Semantic Coherence: GamersNexus – Signal Evidence & AI Readability

[H5] GN Mega Charts: CPU Benchmarks & Comparison
CPUs Features
[H5] GN Mega Charts: CPU Cooler Benchmarks & Comparisons
Features Coolers
[H5] GN Mega Charts: CPU Power Consumption
Features CPUs
[IMG: Crimson Desert GPU Benchmarks]
[H5] Crimson Desert GPU Benchmarks, Bugs, Simulation Error Tests, & Intel Arc Gets Screwed
GPUs Game Benchmarks & Graphics Guides
[IMG: AMD RX 9070 GRE Yeston Review]
[H5] China-Only AMD RX 9070 GRE Yeston Waifu: Thermals, Gaming, Noise, & Benchmarks
GPUs
[IMG: AI GPU Black Market]
[H5] THE NVIDIA AI GPU BLACK MARKET | Investigating Smuggling, Corruption, & Governments
GPUs Deep Dive News
[IMG: GPU Pricing Update]
[H5] GPU Prices Crater Before Inevitable Opportunity to Screw Consumers
GPUs Subscribe to
[IMG: Crimson Desert GPU Benchmarks]
[H5] Crimson Desert GPU Benchmarks, Bugs, Simulation Error Tests, & Intel Arc Gets Screwed
GPUs Game Benchmarks & Graphics Guides
[IMG: Dragon]
[H5] Dragon's Dogma 2 is a Mess: GPU & CPU Benchmarks, Bottlenecks, & Crashes
Game Benchmarks & Graphics Guides
[IMG: City Skylines 2]
[H5] Terrible Optimization: Cities Skylines 2 GPU Benchmarks & Graphics Optimization Guide
Game Benchmarks & Graphics Guides
[H5] Baldur's Gate 3 GPU Benchmarks: The Best Video Cards for Baldur's Gate 3
Game Benchmarks & Graphics Guides

Get fucked, Bambu.3D printer manufacturer Bambu Lab has legally threatened a software developer who built a slicer tool called "OrcaSlicer-BambuLab," which is software used to translate 3D models into printable files. The developer, Pawel Jarczak, has taken down his own software out of an abundance of caution and in order to avoid further legal ramifications from Bambu Lab; however, from what our lawyers have seen and what we think, we believe Pawel is in the right to both make and upload his software. As a result of this, GN is putting its position of strength behind the software alongside our friend Louis Rossmann and will be, with permission from the developer, hosting the files for download. You can download them here:
[H3] DOWNLOAD ORCASLICER FOR BAMBU LAB
OrcaSlicer-BambuLab by Pawel Jarczak (hosted on GN)OrcaSlicer-BambuLab by Pawl Jarczak (hosted on Github by FULU / Louis Rossmann; mirror)We have a full standalone piece we’re working on for GNCA that we’ll publish as a deep dive into Bambu’s anti-consumer actions, but we’ll start with the basics while working on that writeup and research. This is also being published in HW News.
