Home Creators Posts Import Register Favorites Logout

Downloads

Content

Moira and Adrian gather for a quick reflection on the state of the Jensen Huang/Nvidia scandal, Silicon Valley masculinity and how one man's leather jacket-phase is taking a pickax to your 401K! DISCLAIMER: This story is of course still in flux, if you're listening to this way down the line, please forgive any inaccuracies!

Comments

audioboy

I was somewhat involved in computers in the late 00's (fixing other people's -- didn't have enough to spend on building my own), but I think Adrian's assessment that they were pole position in the GPU market is more accurate. I think Moira's right that Intel was more of a nerd darling at the time, but that was mostly a competition against AMD (which has since flipped the script a bit); Intel wasn't really in the discrete graphics card business, which was Nvidia's ENTIRE business; they weren't strong competitors at the time.

Betti Steam

I keep getting sent memes about it

Karen Cox

Thank you! I’ve spent the last week wondering why one guy’s family problems were costing my family — I AM retired already — so much money. You’ve explained what the Chinese spy was actually doing.

Autumn

Worth noting that Jensen donated money to Trump's ballroom specifically, rather than republican political campaigns. Completely in line with your point that everything goes through Trump personally

Sam D.

It's really not about the infidelity- it's about the national security issues and congress. And while the Board undoubtedly was fine keeping Huang, the company lawyers were looking at months of congressional hearings digging into the potential corruption and the definite national security issues. There was no way they weren't looking at months of Huang being in the headlines in very damning ways, which would inevitably drag the stock value down for an extended period. And possibly more. He doesn't have a majority share in the company (kinda unique structure for tech companies) so when the lawyers went "Yeah, NO", he had to take the L.

Ariel

Sleep deprived, so we’ll see if I can get this down in a coherent manner. One, everyone in the upper echelons absolutely knew that Jensen Huang was having an affair with Alice Huang and (at least) two subordinates. There’s been reporting that he was extremely open about all of these affairs. Quite frankly, I think that also makes it more likely that there was a ton of sexual harassment/inappropriate sexual relations going on at that company, because clearly their culture was terrible. Which brings me to point two, in addition to tanking the stock price and opening the company up to so very many investigations, presumably NVIDIA would like to be able to hire women? So I think that they probably also felt it behooved them to push him out immediately once it came out that he was having sex with a subordinate as well as a possible spy. You know, “As soon as it came to our attention our CEO was abusing his power over his female employees we showed him the door. Please do not investigate further.” Plus now they have the classic sacrificial woman as interim CEO, which might make women feel safer applying. It shouldn’t, but there’s the optics.

Relkia

I think that at the end of this it will turn out that the relationship between Alice Huang and Chinese intelligence is one more of mutual benefit than a directed mission to entrap Jensen Huang, mostly because it would make importing the chips in question from Singapore to China much easier if you have a major state office working the refs for you. Its like how the majority of the defectors during the Cold War were less motivated by ideology and more by being in some kind of financial crisis. There is always the possibility of this being the exception that proves the rule but I wouldn't stake any chips on it.

Relkia

Then again, I have an Nvidia manufactured chip & card in the computer I'm typing this on so maybe *I'm* the spy?

生きてるだけで、りん (edited)

