Stock Market Trends

Jeff Bezos' $6.2 Billion Project Prometheus: Why the AI Arms Race Boosts Nvidia's Hardware Dominance

Headshot image for David Engle
By David Engle
Published December 9, 2025 |  Updated December 9, 2025
Featured

It has been more than four years since Jeff Bezos handed over Amazon's (AMZN) keys to current CEO Andy Jassy. But now, Bezos is stepping back into a hands-on leadership role with Project Prometheus, a $6.2 billion AI startup focused on industrial workflows and the physical economy.

And its mission is ambitious.

Project Prometheus aims to build a new type of AI that can help design physical items like machines, parts, and even buildings. Prometheus, as it calls itself, is "AI for the physical economy."

Let's take a quick step back to really understand how Prometheus will compare with the AI most of us know, and the implications it could have for public companies in the AI ecosystem like Nvidia (NVDA).

Why Project Prometheus' AI Could Learn Like a Scientist

At some point, you've probably requested help from ChatGPT or asked Gemini or Copilot a question. They answer the question and can be helpful. But generative AI models like these have limitations. Namely, they don't truly understand the actual, physical world. These models are able to scour huge amounts of digital text, find a general consensus from these sources, and provide a usually serviceable answer. In that way, they're not too dissimilar from a traditional search engine.

But Project Prometheus' goal is to create AI that can learn from the physical world as well as the digital. In theory, this AI model could "observe" real-world experiments, "remember" the results, and iterate on new experiments from there. In that way, these models act like a scientist or an engineer... not just a search engine.

If Prometheus succeeds, American innovation and manufacturing could accelerate like never before.

The main challenge Prometheus faces? This advanced level of AI will require a massive amount of power. Well beyond the already-significant amount of power AI workloads consume today. That's great news for investors in AI infrastructure, because demand will continue to grow.

It may be even better news for those who invest in Nvidia (NVDA). The chipmaker behemoth creates and provides the graphics processing units ("GPUs"), software foundation, and networking systems necessary to design and train this next-level physical-economy AI model. That means Nvidia stands to benefit in a big way from Prometheus.

Project Prometheus: The $6.2 Billion Shockwave Fueling Physical AI

When Project Prometheus was revealed in mid-November, we learned that it entered the market backed by $6.2 billion in capital. From day one, Prometheus was one of the best-financed early-stage startups in the world.

And it looks like Prometheus will need every penny of that seed funding. Primarily because the cost structure of physical AI – the systems that interact with the real world by sensing, reasoning, and acting through physical machines like robots and autonomous vehicles – is extremely expensive.

Physical AI requires specialized infrastructure and proprietary data systems (among other technologies). And those costs drive physical AI well beyond the cost of software-only large language models ("LLMs").

For context, let's look at some of those LLM costs. Alphabet (GOOGL) spent around $190 million to train its Gemini Ultra GPT. Today's LLM models may require around $1 billion to train. And future LLM AI models may cost in the tens or even hundreds of billions.

Looking at energy usage, a single 100-megawatt ("MW") hyperscale LLM AI data center can use up to 2,400 megawatt-hours ("MWh") of electricity per day. That's about as much as it takes to power more than 85,000 average American homes a day. Or a city the size of Boulder, Colorado.

What it will take to power future Project Prometheus advanced AI data centers is still unknown. What we can safely predict, however, is that those costs will likely make current LLM hyperscale data-center costs look like pocket change. And that's not even considering the hardware and physical-infrastructure expenses.

Those figures aren't publicly available. But it's not a stretch to surmise that a good chunk of Prometheus' $6.2 billion in funding will be used on infrastructure capital expenditure ("capex"). This may include:

  • Advanced robotic equipment designed to construct industrial complexes and other buildings
  • AI tools that can manufacture spacecraft and vehicles
  • AI-native engineering, which uses AI to design and build physical systems from start to finish
  • Advanced manufacturing components and other types of custom hardware
  • Sensors and cameras that learn and understand from interactions with the physical world

For investors, there's a lot about this new physical AI world to be excited about. But it's important to keep that massive capex in mind as physical AI gains its footing.

The New Billionaire Rivalry: Application-Layer Competition Drives AI Hardware Acceleration

Simmering beneath the surface of physical AI's emergence is fierce competition between Prometheus, OpenAI, and xAI. These three AI titans are grappling at the application layer, where products like developer tools and chatbots are provided to customers.

Call it the Battle of the Billionaires between Bezos, Sam Altman, and Elon Musk. The latter two are the CEOs of OpenAI and xAI, respectively.

This competition stems from the desire to build the best AI models at the application level. And that requires spending. Lots of spending. Because whichever billionaire (or potential trillionaire, in Musk's case) isn't increasing their investment in advanced training, model development, and cloud infrastructure is the one who will fall behind in this AI arms race.

This competitive spending has its benefits, though. It turns the application-layer war into quickly rising demand for high-performance hardware to be used across the entire AI industry. In other words, the demand generated by the competition creates bottlenecks and shortages for advanced AI hardware like specialized GPUs, high-performance memory, and networking equipment.

And Nvidia, as well as other tech companies we follow closely, will be there to supply it.

