The $2.5B number is just these guys. It could be 10x in total.

Gamers Nexus did a whole deep dive which basically proved that Chinese researchers had access to whatever they wanted.

https://youtu.be/1H3xQaf7BFI?si=ojlxOC7uiPqZxv0N

edit: not sure if this was sarcasm

Did you think the hesitancy of westerners engaging and relying on Chinese labs was due to vibes? There are fundamental cultural differences at play, wether we are comfortable admitting that or not.

Seems like that's a pretty obvious and straightforward power for a state to have. The state has to make foreign and domestic policy decisions, and to be effective that would have to include trade restrictions. Otherwise you could have situations like businessmen profiting by selling weapons to the enemy to kill his own countrymen--and there are sociopaths who'd do that.

> i never voted for the dingbats who control who is called a terrorist, let alone the people scared of china.

So what?

Same with chips, efficiency, speed, etc all depend on good design, and cutting edge factors, if the main reason your chip isn't faster is because of the distance between your L1 cache and your core is far, then having a bigger node process but bigger chip won't make it quicker.

If you could steal all the designs at TSMC, and you had exactly the process that TSMC uses, you could definitely make counterfeits. If you didn't have TSMC's specific process, you could adapt the designs (to Intel or Samsung) with serious but not epic effort. If you couldn't make the processes similar (ie, want to fab on SMIC), you are basically back to RTL, and can look forward to the most expensive and time-consuming part of chip design.

This is nothing like copying a trivial, non-complex item like a car. Copying a modern jet engine is starting to get close (for instance, single-crystal blades), but even they are much simpler. I mention the latter because the largest, most resourced countries in the world have tried and are still trying.

Not investment advice, do you own research. I'm just someone on the Internet.

Bloomberg's tech coverage is not great from what I've seen. Last year they published a video which was intended to investigate GPUs being smuggled into China, but they couldn't get access to a data center so they basically said we don't know if it's true or not. Meanwhile an independent Youtuber with a fraction of the resources actually met and filmed the smugglers and the middlemen brokering the sales between them and the data centers. Bloomberg responded by filing a DMCA takedown of that video.

(Allegedly) just some Bloomberg (alleged) bullshittery, (allegedly) posted to move the market.

But there's a distinct time/value of investment equation with the current AI boom. The jury is at best still out on what that equation is for the goals of capital (it's increasingly looking like there's no moat), but if you're a national government trying to encourage local bleeding edge expertise in new fields like this it's quite a bit more clear.

They have alternatives, like the Tian supercomputer was originally built with Xeon Phi chips that have been replaced with their own domestic alternatives.

A big limitation is getting access to fab slots. Nvidia and Apple are very aggressive about buying up capacity from TSMC, etc, and China's own domestic fabs are improving fast but still not a real match, particularly for volume.

They can copy it. And no, the software moat is not there if someone choose the blatant copy route. They just can't build it in the scale they want yet.

> what if they just use 12nm and create GPUs with much bigger size but comparable performance

Physics do not work this way :/

Remember the 2018 accusations of spy chips implanted in supermicro motherboards that everyone denied so strongly?

SMCI has a pattern of missteps over the years, I would not qualify them as a solid future bet.

(And in case someone asks the question, no that is not a viable long-term strategy one's retirement savings because it's very much speculating and doesn't work AT ALL when the market is volatile or falling as a whole.)

Something similar has been done in many video game console mod chips. IIRC, some of the mod chips manage it on an encrypted bus (which Bloomberg's claims do not require).

Here's one example of a mod chip for the PS1 which sniffs and modifies BIOS code in transit: https://github.com/kalymos/PsNee

"On PsNee, there are two separate mechanisms. One is the classic PS1 trick of watching the subchannel/Q data stream and injecting the SCEx symbols only when the drive is at the right place; the firmware literally tracks the read pattern with a hysteresis counter and then injects the authentication symbols on the fly. You can see the logic that watches the sector/subchannel pattern and then fires inject_SCEX(...) when the trigger condition is met.