[H4] Credits
[H5] Host, Writing
Steve Burke
[H4] The Basics
While we work on our full piece, here's what you need to know about the situation.The developer, Pawel Jarczak, posted on GitHub to say that Bambu Lab accused him of reverse engineering their software and impersonated Bambu Lab:“In that statement, Bambu Lab described the modification as something that: (1) injected ‘falsified identity metadata’, (2) ‘pretended to be the official Bambu Studio client’, (3) ‘crosses into impersonation’, (4) was ‘bypassing a technical limitation’, (5) could create infrastructure problems comparable to service overload caused by unauthorized traffic. These are very serious public accusations. Bambu Lab did not write to me with these specific public claims first. They also refused my request to publish the full correspondence. Instead, they published a one-sided public statement where I cannot reply directly. In practice, this presents me to the public as someone bypassing security, impersonating their client, and creating a risk to their infrastructure. I reject that characterization.” - Pawel, developer of OrcaSlicer-BambuLabFrom everything we’ve seen, Pawel’s code is a legal third-party tool that is covered under open source software licensing, and in no way has he “impersonated” Bambu Labs or presented as them at any point.We’re not lawyers, but you don’t have to take our word for it.Copyright attorney and YouTuber Leonard French, who is a lawyer and who also covered our Bloomberg situation, jumped in with his own commentary:"The picture, then, looks like this: Bambu sends a Cease & Desist letter to a developer over an open source project, while a legal mechanism is quietly taking shape that would allow any of Bambu's own customers to walk into California State court and demand the corresponding source code for their entire closed networking ecosystem. But you have to zoom out, because the conflict between Bambu and OrcaSlicer is not an isolated event: It is a textbook manifestation of what Right to Repair advocates have started calling 'progressive enclosure,' the strategy by which manufacturers across every sector use software locks to convert one-time hardware sales into ongoing monetizable services. Bambu's enclosure timeline tracks the pattern almost perfectly."In the piece, he goes on to describe a situation that, to us, sounds favorable for the developer and not Bambu.Friend of GN Louis Rossmann also jumped in with a bounty:"I saw Pawel was kind enough to give me an idea of what was going on [...] what I'm going to be doing here is pledging $10,000 for the legal fees that he would have to pay if he keeps up his code and Bambu Labs goes after him. If Bambu Labs goes after you for keeping up your code, I am so confident in your case that I will pay the first $10,000."And we’d like to do the same. If Bambu Labs pursues Pawel for his code, which from what our research suggests is completely within legal bounds, we will also contribute $10,000 to Pawel’s defense against Bambu Labs and would be happy to connect Pawel with our US-based IP attorneys that we used to successfully fight Bloomberg. We'd also be fine putting it towards offense against Bambu.Louis is pursuing this from the angle of fighting Code 1201 of US copyright law and the DMCA, which is the “circumvention of copyright protection systems” code. Bambu appears to be trying to leverage this section of copyright law, which carries with it aggressive penalties and possible federal prison time as a maximum criminal penalty.
[H4] Frustrations for GN, too
We’ve seen some of the communications behind the scenes and it looks like Bambu is trying to claim circumvention of copyright protection systems. Louis is also going to be hosting the OrcaSlicer-BambuLab from Pawel with his permission, so it will still be available even though Pawel isn’t currently hosting it for legal reasons. Speaking with me, Louis said Bambu can sue him if they don’t like it. We will be doing the same. If Bambu doesn’t like it, they can add us to their list of lawsuits or I’d be happy to meet them in Shenzhen.This whole thing is frustrating, because Bambu’s printers are actually good.Late last year, we had just bought our first 3D printers and decided to join the scene. We’re consumers here -- we’re not reviewers of 3D printers, so like you all, we deferred to reviews and researched our purchases. We chose Bambu Lab H2S machines, which we used in our Noctua 3D printed fan content. We’ve been happy with the quality of the prints, but the company is reminding us of NVIDIA: The product quality is good, but the company is run by assholes. As a result of Bambu’s recent actions, we’ve begun conversations with Prusa instead and have already ordered $5,000 of Prusa equipment to start experimenting. We want to make 3D printing a long-term part of GN, but unlike NVIDIA, there are a lot of alternatives to 3D printers that aren’t made by assholes, and we’d rather buy into those ecosystems to support companies that are more open rather than be part of the marketshare capture that a Bambu’s actions are enabling.It’s unfortunate, as we do really like the printers and the quality. We already own them, so we’ll use the ones we have until we phase them out as we evaluate Prusa; however, going forward, we’re already looking to move onto Prusa instead. Hopefully they stay clean. We chose Prusa by recommendation of Wendell, and in talks with Prusa, the company is far more open.We’ll have more on Bambu Lab on the GNCA channel soon.In the meantime, grab your Slicer above.Go fuck yourself, Bambu.

[H4] Table of Contents
AutoTOC Grab a GN Tear-Down Toolkit to support our AD-FREE reviews and IN-DEPTH testing while also getting a high-quality, highly portable 10-piece toolkit that was custom designed for use with video cards for repasting and water block installation. Includes a portable roll bag, hook hangers for pegboards, a storage compartment, and instructional GPU disassembly cards.
[H3] Intro
Turns out, we got access to the buggy mess that is Crimson Desert at the same time as Intel’s Arc GPU team.Editor's note: This was originally published on March 21, 2026 as a video. This content has been adapted to written format for this article and is unchanged from the original publication.
