The AI private market is moving into a different phase.
It is no longer just about startups raising bigger rounds at higher valuations. Increasingly, the biggest players are looking for ways to own the infrastructure, talent, distribution and technology around AI.
That is what makes Nvidia’s reported agreement to acquire Hugging Face for $12.9 billion particularly interesting.
The deal has not been officially confirmed by either company, but if completed, it would mark a major shift in how Nvidia is building its position across the AI ecosystem.
At the same time, Anthropic is reportedly preparing for an IPO that could allow some existing shareholders to sell shares directly in the offering, while the company has also explored acquiring AI chip startup MatX.
Put those developments together with the continued activity around companies such as SpaceX, OpenAI and a growing group of robotics and AI infrastructure startups, and one thing becomes clear:
Private-market liquidity and strategic acquisitions are becoming just as important as fundraising.
Nvidia wants more than the chips
Nvidia has built its AI dominance around one of the most important pieces of the stack: computing.
But the reported Hugging Face acquisition would take Nvidia much further into the software and distribution side of AI.
Hugging Face is widely known as a repository for open-source AI models, datasets and benchmarks. It has become an important meeting point for developers, researchers and companies building with open models.
That makes the reported $12.9 billion acquisition price particularly notable.
Hugging Face last raised money at a $4.5 billion valuation in 2023. The Financial Times had also reported that Nvidia previously considered a $500 million investment in 2025, which would have valued Hugging Face at around $7 billion.
If the reported acquisition goes through at $12.9 billion, it would represent a significant jump from those previous private-market marks.
But the bigger story may not be the valuation increase.
It is what Nvidia is buying.
Rather than acquiring another frontier AI model company, Nvidia would reportedly be buying a platform that sits underneath a large part of the open AI ecosystem.
That could give Nvidia greater exposure to the models and developers that actually use AI infrastructure.
From licensing deals to outright acquisitions
There is another reason the reported Hugging Face deal stands out.
Nvidia has recently pursued several transactions that did not involve buying entire companies.
The company reportedly paid around $20 billion for Groq technology and key leaders in December. It then reportedly entered into a roughly $7 billion licensing and investment deal with Poolside.
Those structures are different from a conventional acquisition.
With a full acquisition, there is a clear headline transaction value and a change in ownership.
With licensing, investment or talent-focused arrangements, the economics can be much more complicated.
For private-market investors, that distinction matters.
An acquisition can create a much clearer liquidity event for existing shareholders, although the actual amount shareholders receive still depends on factors such as dilution, preferences and the structure of the transaction.
That is why the Hugging Face story is worth watching beyond the headline $12.9 billion figure.
Why Hugging Face is strategically different
Hugging Face is not simply another AI model developer.
Its importance comes from its position in the ecosystem.
It provides a platform where developers can discover, share and work with open AI models and datasets.
That makes it closer to an infrastructure and distribution layer than a direct competitor to a frontier AI lab.
For Nvidia, that could be strategically valuable.
Nvidia already benefits when companies buy more GPUs to train and run AI models. Owning a major platform used by developers building those models could give the company another connection to the AI application layer.
The reported deal therefore reflects a broader trend in AI:
The battle is increasingly about controlling the ecosystem around the models, not just building the models themselves.
Anthropic is approaching the public markets differently
While Nvidia is reportedly pursuing an acquisition, Anthropic appears to be preparing for a very different kind of liquidity event.
The company has reportedly been working on an IPO structure that could allow existing shareholders to sell some of their stock as part of the offering.
Anthropic has also reportedly considered lockup periods longer than the traditional 180 days for some shareholders.
The company filed confidentially for an IPO on June 1, and its prospectus is reportedly expected to become public after Labor Day.
A reported valuation of around $1.5 trillion has been discussed by bankers, although the final valuation and offering structure are not yet set.
The important point is that the IPO could potentially give some early investors and employees liquidity at the time of the listing rather than forcing all existing shareholders to wait for lockups to expire.
Why secondary shares matter
This is an important concept for anyone following private markets.