PsNee also includes an optional PSone PAL BIOS patch mode which tells the installer to connect to the BIOS chip’s A18 and D2 pins, then waits for a specific A18 activity pattern and briefly drives D2 low for a few microseconds before releasing it back to high-impedance. That is not replacing the BIOS; it is timing a very short intervention onto the ROM data bus during fetch."

Man, Kazakhstan must be an industrial powerhouse by now with all that German machinery. Can't wait for Kazakh EVs and semiconductors to hit the market.

M - Money/Greed

I - Ideology/Divided Loyalty

C - Coercion/Compromise

E - Ego

Sometimes, I think we look at people who are this wealthy and think they should be immune to these kinds of shenanigans, but I'd wager that the -ICE becomes even easier to exploit in people once they no longer need money, if they were already susceptible to it to begin with.

https://youtube.com/watch?v=iEda8_Mvvo4

With processors it's customary to use the "Fan out of 4" metric as a measurement of the critical paths. It's the notional display for a gate with fan out of 4, which is the typical case for moving between latches/registers. Microprocessor critical paths are usually on the scale of ~10 FO4.

The largest chip at the moment is Cerebras's wafer scale accelerator. There the tile is basically at the reticule limit, and they worked with TSMC to develop a method to wire across the gaps between reticules.

you could certainly use a larger process and clone chips at an area and power penalty. but area is the main factor in yield, and talking about power is really talking about "what's the highest clockrate can you can still cool".

so: a clone would work in physics, but it would be slow and hot and expensive (low yield). I think issues like propagation delay would be second- or third-order (the whole point of GPUs is to be latency-tolerant, after all).

Then it’s a matter of how well your engineering/ops org is setup to deal with silly hardware issues and annoyances. Some orgs will burn dozens of hours on a random failure, some will burn an hour or treat the entire server as disposable due to aforementioned cost differences. If you are not built to run on cheaply engineered gear that has lots of “quality of life” sharp edges (including actual physical sharp edges!) then you are gonna have a bad time. Silly things like rack rails sucking will bite you and run up the costs far more than anyone would expect unless you have experience to predict and plan for such things beforehand.

Of course you do have the risk of a totally shit batch or model of server where all that goes out the window. I got particularly burned by some of their high density blade servers, where it was a similar story to yours. Total loss in the 7 figures on that one!

Totally agreed on their BMC/firmware department. Flashbacks to hours of calls with them trying to explain the basics. My favorite story from that group is arguing with them over what a UUID is - they thought it was just a randomly generated string. Worked until one didn’t pass parsing on some obscure deeply buried library and caused mysterious automation failures due to being keyed against chassis UUID… and that’s when they’d actually burn one into firmware in the first place.

It was also always a tradeoff of having to deal with cheaped out hardware engineering with supermicro or with some horrible enterprise quarterly numbers driven sales process with Dell.

It'd be easy to prove the existence of a pervasive "spy-chip" problem using a camera or a microscope. Unsurprisingly, neither Bloomberg nor it's quoted "experts" ever managed to do so, deapite loudly banging that drum.

DeepSeek v3 was trained on 2,048 NVIDIA H800s. https://arxiv.org/abs/2412.19437

MiniMax M1 used 512 H800s. https://arxiv.org/abs/2506.13585

The H800 wasn't banned in the first round of export controls - but was after October 2023: https://www.cnbc.com/2023/10/17/us-bans-export-of-more-ai-ch...

Z.ai say they used Huawei hardware: https://www.theregister.com/2026/01/15/zhipu_glm_image_huawe...

Qwen and Kimi haven't disclosed their hardware as far as I can tell.

Multiple security companies looked into this and found nothing malicious.

If some market has large margins, it means it has some inefficiencies.

Even if you had 'ai tools' guessing at component blocks on evaluation you would have to have some evaluation of the result.

And, thats assuming NVDA hasn't pulled a Masatoshi Shima type play on their designs (i.e. complex traps that could require lots of analysis to determine if they are real or fake)

Im not sure how much of a speedup even modern tooling/workflow could do reliably.

Even then,

The elephant in the room is that China is working on their own AI accelerators/etc, so while there can be benefit from -studying- the existing designs, however I think they do not want to clone regardless.