[H4] Credits
[H5] Test Lead, Host, Writing
Steve Burke
[H5] Testing, Writing
Patrick Lathan
[H5] Testing
Mike GaglioneTannen Williams
[H5] Video Editing
Vitalii MakhnovetsTim Phetdara
[H5] Writing, Web Editing
Jimmy ThangCrimson Desert was a nightmare to test on launch day. We started right when the game came out, which included a day 0 patch that had some launch optimizations included in our tests, and we immediately ran into problems: Across both NVIDIA and AMD, and across 3 test systems, we experienced various game freezes, full system shutdowns, crashes to desktop… …and whatever the fuck this is, which was available in both pink and yellow. Fortunately, Intel didn’t have any of these problems with its Arc GPUs. That’s because Pearl Abyss refused to work with Intel throughout its development process, so the Arc GPUs just can’t play the game at all. We’ll get into that, too.Beyond the bugs, draw distance issues and pop-in problems, and general noisiness in the game’s graphics, we did actually manage to run 25 GPUs through testing in a span of just 10 hours. It’s also our first gaming benchmark debut of our recently detailed simulation time error testing methodology. We’re using Crimson Desert to test for issues in game object motion separately from framerate and time in charts like this. We had to balance turning this around fast to help as many people as possible with testing as many different graphics settings, areas of the game, and cards as possible, while also trying to balance with our new game tests. We didn’t have early access, so we tested the launch version of the game and its first patch. We ended up going with 25 GPUs, including the GTX 1080 Ti, 1070, and 1060. We ran 3 resolutions, every settings option on a couple of GPUs, and tested across several hours of gameplay, including some distant towns and environments to make sure we understood if the game’s performance changed later. This is in addition to our new simulation time error charts, which we’re excited to debut in the first real-world use case.For testing, we're using our GPU reviews test bench, which has a 9800X3D (read our review) that’s slightly overclocked just to maintain a level clock without any fluctuations. And then we started testing the game as soon as it came out. And saw stuff like this:
[H3] A Buggy Mess -- Sometimes
As we got into Crimson Desert testing, we ran into a ton of bugs that seemed to be from the game rather than from drivers. These issues occurred on both AMD and NVIDIA.There was one driver issue, though: Intel’s GPUs just don’t work. Crimson Desert rejected the B580 (read our review) when detected and told us that it can’t be used to play the game. We asked Intel for its driver support plans. The company informed GamersNexus that Pearl Abyss did not work with or enable Intel to work on drivers in advance of the game’s launch, meaning that they got access at the same time as the public did.Intel stated:“We’re aware that Crimson Desert currently doesn’t launch on systems with Intel GPUs and we’re hugely disappointed that players using Intel graphics hardware can’t jump in to the world of Pywel at launch.Getting games running smoothly is always a partnership between developers and hardware makers. Over the past several years, we’ve reached out to Pearl Abyss many times to help test, validate, and optimize support for Intel graphics, providing early hardware, drivers, and engineering resources across multiple generations, including Alchemist, Battlemage, Meteor Lake, and Lunar Lake.Our teams are deeply committed to helping all studios deliver the best experience possible, providing open tools, documentation, and direct engineering support to make sure their games run well for everyone, including the tens of millions of players using Intel GPUs. We remain ready to assist Pearl Abyss however we can.For details on the choice not to enable Intel support at launch, please reach out directly to Pearl Abyss.”This is about as direct as you can get from Intel. We asked Intel if its between-the-lines comment here is that Pearl Abyss didn’t grant Intel access before launch. Intel’s representative stated: “Correct, Pearl Abyss did not provide early access to game code[s].”Intel is really getting screwed and boxed out here, especially if they're providing hardware and drivers to Pearl Abyss. The fact that, based on what Intel has written here, several media publications had access to this game before Intel did is bizarre. Either way, Intel is probably working on this now, but it sounds like they basically got started the same time we did, which was launch day. Anyways, that's the news on the Intel stuff. We also had issues with hard crashes, and not just to the desktop. We had two machines that experienced a complete shutdown of the system from Crimson Desert. One was on a 1080 Ti and the other was on an RX 7800 XT. It’s possible that there was some other issue on the 1080 Ti system, and so we assumed it was unrelated; however, it then happened again on the 7800 XT (read our review) card and in a completely different computer that hasn’t ever had problems of this sort before.We also had regular issues with all moving objects being rendered with pink boxes around them, resembling hitboxes or something similar. We occasionally saw these with yellow boxes as well. They’d randomly appear and disappear, and we saw these on both AMD and NVIDIA.On a few instances, we had the game lock-up and freeze, though we could kill it with task manager.And that’s just what we found in the test session. The game seems half-baked and not ready for launch, as there should never be any hard shutdowns or game freezing issues.One thing we noticed immediately (and which is just a game issue) was that loading in the game is painfully slow in some situations. We regularly had to sit through a minute to a minute-and-a-half of loading animation and this is on machines with reasonably fast SSDs as well.
