@nvidia bought @huggingface for $12.9 billion. Here's my take.
Nvidia’s real threat is that it's biggest customers are becoming its competitors. Google has TPUs, Amazon has Trainium.
You can fight that by building better GPUs, but Nvidia is playing one layer higher and it’s doing it in two ways now.
One: Own the distribution.
Hugging Face doesn’t own model IP now; it owns something arguably more valuable: where developers go to choose models. A million public model repos. The place where developers decide on the model, the quantization, and the inference provider. If you own that decision point, you own the moment right before the compute decision.
Two: Own the supply.
Nvidia is spending heavily on Nemotron, its own open-weight model family.
Reportedly ~$6B on a model north of a trillion parameters, with 100+ poolside engineers pulled onto the project. The chip company is now shipping open models that compete for the same workloads its hardware powers.
Why now, though?
Open weights are increasingly coming from China. China accounts for roughly 41% of Hub downloads over the past year. Meanwhile most US enterprises still won't deploy those models, even hosted on US infrastructure. That gap is real, and Nvidia just exploited it. It supplies the open models, and owns the shelf they ship from.
Short term - probably good for open source.
Long term - intelligence is still controlled by a few companies. We need to see distributed compute, not just open models.
We're sitting on 100K+ in cloud credits from AWS, Google Cloud, and Azure.
We still don't use them to host most of our applications. We deploy on @render, @supabase, and @vercel instead.
Why pay to host what your credits already cover for free? Because the credits are a trap dressed as a discount.
The savings look obvious on paper. But if you deploy on them, someone owns the DevOps and babysits the deploys. You ship slower, so you hire more engineers to move at the speed you already had for free.
That trade is almost never worth it. Not unless you have a DevOps team sitting idle, which most small teams don't.
We only reach for the big clouds when there's heavy GPU and CPU work and inference. Our fine-tuned model lives there. That's the whole list.
If you're a small team, look hard at your stack before you chase credits.
The build vs. buy rule I use now:
Whatever gets us shipping faster wins.
As an engineer, my instinct is always to build.
But AI has changed the equation.
Building is easier than ever. Maintaining what you build still costs time, attention, and mental bandwidth.
We recently moved our website + blog from Framer CMS to our own Git-based CMS.
Why?
Our marketers and PMs already live in Git + Claude Code. With Framer, formatting, editing, uploading, and publishing took longer than writing the content itself.
Now, updates are faster, and blogs take a fraction of the time.
So I don't ask: “Is it cheaper to build?”
I ask: “Which option helps us move faster?”
In the AI era, speed is the moat.
What’s the last thing you bought that you should have built or built that you should have bought?
@Davidjpark96 I found something interesting: Their search traffic has not gone down yet; in fact, it increased slightly YOY. Anyway, sad to see their stock going down, hopefully it recovers.
🎓✨ Attention Students & Researchers!
@paperguideai Writer is your all-in-one writing assistant, now live on Product Hunt!
Generate outlines, handle citations effortlessly, paraphrase accurately, and detect plagiarism – plus multilingual support. ✍️💡
📖 Key Features:
📝 Seamless content generation
📚 1000+ citation styles
🔄 Advanced paraphrasing
🛡️ Plagiarism detection
🔍 AI Research Assistant
🌍 Supports multiple languages
Try it now 👉 https://t.co/frU0w6wzbY
Support here 👉
https://t.co/Z3VQLcJ29d
#PaperguideAIWriter #AIWritingTools #ProductivityRevolution
Revolutionize your research with Paperguide! 🚀📚 Discover the power of AI-driven tools to streamline your workflow. 💻✨ #AIcouldDOthat#ResearchTips
https://t.co/XyQT5EM5tf
https://t.co/ud7vrhyEK7
Paperguide was No 1 product of the day! If there is a sufficient interest, will write a detailed and actionable blog post about it. Just reply with "Interested" in the comments.