Home Creators Posts Import Register Favorites Logout

Content

The next episode of Broken Silicon will see us be joined by Ed Zitron once again! Obviously, we will be discussing the repercussions of the US Midterms results, if the AI Bubble is truly starting to pop, and if so - what this means for 2027 hardware.

We are recording a little later than usual, on Tuesday, so it is likely this will either be an evening release on 11/10, or the morning of 11/11. Therefore you have ~48 hours to submit below!



Check out Ed: https://www.wheresyoured.at/

Last time Ed was on: https://youtu.be/WTRQJvzmJrs

MLID AI Bubble & US Election Analysis: https://youtube.com/live/7qtK0BnBI84

Comments

Jen-Hsun Huang, Our Dear Leader

Will companies respect the voters in USA clearly showing that as companies have started buying non-NVIDIA AI hardware, and not liking it, by redirecting all their spending to us? Surely there is no other valid interpretation on the results.

Jihad ROUANI

Hi Tom and Ed, we finally got Anthropic's financial numbers and frankly they do not look bad at all, about $80B people seem to think is debt when in reality is a one time charge to give its big investors some shares, and the rest of the company looks quite healthy and profitable. Do you expect them to survive on their own now that they are able to stand on their feet? Small correction to Tom, you've said many times we don't need more data centers, we actually do, AI right now is being rationed to us via 5 hour limits and weekly limits, none would be in place if these companies were able to meet demand in abundance, we actually still are compute starved.

Shonery

If the Midterms give Big Tech an easy excuse to finally cut 2027 AI spend and blame Washington instead of bad ROI, what does that actually mean for us? Gamers assume a popping bubble means cheaper GPUs, but if Nvidia and AMD lose those 75% data center margins, won't they just choke supply and price-gouge consumers even harder to cover the losses?

Girthmaxxer

Making the big assumption that Democrats grow a spine and use their power, what elements of the bubble should the next Congress investigate like GPU smuggling around export restrictions, antitrust or memory price fixing? Who would you most want to see be called to testify publicly?

Girthmaxxer

How much do you think the post-midterms red wave on the stock market was motivated by investor fear about Anthropic's likely junk credit rating versus a slowdown in data center projects, some of which only existed on paper?

magneticmanul

How are tech CEOs still so tone deaf when talking publicly about datacenters, after the election made it crystal clear that people of all parts of society hate datacenters? Recent examples Sam Give Me All Your RAM Altman and Lisa Su bragging on TV about AMD being a "datacenter first" company one day after dozens of datacenter bans or curbs were passed across the county? They haven't even tried to manufacture consensus among the public in the way that most industries do in America. And then all the chipmakers had their stock prices crash by about 20% in a single day over the fear of "no more datacenters" and the tech CEOs still have no idea what is going on.

AnotherChris

Hi Ed, is there any chance of a rug pull from someone like google absolutely obliterating the competition on price with there use of custom hardware or a dark horse model that is super cheap and has great quality and gets people close to an unlimited useage plan?

TheParadoxy

I (Nothin) posted a link in reader mail technical, Jen-Hsun Huang has already linked to it (thanks!!!). I just wanted to add that if you're able to get it to Ed Zitron ahead of time and he's game for it, it has a spreadsheet that professor Damodaran used to value Anthropic. The spreadsheet is a bit complicated, so I think Ed Zitron would want to get an understanding of it before the stream. But if he does, I've highlighted that it can produce the result that Anthropic is worth NEGATIVE hundreds of billions of dollars.

TheParadoxy

Follow-up question for Ed, is there any politician that you think would be a good person to lead an antitrust investigation?

FloridaMan

As a software engineer, I do admittedly find AI to be quite useful. For OpenAI to stay in business, how expensive would their models need to be? At what point do you think big companies will move entirely to local AI, if at all?

Girthmaxxer

Times like these make me miss Lina Khan at the FTC when the FTC could briefly do its job and enforced antitrust. Elizabeth Warren is one of the best members of Congress on antitrust issues and tech regulation.