Why Nvidia Remains the Core Infrastructure Supplier for Advanced AI

Nvidia's stock price is significantly lower than its 52-week high of $212.19 from October 28. But it's still widely considered the gold standard in the AI development and infrastructure industry. Consider:

  • About 90% of the servers built for AI use Nvidia GPUs.
  • Leading AI and cloud businesses depend on Nvidia hardware and architecture for their infrastructure. These include OpenAI and its cloud partners Microsoft Azure, CoreWeave (CRWV), and Oracle Cloud Infrastructure, as well as Google Cloud, Meta AI, xAI, Anthropic, and Amazon's AWS.
  • Nvidia's networking and hardware are used by companies like Dell Technologies (DELL), Cisco Systems (CSCO), Hewlett Packard Enterprise (HPE), Lenovo, Fujitsu, Asus, Gigabyte, Inspur, and several others within their server and storage systems.

Nvidia's fingerprints are all over everything AI-related. And it's easy to explain why.

Nvidia makes high-level products.

Its GPUs are the fastest and most capable of building large and complex systems. Its chips and networking hardware are baked into how virtually all AI models are trained and deployed. And Nvidia's hardware at the infrastructure layer is generally considered the industry standard.

So, it's safe to say that Nvidia is, more or less, the heartbeat of AI infrastructure. And that advanced AI innovators are relying on Nvidia's products to power the next generation of AI – the physical AI that may change how nearly every industry operates.

How Large AI Investments Support Nvidia's Near-Term Demand Outlook

After reaching its 52-week high on October 28, Nvidia stock plummeted to $177 just a month later, despite a very strong Q3 2025 that saw Nvidia report $57 billion in revenue and $32 billion in GAAP net income. Both figures exceeded expectations and marked year-over-year increases of 62% and 65%, respectively.

But overall concerns of an AI bubble have driven several AI and tech stocks (including Nvidia) down in recent weeks. Some analysts believe this is the market and valuation correction the AI sector needs. Jeff Bezos, however, doesn't truly believe in the AI bubble many analysts worry about. He stated in early October at the Italian Tech Week conference that "this is kind of an industrial bubble as opposed to financial bubbles." Bezos continued:

[A bubble] can even be good, because when the dust settles and you see who are the winners, societies benefit from those investors. That is what is going to happen here too. This is real. The benefits to society from AI are going to be gigantic.

That's why Project Prometheus' backers are making a huge financial commitment to advanced AI development. And that development will need a ton of powerful computing components... like the products Nvidia has mastered.

That could fuel a resurgence for stocks like NVDA. And that should benefit the AI and tech companies that continue investing in AI hardware and technology companies like Nvidia. Even as they navigate through a turbulent AI market in the near term.

My colleague Steven Longenecker wrote recently about Nvidia being one of those companies that will survive the turbulence:

Investors who buy Nvidia today might well be overpaying. But they aren't crazy.

Nvidia's stock could drop 20% or more tomorrow, especially if its earnings show any sign of slowing down. But Nvidia will survive as a company.

It's the most valuable in the world... and for good reason.

  • It is by far the No. 1 provider of graphics processing units ("GPUs"), without which AI data centers simply couldn't function.
  • Those GPUs are in exceptional demand. Nvidia boasts a backlog worth at least some $300 billion in future revenue for its cutting-edge chips over the next year.
  • And Nvidia is both insanely profitable... and still growing at an incredible pace.

All further evidence shows that current and future AI investments – and there will be many – will continue to benefit Nvidia and its investors.

The Bottom Line: Quantifying the Infrastructure Capture

Project Prometheus is just the beginning. Big Tech knows the value of advanced, high-performance AI systems that can perform tasks and interact with the world around them. That's why OpenAI, Meta, and Google DeepMind are following Project Prometheus and getting in on physical AI.

It's not hyperbole to suggest that this technology will change life as we know it.

Robots will move from rigid automation to actual autonomy. This will transform factory and warehouse work as advanced AI powers these robots to navigate complex environments and perform a variety of new tasks that require logic.

It will change the medical industry, as surgical robots learn how to perform minor procedures like administering stitches.

Physical AI will soon make autonomous vehicles commonplace on America's roads. Innovations will allow these vehicles to understand their surroundings and make informed decisions on any type of road, in any type of weather, and in many types of situations.

All of this is happening. And all of it requires an expensive and long-term commitment to advanced-AI development. That includes the hardware and infrastructure needed to support these types of AI systems. And that layer, the infrastructure layer, is where investors can find some stability in an otherwise bumpy AI landscape.

To move this technology forward, AI and tech companies will need networking equipment, chips, and data-center systems. That holds true no matter who is leading the AI race at the application level.

Nvidia is a critical player within the AI-infrastructure layer, with its technology used by nearly every tech and AI company to train and deploy advanced AI models... including physical AI.

Regards,

David Engle

Editor's Note: Amazon just slashed 30,000 jobs – the largest layoff in its history – and almost no one's talking about the real reason why.

Former hedge-fund manager Whitney Tilson says it's part of a much bigger shift. One that could reshape how we all work, invest, and build wealth in the years ahead.

He has spent the past decade preparing for this moment... and he just released something that could help everyday Americans get ahead.

Click here to learn more about this market move while there's still time.

Recent Articles
Back to Top