Why Nvidia invests in AI startups: The hidden strategy behind its AI empire

Most investors don't know this about Nvidia (NVDA).

It's been one of the main investors in AI startups like OpenAI and Anthropic. It invested $40 billion in those two alone.

Nvidia supports emerging AI companies so they keep building their systems on Nvidia's hardware and software.

It's basically funding itself.

As you'll see, this hidden strategy has been very profitable for Nvidia.

Nvidia invested in all the big AI startups.

Nvidia invested in the following AI startups over the years:

• OpenAI,

• Anthropic,

• xAI,

• Perplexity AI,

• Runway Cohere,

• Inflection AI,

• Adept,

• Mistral,

• and several other smaller AI companies.

At the same time, Nvidia strengthened ties with the cloud providers that host those models, including Microsoft Azure, Google Cloud, and Amazon Web Services.

Nvidia didn’t need to predict which AI lab would win.

If multiple labs succeeded, even better. The demand for Nvidia’s chips would grow across the entire AI industry.

Even a sizable venture investment of $100 million to $500 million is tiny compared with what these AI labs eventually spend on computing power.

If backing an AI startups helps steer it toward building a 30,000-GPU cluster, and each GPU costs around $35,000, that’s more than $1 billion in potential hardware demand.

So in many cases, Nvidia can treat these venture investments almost like customer acquisition costs for billion dollar infrastructure deals.

And that spending doesn’t stop once the customer stops training the model.

CUDA locks developers into Nvidia for good.

Once the hardware cluster is deployed, developers build their software on CUDA. It's Nvidia’s proprietary programming platform for GPU computing.

And because it's free, it's the logical first choice for AI startups.

CUDA has been under development for nearly two decades and now supports more than 4 million developers worldwide.

The platform includes thousands of optimized AI libraries, frameworks, and development tools.

This creates an important economic effect.

Once AI systems are built around CUDA, switching to a different chip architecture becomes extremely difficult.

Software must be rewritten. Performance often drops during migration. Development slows.

For companies racing to release new AI models, those switching costs are enormous.

The result is a powerful ecosystem lock in.

The best thing is Nvidia doesn't need to fund AI startups forever to earn money from them.

Many AI companies it helped get off the ground aren’t small startups anymore.

They’re turning into infrastructure companies in their own right.

Take OpenAI. It’s reportedly raising money at valuations close to $700 billion. Anthropic recently raised around $30 billion at valuations near $380 billion.

At that size, these companies don’t need much help funding their growth. They can raise huge rounds, sign long-term cloud deals, and build massive computing clusters on their own.

Which is why Jensen Huang said Nvidia will stop investing in OpenAI and Anthropi altogether.

Their ecosystems are already running on Nvidia hardware and software. The moat Nvidia set out to build is already in place.

Nvidia can keep collecting money from OpenAI and Anthropic, despite not funding them anymore.

And there's almost no risk of them switching to a different supplier at this point. The cost and the risk to fall behind in the AI race are just too great.

Hats off to Jensen for this brilliant strategy.

Nvidia (NVDA) has a Disruption Score of 4. Click here to learn how we calculate the Disruption Score.  

Nvidia is also part of the Disruption Aristocrats, our quarterly list of the world’s top disruptive stocks.