Zane

Hi Tom and Ed, I became a Zitronium addict after the first Broken Silicon with Ed. I saw on Ed's Wikipedia that he once was a writer for a video games magazine. Can you share your history with games and what games, if any, you've been enjoying lately? Thanks Zane

Xavbeat03

As a fan of Ed and Tom, Do either of you think there will be adverse effects to not just AI Datacenters but normal datacenters as well from the backlash? What do you both think of as good limitations to datacenters in general and what kinds of regulation do you think datacenter development should have?

Mark Smith

Tom and Ed, how did SpaceXAI fail to anticipate Grok Bot's popularity, leading it to revise its token allowance for users? Based on what I have observed on social media, it seems some users felt that the rug was pulled out from under them. Keep up the good work, Tom and Ed. Thank you.

Gob

Hi Ed and Tom, Longtime listener of Better Offline and wanted to say thank you for recommending Deadzone Rogue. Really enjoyed it and was wondering if you’ve played the demo for Deadzone Rogue 2? Are there any other games you’ve been playing that you’ve enjoyed lately? What are your thoughts on local AI inference being the future?

Fipa

Out all the AI giants, who has the best chance not to collapse? I'd assume all the AI subscriptions would stop being subsidized

WujoFefer (edited)

Comment edits

2026-11-11 10:00 I have question about Chinese token prices and models. How they can compete with US AI Labs for the fraction of the price? How they can train models and run them for 10-100x less than OpenAI and Anthropic? How is this possible that they can handle massive use of their models? Did they create some technology advantage or it’s something else? Why Mag7 still invest unlimited resources to build Data Centers when Chinese competitors can do 80-90% of work for fraction of the price? I’m really confused because I’m using GLM5.3 hosted in EU and I have comparison to Sonnet 5.5 and Fable and they are very expensive like tens of cents for one simple question when I made around 1000 requests to GLM for less than 5€… WTF?!
2026-11-08 10:56 I have question about Chinese token prices and models. How they can compete with US AI Labs for the fraction of the price? How they can train models and run them for 10-100x less than OpenAI and Anthropic? How is this possible that they can handle massive use of their models? Did they create some technology advantage or it’s something else? Why Mag7 still invest unlimited resources to build Data Centers when Chinese competitors can do 80-90% of work for fraction of the price? I’m really confused because I’m using GLM5.3 hosted in EU and I have comparison to Sonnet 5.5 and Fable and they are very expensive like tens of cents for one simple question when I made around 1000 requests to GLM for less than 5€… WTF?! For more complex tasks this difference is even bigger. I don’t understand this difference and I think distillation is not the only answer for this. There must be few other reasons for this.

I have question about Chinese token prices and models. How they can compete with US AI Labs for the fraction of the price? How they can train models and run them for 10-100x less than OpenAI and Anthropic? How is this possible that they can handle massive use of their models? Did they create some technology advantage or it’s something else? Why Mag7 still invest unlimited resources to build Data Centers when Chinese competitors can do 80-90% of work for fraction of the price? I’m really confused because I’m using GLM5.3 hosted in EU and I have comparison to Sonnet 5.5 and Fable and they are very expensive like tens of cents for one simple question when I made around 1000 requests to GLM for less than 5€… WTF?! For more complex tasks this difference is even bigger. I don’t understand this difference and I think distillation is not the only answer for this. There must be few other reasons for this.

Chris Rijk

Let’s say that the newly elected Democrats subpoena Jensen Huang to testify under oath - what questions should be asked? Do we need new laws about what publicly traded companies are required to disclose to investors and if so what specific data do you think is needed?

WujoFefer

For more complex tasks this difference is even bigger. I don’t understand this difference and I think distillation is not the only answer for this. There must be few reasons for this.

Chris Rijk

Anthropic was originally expected to IPO with a valuation of about $2Tn but now the talk is of a $1Tn valuation. But they’re still committed to spend hundreds of billions of dollars on compute in the next couple of years. Even if everything else goes to plan, isn’t this reduction in their IPO alone going to cause them a lot of headaches?