[H4] Pop-In
We won’t get too into this, but we also noticed that pop-in is extremely bad in the game. You can see small elements, like grass and plants, popping in just 5 feet in front of the character. At that point, it’s more distracting to have the objects than not. The same is true for objects in the background and mid-ground, where even the highest-end of systems at hundreds of frames per second aren’t given the ability to extend draw distance. We weren’t able to find a way to do this in a configuration file from a quick look, but this seems like a likely mod in the future.The game also just has a lot of shimmery noise in general.
[H4] Scaling in Game Areas (5060 Ti)
This next section goes over our research for designating a test area candidate. The point of this is to find something that’s representative of play without being difficult to replicate, like combat. For this process, we played the game for a couple hours each and wandered around, including to some outer areas near Demeniss, the Serkis Estate, and near the first town of Hernand. We manually capture these and assign names to each, with the point being that we want to learn and understand how the game performs in different situations before committing to a test scenario.Here’s the result with the 5060 Ti 16GB (read our review), which we used specifically because it’s a moderate performer. This is at 1440p. Like most of our testing, we do not use upscaling of any kind unless noted.The range is huge for the lows, where some situations resulted in heavy hitching and stuttering; primarily, we noticed these during what seemed like invisible loading sequences when changing from one outdoor cell to another, such as crossing the bridge on the horse. Lows dropped to 12.5 FPS during this brief sequence, leading to temporary, brief stuttering when at 1440p Ultra without RT enabled. This is in spite of a good average at nearly 70 FPS.We also observed this in our “Inner City Roaming 3” test sequence, where we walked from the lowest part of the Hernand castle outer town to the castle gates. We saw a drop to 21.6 FPS for the lows, despite a great average framerate. The 1% lows also suffer here, at 40 FPS. The real problem is the disparity between the average and lows, with the gap being noticeable.Most of the rest of the results are relatively consistent. The 1% lows are still more distant from the average than we see in some other games, but the 0.1% and 1% lows are mostly near each other, which is good. The range for average FPS here was nearly 28 FPS AVG, from 77 FPS in the fields to 49.8 FPS in one of the fights. That’s a large range.The average of all scenes measured, which spans a few hours of gameplay, was 67 FPS. Our final test candidate was in a city, which averaged 70 FPS. This is higher than the overall average, but it’s the most consistent run-to-run test we could find. Areas that are lower than this in any meaningful way often involved combat, which is not repeatable run-to-run.
[H4] Frametimes - Bridge Draw Distance (5060 Ti)
Here’s a frametime plot for the bridge scene, where we have long-range visibility. Everything looks fine if you truncate the scale to 50ms, with frame-to-frame consistency overall good. That framerate itself is OK, but it’s really the consistency of the pacing that we care about. The problem emerges at the end, where we encountered a massive spike.The chart now shows the worst frametime spike we’ve plotted in years, jumping up to 691 ms. That means we were staring at the same frame for nearly an entire second without any kind of movement -- enough for you to think the game’s going to crash. It recovered, but it was a hard stutter. This doesn’t repeat in the same spot every time, but it does happen just in general during play on some GPUs. The 5060 Ti was holding a 13ms-15ms frametime on average here, which is greater than the 60 FPS that’d plot at 16.667ms. It’s not like the framerate was bad, but this spike definitely was, and this wasn’t a one-off.