Comment edits

2026-11-16 04:00 yeh. intel never got much of anywhere in the discrete graphics market. was at sc13 (2013 supercomputing conference), and they were showing off a new line of discrete chips for that usecase some people got excited of, but never really went anywhere. intel did have integrated graphics cornered at the time with iris, but that was for "workstations you want to play a 1080p video on without buying a discrete card". nvidia was discrete-gpu dominant (and everyone hated it, because their closed-source drivers sucked and were the most likely culprit to kernel panic your system, and even more annoying to work with than intel's proprietary c compiler). the use of gpus for supercomputing was already commonplace, because plenty of e.g. physics simulation programs, like wrf and openfoam, benefited from an embarrassingly parallel approach (splitting a task into many little tasks that can be run simultaneously, which gpus were good at because they were made to update every pixel on a display at once). what alexnet in 2012 did was show that the old neural net ideas from the 1980s and 90s could be bruteforced now by harnessing that software, and we've been in a cycle of "but what if we did the exact same things, just a little faster and with more data" ever since, with nvidia riding that train of selling the same product +1 to the same customers over and over every year ていうか this is also why llms are hitting a diminishing returns wall. in a brain, signals are passed backwards and forwards between neuron clusters and brain regions (e.g. your eyes talk to your visual cortex, telling it what they see, but your visual cortex also talks to your eyes, telling them how to look based on what you've seen in the past). this sort of architecture is clearly necessary to build a system that can efficiently work with abstracts (like "human" and "chair"), but llms can't do this because they're working with embarrassingly parallel hardware that only does one-direction communication. transformer was a tiny step in the right direction, and so is recent trend of "agents using agents", but, as long as we're building on this kind of hardware, we're going to have obscenely inefficient systems that are inhumanely good at detailing but sometimes fail simple object permanence tasks a child can do. have been a few small steps towards this kind of machine from people like tenstorrent, but the complete industry revolution necessary i think is nowhere near addendum: also yes, tech execs go to conferences like the above and get SO drunk. they rented out the entire denver natural history museum for an afterparty, and i was wandering around chatting as a young grad student with all these bad-wine wobbly intel (and ibm and honeywell and whatever) guys who were "very intently listening with no ulterior motives" to my drunken rambling about like paraceratherium. there was also a mythbusters interactive exhibit 😅
2026-11-15 00:01 yeh. intel never got much of anywhere in the discrete graphics market. was at sc13 (2013 supercomputing conference), and they were showing off a new line of discrete chips for that usecase some people got excited of, but never really went anywhere. intel did have integrated graphics cornered at the time with iris, but that was for "workstations you want to play a 1080p video on without buying a discrete card". nvidia was discrete-gpu dominant (and everyone hated it, because their closed-source drivers sucked and were the most likely culprit to kernel panic your system, and even more annoying to work with than intel's proprietary c compiler). the use of gpus for supercomputing was already commonplace, because plenty of e.g. physics simulation programs, like wrf and openfoam, benefited from an embarrassingly parallel approach (splitting a task into many little tasks that can be run simultaneously, which gpus were good at because they were made to update every pixel on a display at once). what alexnet in 2012 did was show that the old neural net ideas from the 1980s and 90s could be bruteforced now by harnessing that hardware, and we've been in a cycle of "but what if we did the exact same things, just a little faster and with more data" ever since, with nvidia riding that train of selling the same product +1 to the same customers over and over every year ていうか this is also why llms are hitting a diminishing returns wall. in a brain, signals are passed backwards and forwards between neuron clusters and brain regions (e.g. your eyes talk to your visual cortex, telling it what they see, but your visual cortex also talks to your eyes, telling them how to look based on what you've seen in the past). this sort of architecture is clearly necessary to build a system that can efficiently work with abstracts (like "human" and "chair"), but llms can't do this because they're working with embarrassingly parallel hardware that only does one-direction communication. transformer was a tiny step in the right direction, and so is recent trend of "agents using agents", but, as long as we're building on this kind of hardware, we're going to have obscenely inefficient systems that are inhumanely good at detailing but sometimes fail simple object permanence tasks a child can do. have been a few small steps towards this kind of machine from people like tenstorrent, but the complete industry revolution necessary i think is nowhere near addendum: also yes, tech execs go to conferences like the above and get SO drunk. they rented out the entire denver natural history museum for an afterparty, and i was wandering around chatting as a young grad student with all these bad-wine wobbly intel (and ibm and honeywell and whatever) guys who were "very intently listening with no ulterior motives" to my drunken rambling about like paraceratherium. there was also a mythbusters interactive exhibit 😅

yeh. intel never got much of anywhere in the discrete graphics market. was at sc13 (2013 supercomputing conference), and they were showing off a new line of discrete chips for that usecase some people got excited of, but never really went anywhere. intel did have integrated graphics cornered at the time with iris, but that was for "workstations you want to play a 1080p video on without buying a discrete card". nvidia was discrete-gpu dominant (and everyone hated it, because their closed-source drivers sucked and were the most likely culprit to kernel panic your system, and even more annoying to work with than intel's proprietary c compiler). the use of gpus for supercomputing was already commonplace, because plenty of e.g. physics simulation programs, like wrf and openfoam, benefited from an embarrassingly parallel approach (splitting a task into many little tasks that can be run simultaneously, which gpus were good at because they were made to update every pixel on a display at once). what alexnet in 2012 did was show that the old neural net ideas from the 1980s and 90s could be bruteforced now by harnessing that hardware, and we've been in a cycle of "but what if we did the exact same things, just a little faster and with more data" ever since, with nvidia riding that train of selling the same product +1 to the same customers over and over every year ていうか this is also why llms are hitting a diminishing returns wall. in a brain, signals are passed backwards and forwards between neuron clusters and brain regions (e.g. your eyes talk to your visual cortex, telling it what they see, but your visual cortex also talks to your eyes, telling them how to look based on what you've seen in the past). this sort of architecture is clearly necessary to build a system that can efficiently work with abstracts (like "human" and "chair"), but llms can't do this because they're working with embarrassingly parallel hardware that only does one-direction communication. transformer was a tiny step in the right direction, and so is recent trend of "agents using agents", but, as long as we're building on this kind of hardware, we're going to have obscenely inefficient systems that are inhumanely good at detailing but sometimes fail simple object permanence tasks a child can do. have been a few small steps towards this kind of machine from people like tenstorrent, but the complete industry revolution necessary i think is nowhere near addendum: also yes, tech execs go to conferences like the above and get SO drunk. they rented out the entire denver natural history museum for an afterparty, and i was wandering around chatting as a young grad student with all these bad-wine wobbly intel (and ibm and honeywell and whatever) guys who were "very intently listening with no ulterior motives" to my drunken rambling about like paraceratherium. there was also a mythbusters interactive exhibit 😅

