# Nvidia's Board May Discuss Jensen Huang's Future—No Ouster Yet

Canonical URL: https://superpowerdaily.com/videos/bd9c992d-947b-433d-9ca2-dcb594640d35
Published: 2026-11-12T09:00:11.000Z
Episode date: 2026-11-12
Language: English
Duration: 11:08
Publisher: Superpower Daily
YouTube: https://www.youtube.com/watch?v=molzeH4w2js

The WSJ reports Nvidia directors may discuss Jensen Huang’s future, but no dismissal has been decided; meanwhile, DeepMind’s AlphaGenome 2 targets cell-specific genetic effects for research, not clinical use. OpenAI’s reportedly surging Q3 revenue came with a wider adjusted dollar loss, while Google’s Gemini Decisions API scores bounded choices, though its speed claims cover shorter inputs.

## Key takeaways

- Nvidia faces reported board discussions about Jensen Huang’s
- future, but no completed dismissal or final decision has
- been announced.
- AlphaGenome 2 conditions genetic predictions on specificcell

## Chapters

- 0:34 Nvidia leadership: https://superpowerdaily.com/videos/bd9c992d-947b-433d-9ca2-dcb594640d35?t=34
- 2:30 Cell-specific DNA: https://superpowerdaily.com/videos/bd9c992d-947b-433d-9ca2-dcb594640d35?t=150
- 5:08 Growth and losses: https://superpowerdaily.com/videos/bd9c992d-947b-433d-9ca2-dcb594640d35?t=308
- 7:15 Bounded decisions: https://superpowerdaily.com/videos/bd9c992d-947b-433d-9ca2-dcb594640d35?t=435

## Stories covered

- Nvidia board prepares to oust Jensen Huang, WSJ reports; shares slide: https://superpowerdaily.com/posts/nvidia-board-prepares-to-oust-jensen-huang-wsj-reports-shares-slide
  Directors could meet as soon as Wednesday to discuss the co-founder’s future. TechCrunch says Nvidia has no obvious successor.
- Google DeepMind releases DNA AI that predicts genetic effects in specific cells: https://superpowerdaily.com/posts/google-deepmind-releases-dna-ai-that-predicts-genetic-effects-in-specific-cells
  AlphaGenome 2 replaces broad tissue predictions with cell-specific analysis. DeepMind reports stronger immune-cell results, while clinical use remains unvalidated.
- OpenAI’s Q3 revenue reportedly jumps about 50% as adjusted operating loss widens: https://superpowerdaily.com/posts/openai-s-q3-revenue-reportedly-jumps-about-50-as-adjusted-operating-loss-widens
  Business customers now supply most of OpenAI’s revenue, according to Greg Brockman. But improving margins have not yet translated into smaller dollar losses.
- Google launches an AI decision API that scores up to 1,000 questions per request: https://superpowerdaily.com/posts/google-launches-an-ai-decision-api-that-scores-up-to-1-000-questions-per-request
  Gemini Decision 1 evaluates supplied options instead of generating text. Google claims fast results, but its published speed figures cover shorter contexts than the API’s maximum input.

## Transcript

Automatically transcribed from the narration.

[0:00] Welcome to the Superpower Daily Explainer Let's dive straight into our fast paced briefing of today's four most critical AI developments In under 25 seconds here's our roadmap for today First we'll cover the Wall Street Journal's report on a potential ouster of NVIDIA's CEO Jensen Huang Then we'll look at Google DeepMind's Alpha Genome 2 and its shift to highly specific cell targeting Third we'll explore OpenAI's wild paradox of soaring revenues but widening dollar losses And finally we'll

[0:27] break down Google's brand new AI Decision API before wrapping it all up with four essential takeaways Let's get to it NVIDIA's leadership shakeup WSJ reports potential CEO ouster We're starting off today with some unexpected corporate drama right at the very heart of the AI hardware boom So the Wall Street Journal is reporting that NVIDIA's board is getting ready to discuss the future of their chief executive Jensen Huang and the stakes here they're astronomically high I mean