[H4] Scaling in Game Areas (9070 XT)
Here’s the same in-game scaling chart for the 9070 XT (read our review) now.The range for average FPS is about 39 FPS AVG, from 115 FPS to about 77 FPS. The 0.1% lows range from 99 FPS to 19 FPS, or about 80 FPS 0.1% low range.The 9070 XT experienced the worst performance during a loading screen to the campfire, but fortunately, that’s not an area where you need responsive controls. In terms of actual play, we had some hitches when walking to the cabin, but otherwise, this card was much more stable for its frame interval pacing at 1440p than the weaker 5060 Ti. The average for all test areas was 96 FPS. The test candidate that we approved for its reliability and consistency averaged 100 FPS, so it’s about 4 FPS higher than the global average, which we think is a representative match while still being repeatable (unlike combat, which is not repeatable run-to-run).
[H4] Settings Scaling (5060 Ti)
Our next section was also part of our early research. For this, we went through and tested various presets in a fixed test path. Unless otherwise noted, we did most of this without modifying any graphics presets and while keeping upscalers. The purpose of these quick tests is to understand what our scaling is and help quickly choose a preset for baseline testing.We tested from Minimum to Max presets, plus RT toggled on and off, with Max including Cinematic mode, Maximum lighting, ray reconstruction, and DLAA. DLAA was not used for any other settings.Minimum looks like a fuzzy, blurry, muddy mess and is hardly playable, but we still tested it. Cards that are otherwise insufficient can technically run this, but it’s not a good experience. This is how they can technically accommodate the 1060 and similar cards.Here’s the chart with the 5060 Ti with FPS values.The 5060 Ti ran at 151 FPS AVG with 1440p/Minimum, which looks awful. That’s an improvement on the low setting’s 91 FPS AVG of 65.7%. This is where you would gain the most performance, but again, it’s up to you if you want to deal with these visual drawbacks. Personally, we wouldn’t play this game on its lowest settings. It just looks that bad to us, but if you can deal with it, more power to you.Medium ran at 74 FPS AVG here, so Low is better by 24%, but obviously looks much worse. High has a performance cost about the same as medium, with the two indistinguishable in performance in this test.Ultra without RT consistently ran within a few frames per second average faster than Ultra with RT. That’s why we chose to disable RT for most of our tests: It has almost no impact on the framerate performance and we wanted to maintain maximum compatibility for older cards, like the GTX 1080 Ti, by disabling RT. Since RT on vs. off performed about the same on both AMD and NVIDIA, we toggled it off to better accommodate cards that don’t support hardware RT.Cinematic is the next large drop in performance, down to 62 FPS AVG. Ultra performs about 12% better with RT on. Enabling “max” lighting, which is not part of any preset, drops Cinematic massively: It falls from 62 FPS to 37 FPS average.Finally, Max lighting, cinematic presets, and RT reconstruction with DLAA ran at 18.4 FPS. This is on a 5060 Ti, remember. We also ran this on a 5090 (read our review) and it was playable.
[H4] Settings Scaling (9070 XT)
Here’s the same for the 9070 XT, although we didn’t run minimum for this one.Top-to-bottom, there’s a total range of 38 FPS AVG in these results in a like-for-like test. That’s about the same range as we saw game-wide. Medium and high don’t differ much from each other, and High and Ultra are also close together. Toggling RT with Ultra showed about a 3.6% uplift with it off, so not enough to warrant leaving it on for this benchmark process as it’d eliminate the 1080 Ti for a like-for-like comparison that we wanted to include.
[H4] Explainer: Animation Error
Grab a GN15 Large Anti-Static Modmat to celebrate our 15th Anniversary and for a high-quality PC building work surface. The Modmat features useful PC building diagrams and is anti-static conductive. Purchases directly fund our work! (or consider a direct donation or a Patreon contribution!)Up next is simulation time error, AKA animation error, which is a mismatch between the pace at which frames are shown and the event

[H4] Table of Contents
AutoTOC Visit our Patreon page to contribute a few dollars toward this website's operation (or consider a direct donation or buying something from our GN Store!) Additionally, when you purchase through links to retailers on our site, we may earn a small affiliate commission.
