Nvidia GTC 2025: Five (or six) developments you need to know about

Welcome to AI Collision 💥,

In today’s collision between AI and our world:

• Five takeaways from GTC 2025, maybe six, or seven (it’s a lot)
• AI factories
• The Derp Knight

If that’s enough to get the pirates sailing, read on…

AI Collision 💥

No one competes with Hopper. I’m yet to see any competition for Hopper.

That’s what Jensen Huang said this week about Nvidia’s tech and its competitors.

Hopper are the H-series chips that Nvidia has been selling since they announced them in 2022.

These are the AI chips (notably the H100s) that Meta, xAI, Tesla, Google, Microsoft, and many, many more have been buying…at least until the GB200 chips get delivered. These are the chips the US put a ban on selling into China, that’s how good they are.

And now Nvidia is selling and shipping, GB200s, and the recently announced Vera Rubin series of AI chips.

While there’s still no competition for Hopper (according to Huang at least) NVidia is pressing on another two generations ahead.

So, yeah, apologies if we sound like Nvidia fanboys from time to time, but the gap from Nvidia to everyone else in the development of AI chips is astonishing. The gap is like a Formula 1 car to my 3-year-old’s balance bike.

Okay, maybe not that big, but it’s big.

There’s a little video at the end of today’s essay where you can hear Huang speak again yourself, but aside from the gigantic leaps they’re taking there was a lot of other stuff Nvidia announced, explained, and dropped into the market this week, so let’s take a look and see where they’re heading next.

AI Platforms and Reasoning Models

They’re now talking about a move away from “generative AI” towards “reasoning AI.” in other words, AI that’s capable of complex decision-making and planning. To get there, Nvidia introduced the open source “Dynamo AI” inference software. They also launched the “Llama Nemotron” advanced reasoning models built upon Meta’s LLaMA architecture.

What I’m loving about the current state of AI development is that it’s all open source. I’ve long held the view that the fastest way to accelerate development in any given are of technology is to open source the information. I think that’s a significant contributor to the pace at which AI is developing now.

Anyway, amongst these things, Nvidia also started to change the narrative around datacentres, and pretty soon we’ll all just be calling them…AI FACTORIES.

Data Centres and AI Factories

Central to Nvidia’s strategy for their AI factories are the GB200 AI chips and the server racks they’re building with these. But they also unveiled the GB300 NV-Link 72—a powerhouse rack system integrating 72 Blackwell GPUs with 36 Grace CPUs, described as an “AI factory in a rack.”

Further to this manic roadmap, Nvidia teased upcoming AI chips, “Vera Rubin” for 2026 and “Feynman” for 2028.

It’s almost getting silly now. And they’re running out of brilliant mathematicians…

Then they showed us the DGX Spark and DGX Station personal AI supercomputers which make my newly build PC soon to be redundant.

AWS, Microsoft Azure, Dell, Lenovo, and Cisco, just some of the cloud providers that have committed to deploying these newer “Blackwell Ultra” systems.

And just like selling the unbelievable Oxi-Clean…but wait, there’s more!

Autonomous Vehicles and Robotics

They announced a major collaboration with General Motors. GM will adopt Nvidia’s Omniverse and Cosmos platforms to create digital twins for their factories , next-generation vehicles, and robot automation. Future GM vehicles will also use Nvidia’s Drive AGX supercomputers, so you bet the future of self-driving cars is Nvidia dominated too.

Taking that even further, Nvidia introduced “Halos,” a safety platform integrating AI chips, software, and mind-bending simulations, designed to get self-driving cars to market, faster and safer. Like airbags and seatbelts became industry standard, I’d expect Halos also becomes an industry standard in self-driving safety.

In robotics, Nvidia unveiled Isaac GR00T another open-source AI foundation model. This one of course tailored specifically for humanoid robots.

To demonstrate this, why not just roll out a cute Disney robot?

Quantum Computing (sort of)

Walking back (somewhat) his comments earlier this year about quantum computing, Nvidia established the Nvidia Accelerated Quantum Computing Centre (NVAQC) in Boston.

This facility, equipped with 576 Blackwell GPUs is for QC companies to accelerate quantum research by simulating quantum circuits and integrating classical and quantum computing.

Partnerships with leading quantum firms, such as Quantinuum, QuEra, and IonQ, alongside reputable schools like Harvard and MIT, give NVidia now at least a foothold in QCs.

And the stock price…

For all the insane, incredible, mind-bending announcements this week, the stock price…did nothing really.

That’s just fine, because I still see this as the most valuable company on earth for at least the next decade. At these prices, long term, I think most investors would have to be a little bonkers to not at least be considering having this as a core part of their portfolio.

But each to their own, I guess.

Finally, I’ll leave you with this too. Note his comment about, “not datacentres, but AI factories.” That’s where we’re heading folks.

#AD

The World’s Largest Oil Discovery – on British Soil!

A report given to the House of Commons Environment Audit Committee confirms the world’s largest oil field has been discovered on British soil!

With estimated reserves totalling 511 billion barrels of oil. At current prices, this oilfield is potentially worth £33 TRILLION. To see how to get in on this opportunity,

Capital at risk. Forecasts are not a reliable indicator of future results.

Boomers & Busters 💰

AI and AI-related stocks moving and shaking up the markets this week. (All performance data below over the rolling week).

Boom 📈

• Team Internet Group (LSE:TIG) up 18%
• Hewlett Packard Enterprise (NYSE:HPE) up 10%
• Palantir (NASDAQ:PLTR)

Bust 📉

• BigBear.ai (NYSE:BBAI) down 8%
• Tesla (NASDAQ:TSLA) down 2%
• IBM (NYSE:IBM) down 1%

From the hive mind 🧠

• I’m all for AI being able to more accurately predict weather. But once, just once, I’d like it to tell me the actual temperature not what it feels like.
• Meta chatbot ahora disponible en español. Meta chatbot ora disponibile in italiano. Le méta-chatbot est désormais disponible en français. Meta-Chatbot jetzt auch auf Deutsch verfügbar.
• “Could be game-changing for the government’s Plan for Change…” ha, could be. Or could just be a really innovative company using AI to do even more innovative things.

Artificial Polltelligence 🗳️

Weirdest AI image of the day

The Derp Knight

ChatGPT’s random quote of the day

Thanks for reading, and don’t forget to leave comments and questions below,

Sam Volkering

Editor-in-Chief

AI Collision