Chantelle

One of the things I think separates Huang from the people in the Epstein files is attention. People in the Epstein files were somewhat shielded by sheer numbers, whereas this news broke containment and everyday normal people know about this guy and his behaviour. NVIDIA will now be tied to this scandal for a long time, but for the everyday person, him retiring might obscure that this was known in the tech community.

Jon Smith

There are so many layers to this I don’t think any one can capture this quickly. 1) nvidia really pionered designing systems that transfered complex calculations from cpu that GPU were bettered designed for 2) they used a lot of monopolistic practices to lock in people into the platform. Their main competitor is AMD which bought ATI a graphics card manufacturer. 3) as a part of the 2008 finacial crisis there were laws passed that defaulted retirement money to put into the sp500. This has concentrated wealth into fewer companies. As retirement plans are forced buyers even if the companies are shit. 4) nvidia and AMD made a bunch of money selling GPU for crypto in the mid 2010. 5) nvidia as part of his tatics has bought out more capacity TSMC than they actually can sell the chips. 6) jensen skill over the past few years has been figuring out how to create circular financing to help sell more chips than their customers can actually afford. 7) because of its crazy quick rise in valuation it has become a huge fraction of the SP500 so everyone now has a conflict of interest in Nvidia 8) the craziest part of this to me about this story is how superfurlous the sex is to the motivation for this story. American tech has been trying to get China money forever. There are other nvida partner companiesthat have male executives that have been doing the same thing that alice huang has done, but hasn’t really gotten any traction in the general public. 9) nvida valuation was going to crash because its books are cooked like Enron and independent analyst have been reporting this for years. I can keep going but this story is wrapped up in every layer how broken our society is from finacial markets, gender relations in companies, government corruption, media complicity etc.

Theo

The Intel/NVIDIA competition is really niche. NVIDIA always made graphics chips, and they were always better than Intel’s. However, in the mid 1990s, Intel started building very inferior graphics into their North Bridge chips. At the time, CPUs did not have built-in RAM controllers or video processors, the way those functions are built-in now. So, because Intel built the graphics chip into the North Bridge, for a while Intel was the company that shipped the most graphics chips out of any company in the world. In the late 1990s, early 2000s, NVIDIA tried to invade Intel’s market of North Bridge chips. (I had a computer with an NVIDIA nForce North Bridge at the time. It sucked.) Intel responded with anti-competitive actions, and to be honest AMD also coveted the North Bridge dollars, so now NVIDIA has been locked out of that market. But NVIDIA continues to make the best graphics processors. NVIDIA has also engaged in anti-competitive behavior. Programmers use programming languages to command the hardware to do stuff. When researchers started using GPUs to do AI and other such stuff, at first they used graphics programming languages, because that’s all they had available. NVIDIA recognized the value of controlling the programmers, and they created the CUDA programming platform to trap programmers into writing for NVIDIA chips. (To be fair, CUDA is a better programming environment than what AMD and Intel and Apple produced as a collaboration in response to CUDA.) Programs that are coded for CUDA do not run at all on competitors’ chips.

CkeverName

Well but also have you ever read anything about ideologically motivated spies? People don’t understand or trust their motives. The USSR could not figure out why the Cambridge spies were feeding them so much excellent information, they thought this must be a trap or a double cross. They kept questioning it, they didn’t get that they were just dorks and keeners. Around the time of WWII a German Catholic was a source for the UK & they were so irritated by his need to like pray and think things over. My guess is that defectors with financial problems are less irritating and more understood by whoever they’re defecting to. No one wants to deal with some ultra religious person, ideologue etc. If they’re extreme in one direction today who knows what they’ll switch to or do tomorrow, because a lot of extreme type people switch to the opposite idea. There’s a vegan feminist from about 10 years ago, she was all over YouTube, who became an organ meats promoting TERF in recent years. A bunch of hippies became yuppies. Ideology is pretty malleable shockingly enough. Being broke = clear, understandable, relatable. Less likely to blow up in your face in some ways, IMO

Henry Bish

It’s not just the democrats that changed after these midterms. There were very unhappy republican “China hawks” that were mad about how trump was handling nvidia. I heard it explained that what Jensen really lost this election was the massive hit to trumps ability to offer protection.