In fact there is an open gap that I'd expect it to close around $16.30 and another one around $19

And if it were posted to move the market, that would have been about the most cut and dry SEC violation possible, posted at a time when the federal government still enforced such things.

For example: https://www.tomshardware.com/tech-industry/artificial-intell...

His take was that it was very unlikely that it impacted exclusively Supermicro, though.

It was covered various places, including The Register https://www.theregister.com/2018/10/09/bloomberg_super_micro...

And Bloomberg did a DMCA takedown through youtube, copystrike in parlance which pulled the video down for a week. GN had no recourse other than to wait and counterclaim.

Week timed out, Bloomberg did nothing but be the bully.

Louis Rossmann's excellent explainer video here on the Bloomberg bit: https://www.youtube.com/watch?v=6RJvrTC6oTI

You either take a gamble on something and hope it's good, or try to buy the same thing that someone else bought and reviewed.

I don’t think it’s real. Yes, it’s plausible. But first of all, if someone actually surreptitiously put malicious chips onto motherboards en masse, we would have seen a photo of the alleged chip already. And second, there are easier, more effective, and less obvious ways of adding backdoors to networking equipment.

https://www.schneier.com/blog/archives/2018/11/that_bloomber...

From what i recall, the story was very vague, there were no pictures of the specific chip, no pictures of the motherboard of the motherboard that would include serial, i.e. no details that would accompany a serious security research.

As always, Louis is being a bit sensationalist and stretches the truth to whip up outrage. Contrary to what he claims, GN could have easily quoted the president without Bloomberg's video, and that would be fine. "that outlet now has a monopoly on who is able to quote the president" is just a totally false premise. Moreover he tries to argue that GN's video falls under fair use, because it's a 1 minute clip in a 3 hour video. However it's not hard to think of a rebuttal to this. If news organizations can copy each other's clips of official speeches, who would bother going out and making such recordings? Usually how this would be resolved would be by citing precedents, but he doesn't bother citing any.

Gross margin of zero would be mean you sell at exactly the cost to produce. Net margin of zero means you cover all your expenses including COGS. The only really difficult, practically impossible, thing would be doing both at the same time. Though, I could also see a case where you drive down net margins once sunk costs are paid and achieve both.

Doing so practically, or sustainably, in most circumstances would be uhh crazy… but it’s not impossible. Even then I think aiming for zero margin is a pretty credible tactic in eliminating competition if you can out sustain them.

TLDR; Weird? Sure. But not impossible. And even sort of likely if you’re trying to atrophy your competition out of existence.

I thought about quite often while visiting a pub owned by the land lord renting out 150 rooms above. Each floor had a large industrial shared kitchen, shared bathrooms, toilets and a large shared living room. If people had 1-2 guests they would stay in their room, if they had 2-10 guests they would use the shared space, if they had 4-80 guests they would take the elevator to the pub. When one was bored with the guests or didn't have time they were left in the pub. Technically people had bar shifts in their rent contract (that you could buy your way out of) but there were plenty who enjoyed running the bar for free. Drinks were at cost. If you tried to tip or didn't take your change they left it on the counter and it would sit there for a day or two. The problem of the pinball machine earnings they solved with rounds of free drinks and chips.

When asked the owner said exploiting a bar was entirely to much work. If he wanted more money from the people living there he could just increase the rent?

HNers are acting reflexively skeptical (which isn't always a bad thing), but targeted supply chain based attacks conducted by a nation statein the manner described are actually doable, and back when I was still a line-level SWE this was when we started putting significant engineering effort into hardware tampering protections back in the 2015-17 period.

The hardware supply chain incident itself most likely happened in the late 2000s to early 2010s when hardware supply chain security wasn't top of mind as an attack surface.

Modchips targeting contemporaneous gaming systems like the PS1 and PS2 use a similar approach to the SuperMicro incident.

It's still nothing concrete, though. Their CEO basically said that they'd found one and that they couldn't say much more about it due to an NDA.

I thought it was supposed to be an incredibly tiny micro sitting on the bmc's boot flash to break inject vulnerabilities.

Yep. This was why there was a significant movement around mandating Hardware BOMs in both US and EU procurement in the early 2020s.