[H3] Intro
We’re multiple administrations deep in a technological cold war over processing power between the United States and China. China’s Cyberspace Administration has labeled some US graphics processing products as a security risk, seeking answers about US government backdoors in the silicon. Meanwhile, the United States has imposed heavy restrictions on exports of graphics processing units, or GPUs, being sold to Chinese companies by American companies. The sale requires rarely-granted licenses for each import scenario to legally export GPUs above a certain performance level, with the stated objective being to restrict progress of private and government projects, including AI development, in China, while trying to maintain the US’ claimed AI leadership. The United States takes this so seriously that, just this week, the Department of Justice had two Chinese nationals arrested in California for what the DOJ alleges is the smuggling of tens of millions of dollars of GPUs.But where there’s prohibition, there’s smuggling.Editor's note: This was originally published on August 17, 2025 as a video. This content has been adapted to written format for this article and is unchanged from the original publication. This particular story had a saga of what we felt was censorship behind it, thanks to Bloomberg L.P., which we've detailed extensively: Part 1 (Our Channel Could be Deleted) and Part 2 (YouTube's Systematic Punishment). This was written to be seen as a video, so the adaptation sticks to a more viewable/colloquial use of language.
[H4] Credits
[H5] Host, Writing, Lead Editing
Steve Burke
[H5] Editing
Vitalii MakhnovetsTim Phetdara
[H5] Editing, Graphics
Andrew Coleman
[H5] Camera
Tannen Williams
[H5] Research and Writing
Ben Benson
[H3] Acquiring "Illegal" GPUs in China
In China, it's not an illegal market -- it's just a market.We spoke to everyone about this NVIDIA AI GPU black market: We found middlemen who connect buyers and suppliers, users who can understand the demand and explain it, using the most dystopian definition of wealth -- how many GPUs one has.We also found independent repair shops who, simply doing their jobs, salvage valuable silicon components from dead PCBs of banned GPUs, innovatively hand-modifying them to be better than and have more VRAM than NVIDIA’s own official product SKUs. These shops are not explicitly a part of any "black market," they're just repair shops that happen to sometimes work on export-controlled GPUs.We met multiple people who, when asked the same question, gave the same passcode-like Chinese idiomatic expression, or chengyu (成语), which translates to “open one eye, close one eye.” In other words, it means, “to turn a blind eye.” We heard this saying so much that we made a T-shirt based on it to help fund this investigative report."There's a new kind of black market, and it's high-end AI GPUs. This particular black market is worth billions of dollars a year"And among others, we even spoke with a US-based Chinese national buying video cards to strip them and ship the GPUs to Chinese companies, which violates US export control law.One of our viewers was able to meet with a GPU smuggler, whom we’ll call “The Plug.” The smuggler spoke limited English, but they both understood one universal truth: money. He operated a GPU testing rig inside of a Prius that he drives around multiple states in the Western United States. The least suspicious thing in his car was a spare license plate in the trunk, but we’ll come back to “The Plug” towards the end."Black market" is normally a phrase associated with drugs or guns, but there's a new kind of black market: high-end, AI GPUs. This particular black market is worth billions of dollars a year and is hiding in plain sight.Hong Kong is our first stop along our journey. By skyscraper count, Hong Kong would be the tallest city on earth. The density is unbelievable. We spent a few days here for this story, wandering markets and meeting sources.We went to Hong Kong to learn what the demand drivers are for the banned GPUs, how they get into China, and more about the illicit side of the GPU smuggling business.
[H3] Big Adventure
To provide a big overview of our big adventure, it began when we booked a 24-hour plane ticket to Hong Kong and 20 days of hotels across Hong Kong, Shenzhen, Zhengzhou, Huizhou, Taipei, and more. At least one of us got detained by at least one of the governments involved in this story, but we can't talk about these indecipherably singular or plural instances or instance of any aforementioned detainment.Then we talked to a lawyer about the previous sentence, then investigated whether or not GPUs are actually smuggled with lobsters into Hong Kong.We eventually found a pile of GPUs that are now export controlled and met a professor who really wants to make sure you know that they were “legally obtained.”After leaving Hong Kong, we booked a boat to Shekou, China. Then headed to Shenzhen, got kicked out of a warehouse, and used high-speed rail to go deep into China, where we met a guy who thinks that desoldering a GPU and reballing it is no big deal. We got a ton of information about GPU smuggling from a guy named “5.” After that, we went back to Shenzhen for the third time in two weeks, went back to Hong Kong, flew to Taiwan, before finally getting back on the plane to the US.