Frangipani

Yes, the moment I saw the sacrificial CEO I was thinking, there is more going on and ah, so sorry for you - I am actually going to watch them more closely now cause it will tell us everything we need to know, how they treat her

Forest 🐞

What, mom and dad have work and aren't exclusively mom and dad? inconceivable! Joking aside, it's lovely to hear from you, even if it's in a shorter than usual format 💜

West

An important gender angle on this story not covered much is that Jensen Huang’s (temporary) replacement as CEO is Collete Kress, the company’s CFO. And seemingly another wonderful case of the “Glass Cliff”, where a men heavy executive group brings in a woman to “clean up” the man’s mess. TBC, this isn’t to say she isn’t eminently qualified to run the firm. Just indicative of standards & practices within corporate boys clubs.

West

https://respawnfirst.com/before-resigning-jensen-huang-admitted-he-was-facing-an-avalanche-of-excess-wafers-and-memory-new-report-sheds-more-light-on-the-gpu-ram-crisis/ And 👆🏻 the above issue also points towards the many problems with NVIDIA. It had become “the bank of AI” to use Ed Zitron’s phrasing like GM did for car loans back in the mid 2000s. And due to seemingly overplaying their hand, NVIDIA had an oversupply of hardware and has been strong arming buyers to get shit off its own balance sheets.

Dayin

Also just wanted to mention how I’ve been seeing all these memes about Huang/11 but in a good 70% of them Alice Huang (despite us knowing and having many pictures of her!) is AI-ed to look like a 21 year old K-pop girl. Can we talk about that? It’s egregious. I’ve seen images put on Twitter where they literally keep everyone else in the photo the same but feel the urge to AI her face. Are we literally incapable of accepting an Asian woman as anything other than grok goonbait? Even when she was a literal honey pot spy trap? Help.

Maud McInerney

And the Lutnick connection… please consider this. Lutnick is an Epstein island guy as well as the guy who apparently decided to ignore the China stuff…

Cate Kneale

Jensen Huang is not the new Bond, but Alice might be.

Milo Crothers

I knew my favs would have thoughts on this tech clusterf*ck, but this ep showed me how little I understand the particular world that these people inhabit. Fascinated to get a little inside baseball on the bay area/techy/megarich stuff bc I will admit that is not my bag. Since this story dropped, I have been thinking a lot about the politics of cheating/cheating scandals. It seems like the Huang marriage was maybe closer to that of 18th century monarchs than anything a normal human might consider romance, and there has been a material hit to their stockholders, so this might actually be an outlier in this...... But I've also been following the Ethan Klein debacle and thinking back on similar cheating scandals I've been (parasocial) witness to (Coldplay couple, Ned Fulmer, Adam Levine, Scandoval, etc). To be clear, I think these guys are generally gross, and cheating IS bad, but in a society that can't even agree that rape, DV, genocide, etc. are bad, why is it that we all feel so decisively outraged by these particular incidents? Is cheating really the worst thing we can do in our relationships, or is it just that betrayal, jealousy, scorn, etc. are essentially universal feelings that almost anyone can relate to regardless of if they have been cheated on personally? Has the endless media around cheating led to us feeling so strongly about this? Does a public persona (H3H3, Try Guys) make one uniquely culpable in cheating, almost as if the betrayal is to the followers as well? Is it because we (royal we) bought into their Wife Guy brand and feel as if we've been duped when we find out about their infidelities? Btw, can someone smarter than me explain why the "wife guy" thing gives me such an ick? Also, I feel like we as a culture kind of love a scorned woman (shoutout Slayyyter's new single), is that because we find it relatable? Is it because its a socially acceptable avenue for rage? Do we just love the drama? Is it a more sinister desire to watch a woman be unraveled by the actions of her male partner? Anyway just some of my casual thoughts about recent events, much to ponder in the world of creeps and perverts (I think the preachers might have stayed home for this one lol). -Milo

Jenny

Gentle reminder that "we" the people have very different feelings toward celebrity infidelity. For example, I and most people I know can't care less about whether famous people are cheating on their spouses. Your feelings are yours to examine! From the media though, I do think that these stories are often breathlessly reported because sex/marital scandals are one of very few ways anymore that still seem capable of bringing famous and powerful people down a peg. So, the preacher is here after all - and it's the media, wearing preacher garb, even if it's just a creep/pervert/piles of money underneath the clothes.