Also, the time period that the Bloomberg story took place was the late 2000s and early 2010s, when hardware supply chain security was much less mature.

Brother, wait until you learn about the associate press.

In U.S. copyright law, the four factors evaluated to judge fair use are:

1: Purpose and character of the use: including whether the use is commercial or nonprofit educational, and whether it is transformative.

2: Nature of the copyrighted work: for example, whether the work is more factual or more creative.

3: Amount and substantiality used: both how much was taken and whether it was a qualitatively important part of the work.

4: Effect on the market: whether the use harms the potential market for or value of the original work.

Courts weigh all four factors together. There is no fixed rule like "under 30 seconds" or "under 10%." GN's use seems to satisfy all four factors.

The same AP that licenses content to its members and charges non-members for the privilege of reusing their content?

"Many newspapers and broadcasters outside the United States are AP subscribers, paying a fee to use AP material without being contributing members of the cooperative. As part of their cooperative agreement with the AP, most member news organizations grant automatic permission for the AP to distribute their local news reports. "

> GN's use seems to satisfy all four factors.

It's weakest at #1 and #4.

#1: it's a commercial piece of work (so far as I can tell GN isn't a non-profit), and the use of the clip specifically isn't critical to the work. If you're critiquing a movie or something, and need to show a screengrab to get your point across, then that makes sense, but if the purpose of the video is just to establish "Trump said this", the video isn't really needed.

#4: see above regarding making recordings of official speeches.

Moreover I'm not trying to argue that GN is definitely not fair use, only that there's a plausible case otherwise. If there's actual disagreement over it's fair use or not, then the DMCA process is working as intended, and Bloomberg isn't abusing it as Louis implies.

[responses to edited-out portion of parent comment]

Re: #1, GN's work while commercial is an educational investigative journalism / documentary piece which are well established users of Free Use protection. GN's use is absolutely transformative.

#4: Bloomberg would have to prove a financial loss to have standing. That would mean that GN must have no other option than to use Bloomberg's clip, and pay the license, which I don't think would fly. GN would have just produced the segment differently.

readded.

edit: responses

>Re: #1, GN's work while commercial is an educational investigative journalism / documentary piece which are well established users of Free Use protection. GN's use is absolutely transformative.

I'm not going to argue too hard over whether taking a 1 minute clip for a 3 hr video counts as "transformative" because this is getting enough into the legal weeds that you'd want to start citing precedents, rather than having two armchair lawyers duking it out with random arguments.

That said, "investigative journalism / documentary piece" angle seems weak. It's not more "educational" than any other news organization (eg. Bloomberg or The New York Times), but apparently they still go out to record speeches, even though they can supposedly piggy back off another organization's footage under fair use.

>#4: Bloomberg would have to prove a financial loss to have standing. That would mean that GN must have no other option than to use Bloomberg's clip, and pay the license, which I don't think would fly. GN would have just produced the segment differently.

Right, but the purpose of DMCA is to take down infringing works, not to award damages. Whether they have losses or not is irrelevant. Moreover the implied argument of "it might be copyright infringement but Bloomberg isn't losing any money so they shouldn't be able to do takedowns" seems... questionable.

A practical way to think about it is this:

What is the new use for? Courts look first at whether the secondary use serves a different purpose from the original, not just whether it looks different. Uses for criticism, commentary, parody, scholarship, search/indexing, or other new functions often have a stronger transformative argument.

Is there new expression, meaning, or message? That still matters, but after Warhol, a claimed new meaning by itself is usually not enough, especially when the secondary use is being exploited in a similar commercial market as the original. The Court emphasized that the inquiry is tied to the specific use at issue and whether that use has a distinct purpose.

Does it substitute for the original in the same market? Even if the new work has some new meaning, it looks less transformative if it is serving basically the same licensing or audience function as the original. That overlaps with factor 4 as well.

How much was taken, and was that amount justified by the new purpose? A use is more defensible when it takes only what is reasonably needed for the transformative aim. In parody, for example, some copying may be necessary to “conjure up” the original, but not more than needed.

All of which I think can fairly be evaluated in GN's favor. Though as you point out, the lawyers are paid to argue each point.