[H3] Perspective
We shot over 12 hours of interviewsWe spoke to a lot of people on the record for this story and learned about the dystopian world of high-tech GPU smuggling. We spoke to people ranging from owner-operator trading companies to professors of economics building datacenters. We even tried to talk to the US Department of Commerce, but they didn’t reply, and every person on the chain from the US Department of State had out of office auto responders because we coincidentally emailed them the same week of an especially problematic story that enigmatically involved namedropping. We wanted to connect as many pieces of this puzzle together as we can today, and that’s why we shot over 12 hours of interviews that we just spent weeks cutting down. The point is finding people who know people, and each person in our lineup today led us to at least one other person in this video, eventually building the full pipeline of smuggler-to-user. Although this story isn’t about drugs, it is about a different thing that billionaire executives get a high from: AI. But before we get to any of those interviews, we need to establish the basics of this geopolitical mess.The story is complicated, so we'll start with defining these key facts: Why these GPUs are banned, the new 15% license (which only applies to two GPU models), who buying and selling is legal or illegal for, and then the timeline. We have over 400 pages of research that went into finding sources and understanding the laws.
[H3] Why They’re Banned
AI GPUs have been in the mainstream news constantly.This story has been a complete mess to follow. It has spanned years and two US administrations. We have over 400 pages of research that went into finding sources and understanding the laws. Reports warned of NVIDIA product use in nuclear weapons researchHere’s why governments care about AI GPUs:NVIDIA functionally holds a GPU monopoly in our sector of the industry, which is building computers to play video games. That now feels relatively innocent by comparison to AI. The company leveraged decades of gaming domination to build a foundation for what it now focuses on, which is making the most powerful GPUs for AI in the world. Reports warned of NVIDIA product use in nuclear weapons research, facial recognition technology allegedly used in Russia to suppress dissent and growing concerns of AI facial recognition use in the US for similar deployments, alongside reports of use in international spying and in drone warfare. NVIDIA finds itself in the middle of all of this. Even though NVIDIA disputes selling to some of these entities -- for example, it says that it doesn’t sell GPUs to Russia -- the products still find their way there. NVIDIA is making money one way or the other. Someone is buying it, maybe from someone else who bought it from someone else, and it may be transacted through smuggling. But NVIDIA does end up selling the device ultimately to somebody. Regardless of who they sell to, NVIDIA plays a big part in this worldwide obsession of AI.And we think it’s playing all sides, but we’ll talk about that more at the end. Besides, when there’s a gold rush, it’s better to sell the pickaxe than swing it.Although the US doesn’t talk too much about its own use of AI, it spends a lot of time talking about China’s.The US restricts NVIDIA’s GPUs through export control rules that ban the sale of certain GPUs into China. The restriction is for the sale of GPUs by American companies or companies that want to transact business in America to companies that are in China or the Chinese government itself. Some examples of export-controlled GPUs include gaming GPUs like the RTX 5090 and 4090, which are useful in AI applications (mostly for their high VRAM capacity), and data center/AI GPUs like the A100, H100, H200, and B100, as well as the others shown in the image above. This list is constantly in flux. There are some new and incoming exceptions for the NVIDIA H20 specifically, which has faced Schrodinger’s GPU ban depending on whether CEO Jensen Huang had a one-million-dollar dinner at Mar-a-Lago with Donald Trump on a given week.
[H4] GPU Export Control Timeline
Now, we’ll get into the timeline of GPU export controls across multiple administrations. We have a separate article with the full, bulleted timeline here.
