Michael Jackson R&B cover (reminds me of Tory Lanez)
Ms. Jackson — OutKast 1950’s Soul version
My Favorite AI Tool This Week: Wispr Flow
I came across Wispr Flow in a post by its web designer, @hunterthompson. The app’s design and animations are exceptional, and after spending some time with it, I can confidently say it’s become one of my favorite AI tools.
Most people type around 45 words per minute but speak at 220.
Wispr Flow bridges that gap with a clean UI that turns speech into clear, structured text. Removing fillers, fixing grammar, and adapting tone for any app.
I’m still building it into my routine, but when I do, it’s a huge productivity boost, especially for notes and stream-of-thought messaging.
You can download it for free here.
ChatGPT not giving legal advice is a bit a of joke but I think it will be good for law firms because it will both help users and then suggest that they contact a law firm. My law firm has been getting clients after business owners are trying to solve problems in Perplexity and ChatGPT and then the tools suggest law firms to follow up with to actually get legal advice.
Also in AI, Perplexity just added Tasks (not sure if only for Pro users). You have to click on your account and then you'll see it in the menu. I'm having it do a daily comprehensive search on whether there are any publications in the last 24 hours on discussing the impact that AI is having on the law firm business model. I'm going to experiment with what it can do in Comet based on these tasks.
The fact that Satya said they have chips sitting in inventory really drives home why IREN's 750MW Texas capacity is worth $9.7B to Microsoft. AWS dropping $38B on OpenAI and $5.5B on Cipher shows every hyperscaler is scrambling to lock down powered capacity before it disappears. Your framing of power being the new constraint perfectly captures the shift, and IREN's renewabel energy positioning in Texas gives them leverage that traditional data center operators don't have. This is turning into an infrastructure land grab.