[0:55] Jensen co founded NVIDIA and has led the company for a staggering 33 years Over that time he's built up this incredibly wide management structure with around 60 people reporting directly to him According to TechCrunch he delegates very little But the real kicker is that NVIDIA has no obvious successor or succession plan lined up to manage the company if he were to step down Why does this matter to you Well if you use ChatGPT or Clod right

[1:19] now you're relying on Jensen Huang's chips NVIDIA is the undisputed king of this boom Their hardware powers both the initial training of these massive AI models and the large scale inference required to actually run them when you hit send So a leadership vacuum here that sends literal shockwaves through the entire tech ecosystem And you can clearly see that uncertainty in the immediate market reaction Following the report on Wednesday evening NVIDIA shares took an approximately 8% hit

[1:46] in after hours trading Now TechCrunch also pointed out a roughly 11% decline over the preceding five days Though to be totally fair that five day window includes trading before this board report even emerged So we can't pin all of that drop solely on this specific news Still it absolutely highlights the intense sensitivity surrounding NVIDIA's stability But wait here's an absolutely crucial limitation to keep in mind This is only a reported upcoming board discussion Yes the reporting

[2:14] says directors could meet as soon as Wednesday November 11th but this is not a completed dismissal No final decision has been announced and NVIDIA hasn't even responded to requests for comment yet The immediate question is simply whether these directors actually meet and if they do what they decide DeepMind's targeted DNA AI Alpha Genome 2 shifts to cell specific analysis Moving from hardware to biotech Google DeepMind just dropped Alpha Genome 2 on November 11th and it fundamentally

[2:42] changes how AI predicts genetic effects To really get why this is a big deal we have to look at how the model actually makes its predictions The original Alpha Genome took a million letter DNA sequence and tried to predict molecular properties across thousands of experiments and tissues all at once That's a huge broad view but it easily blurred the really subtle effects that might only show up in one particular type of cell Alpha Genome 2 fixes

[3:08] this It replaces those broad tissue averages by zooming in and conditioning its predictions on one chosen biological condition like a specific tissue cell type or cell state DeepMind trained this new model on a massive dataset of more than 107 million individual cell profiles across 68 tissues Just let that sink in for a second We're talking about mapping the biological behavior of over 100 million distinct cells It takes immense computing power but it gives us a much

[3:36] sharper high definition picture of our biology What makes this incredibly powerful is how it handles brand new unseen data Researchers can just give the model a snapshot of gene activity data from a few dozen cells and a cell encoder compresses that into a guide for its predictions completely without needing to be retrained For example in a demonstration using unseen data from a COVID 19 mRNA vaccine study the model successfully distinguished activated amine T cells from resting

[4:04] ones It then accurately predicted that a specific variant linked to severe COVID 19 disrupts gene activity a lot more strongly in the activated state If you were only looking at the unactivated cells you'd miss that state dependent effect entirely Ultimately this means researchers can spot dangerous immune reactions or subtle genetic vulnerabilities in a virtual environment before they ever happen in a physical lab Now on a 28 immune cell benchmark called 10k10k DeepMind reported that Alpha Genome

[4:32] 2 identified 30% of causal variants That's nearly double the 16% of the original model It also jumped from 23% to 43% in identifying directional gene effects But hard stop there is a major limitation here These are company reported research benchmarks not clinical detection rates DeepMind specifies that Alpha Genome 2 is available for non commercial research with commercial access currently in private preview The important gap to understand is that it is strictly not validated and not approved

[5:00] for any clinical or diagnostic use We're looking at a cutting edge research preview here not a tool your doctor is going to use for a diagnosis tomorrow OpenAI's $10 billion quarter Revenue jumps roughly 50% as losses widen Okay let's pivot to the business side and look at OpenAI's newly reported financials which paint a really paradoxical picture of their massive growth So the main takeaway comes down to the raw numbers According to reports from The Wall Street

[5:27] Journal The Information and Bloomberg OpenAI's third quarter revenue shot up roughly 50% jumping from about $6.7 billion in Q2 to roughly $10 billion in Q3 Because their sales grew so incredibly fast their adjusted operating margin actually improved from negative 150% to negative 109% But here is the wild paradox That better percentage didn't mean they're losing less actual cash The absolute adjusted operating loss still widened going from $10.1 billion in Q2 up to $10.9 billion in Q3