[H5] Obama Administration and Early AI Talk
Even at the end of the Obama Administration in 2016, the US Government was just starting to talk about AI in relation to national security in a Wired interview. Then-President Barack Obama said:“Developing international norms, rules, protocols, verification mechanisms around cyber security generally and AI, in particular, is in its infancy. You got a lot of non-state actors who are the biggest players. Part of the problem is that identifying who's doing what is much more difficult. If you're building a bunch of ICBMs, we see them. If somebody's sitting at a keyboard, we don't. And so, we've begun this conversation. A lot of the conversation right now is not at the level of dealing with real sophisticated AI but has more to do with essentially states establishing norms about how they use their cyber capability. Who are you more afraid of: big brother and the state or the guy who's trying to empty out your bank account? Part of the reason that's so difficult is that if we're going to police this wild west, whether it's the internet, or AI, or any of these other areas, then by definition, the government's got to have capabilities. If it's got capabilities, then they're subject to abuse. And, at a time when there's been a lot of mistrust built up about government, that makes it difficult.”
[H5] First Trump Administration and Biden Administration
Those were the early days.the Biden Administration took major action in 2022 by restricting exports to China, Hong Kong, and MacauThe first Trump Administration started a commission advising Congress on maintaining AI leadership, including simply banning the sale of some advanced semiconductor equipment and chips to China. Years of back-and-forth, a pandemic, an election, and Chat GPT’s launch later, then the Biden Administration took major action in 2022 by restricting exports to China, Hong Kong, and Macau. NVIDIA’s Ampere architecture A100 and newer codename “Hopper” H100 GPUs and systems were all restricted. NVIDIA shed tears for $400 million worth of lost sales as a result and was especially sad when many of its export-compliant alternatives to these also got banned, like its newly-created A800, H800, and L40S, in addition to NVIDIA’s RTX 4090 gaming card. NVIDIA said it didn’t expect “near-term meaningful impact” on its financials.
[H5] Second Trump Administration
NVIDIA responded by designing a workaround to the previously worked-around workaround, leading to the H20. Then the government added new rules for high memory bandwidth cards and the Biden admin tried to come up with an AI chip diffusion rule that would limit the quantity of GPUs being sold into different countries rather than only by processing power metrics, because the government really didn’t know how the fuck to measure these things and NVIDIA, knowing more about GPUs, could tweak any dial it wanted to just barely be compliant. Then DeepSeek came out and everyone panicked, stocks plummeted, and the government scrutinized the role of NVIDIA GPUs in it. In February 2025, Fiscal Year 2025 results were posted and NVIDIA’s Singapore revenue skyrocketed to 18% of total revenue based on customer billing location despite shipments to Singapore being claimed to be less than 2% of Fiscal Year 2025 revenue, which caused people to say “wait a minute.” Unrelated: Several GPU smugglers were arrested in Singapore one day after the fiscal year report was posted, which caused people to say “that makes more sense.”In May, 2025, Trump implemented wide-sweeping tariffs and rescinded the Biden chip diffusion rule that would have limited how many AI GPUs Jensen could sell to other countriesAMD spawned out of nowhere to say it wrote-down $800MM of inventory due to export controls. NVIDIA one-upped it with a write-off of $5.5 billion.In May, 2025, Trump implemented wide-sweeping tariffs and rescinded the Biden chip diffusion rule that would have limited how many AI GPUs Jensen could sell to other countries, Jensen then said, “It’s just an incredible vision. I think this is going to be a transformative idea for the next century for us. These 2 initiatives are completely visionary and it’s going to be transformative for America.”The H20 GPU that was created to comply with rules was still okay, then Jensen had dinner with Trump at Mar-a-Lago for $1,000,000, then the H20 got banned -- he must have chosen a bad restaurant. In July, Jensen Huang met with Trump and was permitted to sell H20s again, Huang went to China, China said Jensen’s GPUs have tracking devices and backdoors, NVIDIA denied that, Jensen went to Washington, Trump then spoke very highly of Huang (and Lisa Su), and more importantly, he ragged on Intel’s CEO. Intel’s CEO then went to Washington, so Trump likes him now. Tim Cook materialized from the infernal plane to give Trump a 24-karat gold “gift,” and “gift” is in quotes because that’s not what that’s called. Then we get to this past week, when Trump asked NVIDIA and AM