Chris Rijk

In my day job, even when people are ordered to use more tokens, they rarely do. People are still slow to change and there’s practically new reports every day about AI not bringing the expected return on investment to companies. In short, I’m sceptical that end user demand will continue to rise exponentially. Is there any good data for or against this theory or will we just have to wait and see?

AI Generated Fish

Ed, you have talked about Nvidia channel stuffing and Satya nadella admitted Microsoft has GPUs they can't plug in. I know you've done a lot of work tracking data center builds. Do you have an estimate for how much of Nvidias revenue is just sitting in a warehouse somewhere?

AI Generated Fish

I worked in investments during the 2008 financial crisis. I see a lot more parallels to that than the dot com bubble (markets artificially inflated by sketchy credit and debt). Nvidia lost 77% of their market cap in 2008. Do you think theyll lose more or less market cap in an AI bubble collapse?

AI Generated Fish

Not Ed but keep in mind SEC and department of Justice are both executive agencies so trump. Congress has subpoena powers so they can investigate but they can't actually DO anything.

Jen-Hsun Huang, Our Dear Leader (edited)

Comment edits

2026-11-10 04:00 Ed, I haven’t gotten texts asking me if I want to privately invest in openAI or Anthropic (see: add the link to reader mail question on discord later today) in months. Clearly this means they’re fine, as they don’t need to text random people living in the Bay Area asking for investment!
2026-11-09 00:21 Ed, I haven’t gotten texts asking me if I want to privately invest in openAI or Anthropic (see: https://discord.com/channels/566259563880382476/614176250923319331/1521477706913415208) in months. Clearly this means they’re fine, as they don’t need to text random people living in the Bay Area asking for investment!

Ed, I haven’t gotten texts asking me if I want to privately invest in openAI or Anthropic (see: https://discord.com/channels/566259563880382476/614176250923319331/1521477706913415208) in months. Clearly this means they’re fine, as they don’t need to text random people living in the Bay Area asking for investment!

PCDog

Ed, when the AI bubble fully pops, how many victory laps are you going to take, and how many “I told you so.” are you going to say?

TheParadoxy

Jensen Huang just admitted to having an affair with a Chinese spy. Do you think he actually cheated on his wife, or is this just an excuse for why so many Nvidia GPUs have been going to China?

Frowning Boat 8

In case it's not covered, if the bubble is popping, how quickly could that affect laptop or console pricing, devices with pre installed ram, as opposed to off the shelf ram?

Dangerman1337

Oh he 100% cheated on his wife, big question is if he did anything illegally.

Dangerman1337

Hi Tom and Ed, feels very, very, *very* self congratulatory and probably premature but I feel vindicated about the AI bubble bursting about this time 5 months ago with the previously mentioned 18 year economic cycle (often invovling property) and low and behold it really seems to be truly kicking off. I wonder why so many people just feel in denial, even as we are hearing datacenter cancellations in South East Asia, suppose not long until the US ones start happening.

Dangerman1337

Second question; one potential consequence if Jensen is forced out as CEO is that he's fostered many relationships in the Supply Chain which I think could be partially jeopardized. All those LTAs where pretty much written with "Jensen will be CEO for many more years" and now that assumption is *gone*, what will be the consequences of this?

Dangerman1337

Third and final question; while the US Media conversation about AI has a more skeptical tone in general UK Media conversation is still full of boosterism despite recent evidence showing hitting a wall in token spend especially. I feel astounded how they boosterist talk in the UK ignores the financial constraints especially.

Jensen Popping More Than AI Bubbles Huang 999!

If you were in charge of the hopefully upcoming lawsuits, who do you go after first. Who or what company is the most damning in all of this? OpenAI for evil vilain maxing, maybe nvidia for the circular financing?

Jensen Popping More Than AI Bubbles Huang 999!

As a national security initiative, will the affair draw more eyes to nvidia’s background workings, and how Singapore has a lot of 5090 gamers? Including microcenter selling truckloads of GPU’s to people who apparently share incredibly similar addresses?

Collin Wells

With the chinese getting Jensen to successfully lobby to drop the import restrictions on 5090’s is there any legal trouble for Nvidia. Also is it safe to assume that’s where the 5090 stock went that we couldn’t buy?