[6:02] And the context makes that number even heavier That massive $10.9 billion loss still exceeds their total quarterly sales Plus remember that this is an adjusted figure meaning it completely excludes an extra $2.6 billion in stock based compensation for the quarter Now tech companies obviously use stock to attract and retain the absolute top tier AI talent they need to build these models But leaving it out of the operating loss definitely makes the financial health look a bit

[6:28] rosier on paper than the reality of what it actually costs to keep the lights on and keep that talent in the building OpenAI president Greg Brockman said enterprise revenue doubled in the quarter and now makes up the majority of the company's sales That's driven in part by companies looking for cheaper alternatives through OpenAI's developer API They also have a staggeringly huge consumer base reportedly 1.2 billion weekly active users and 100 million paying subscribers But honestly it's

[6:57] wild that a company with over a billion active users is bleeding cash this fast The ultimate test here is whether this enterprise growth can eventually accelerate enough to outpace their spiraling operating costs Because right now even with sales going through the roof the actual dollar loss is moving in the wrong direction Google's bounded decisions API Gemini Decision One evaluates options instead of writing text For our final story today Google is completely flipping traditional AI text generation

[7:27] on its head with the launch of their new Gemini Decisions API Normally you ask an AI a question and you sit there waiting for it to write out a long open ended conversational response one token at a time Well this API doesn't do that at all Instead it's designed to instantly score up to 1,000 bounded questions against a single shared piece of context in one processing pass This is a massive paradigm shift for developers You supply

[7:53] the options up front and the model fires back probabilities and confidence scores not paragraphs of text And that shared input can be truly multimodal meaning text audio video photos or scanned PDFs Let's look at how this plays out in the real world Imagine you run a massive office building A robot or a security camera snaps a photo of the break room and there's a note about an upcoming meeting on the table You want the AI to

[8:18] analyze this scene In just one single request step one checks for a spill using a simple yes no evaluation Step two chooses a cleanup action from multiple choice options which can take anywhere from two to 255 choices And step three rates the overall mess using rubric scoring on an ordered scale starting at zero Every single answer comes back with a confidence score between zero and one So your software can automatically handle the clear cut answers and

[8:44] pass only the tricky low confidence cases to another system or to a human reviewer What's truly groundbreaking here is how economically efficient this is The service costs just 10 cents per million input tokens but the real savings come from that shared input design The big piece of shared material like a long call recording or a massive PDF is only billed once per request no matter how many questions you ask against it Each new question only adds

[9:10] its own tiny instructions to the bill and there are absolutely zero output token charges since it's just returning scores But as a friendly warning here is the blind spot to watch out for Google is reporting blazing fast end to end latency of 150 to 250 milliseconds claiming the speed stays nearly flat whether you ask one question or a hundred However that benchmark only covers inputs up to 16,000 tokens The API itself accepts massive inputs of up

[9:37] to 256,000 tokens So the key constraint right now is whether those impressive speed figures will actually hold up as real heavy workloads push the API toward its full maximum input limits The four essential takeaways Today's superpower daily synthesis Okay let's pull all of this together with four essential takeaways you can carry with you today Number one NVIDIA's board is reportedly discussing the future of CEO Jensen Huang after his 33 year run raising some serious questions about

[10:08] stability right at the hardware center of the AI boom Number two DeepMind's AlphaGenome2 is zooming in on specific cell types and states to accurately predict genetic effects bypassing the blur of broad tissue averages for much sharper biological insights Number three OpenAI's third quarter revenue accelerated to roughly $10 billion which improved their operating margin percentage but their absolute dollar loss still widened to a massive $10.9 billion And number four Google just built the Gemini Decisions API specifically

[10:40] to rapidly evaluate multiple choice options against shared context proving that AI doesn't always need to chat to be incredibly useful Which leaves us with this final thought to ponder As AI models become increasingly faster and more specialized is the uncertainty in hardware leadership or the widening losses in software margins the bigger risk to this current boom Read more at superpowerdaily com Stay curious

## Editorial standards

https://superpowerdaily.com/editorial-standards

This episode reports information available at publication time. Consult the linked stories and their reporting sources for context and subsequent updates.