When a company goes public, the shares being sold can come from two broad sources.
Primary shares are newly issued by the company. The money goes onto the company’s balance sheet.
Secondary shares are existing shares sold by founders, employees or early investors. The money goes to those shareholders instead.
An IPO can contain a combination of both.
That distinction has become especially relevant after the recent SpaceX listing.
SpaceX reportedly sold newly issued shares in its IPO, meaning existing shareholders had to wait for lockup periods to expire before selling.
As those restrictions expired in waves, more shares entered the public market.
SpaceX subsequently traded below its $135 IPO price during August, while its shares were roughly 42% below the post-IPO high cited in the report.
The lesson for private investors is simple:
An IPO does not automatically mean immediate liquidity for every shareholder.
The structure of the offering and the lockup schedule can have a major impact on when shareholders can actually sell.
Anthropic’s MatX talks reveal another strategy
Anthropic is also reportedly thinking about something else that could have major implications for the AI industry: building more of its own hardware capabilities.
Reuters reported that Anthropic had discussed acquiring MatX, an AI chip startup founded by former Google TPU engineers, for roughly $7 billion.
Those talks are reportedly no longer active and have shifted toward a possible partnership.
MatX is now seeking new funding at a valuation of around $4 billion, according to the report.
These two figures should not be treated as directly comparable.
A $7 billion acquisition would represent a transaction for control of the company.
A $4 billion financing would value a minority investment.
Still, the reported discussions are revealing.
Anthropic has reportedly been hiring chip designers and engineers as it looks to develop custom hardware and reduce its dependence on Nvidia processors.
So while Nvidia is reportedly expanding deeper into the AI software ecosystem through Hugging Face, Anthropic is exploring ways to move deeper into the hardware side.
The AI infrastructure battle is spreading across the entire stack.
The private market is becoming more interconnected
These developments also show how closely the private and public markets are now connected.
SpaceX’s post-IPO trading has influenced private-market pricing.
OpenAI recently completed a $7 billion employee tender at an $852 billion valuation, creating a liquidity window before a potential IPO.
Anthropic is reportedly considering secondary sales alongside its IPO.
Nvidia is making increasingly significant investments in private companies.
These are not isolated events.
They show companies and investors experimenting with different ways to create liquidity while maintaining control and managing the supply of shares.
For private-market investors, that creates an important distinction between valuation and liquidity.
A company can have a very high valuation on paper while shareholders still have limited ability to sell.
AI funding is spreading beyond model companies
The latest funding activity also shows that investor interest is broadening.
Several recent deals highlight where capital is flowing:
- XPeng Robotics raised more than $900 million at a post-money valuation above $6.3 billion.
- Generalist raised a $200 million extension at a reported $3 billion valuation.
- Gatik raised $200 million after expanding its autonomous trucking business.
- REGENT raised $240 million through a combination of equity and debt.
- Socure raised $156 million at a $5.2 billion valuation and acquired agentic AI startup Fravity.
- Stability AI raised $76 million in a Series B, bringing its total funding to $232 million.
The common thread is that AI investment is becoming much broader.
Capital is moving into robotics, autonomous vehicles, AI infrastructure, identity, chips and AI applications, not just foundation models.
That matters because the next phase of the AI market may be less concentrated around a handful of model companies.
Nvidia’s private-market footprint is getting bigger
One number in Nvidia’s latest filing puts the scale into perspective.
Nvidia reported $47.9 billion in holdings in private companies as of late July.
It had also committed another $18 billion to equity investments through the remainder of its fiscal year.
That is a remarkable private-market footprint for a company best known as a public semiconductor giant.
If the reported $12.9 billion Hugging Face acquisition is completed, it would add another major transaction to that strategy.
Nvidia is increasingly becoming more than a supplier to the AI ecosystem.
It is also becoming an investor and potential owner across different layers of the industry.
What this means for private-market investors
For investors watching the private market, the biggest takeaway is not simply that AI valuations are rising.
It is that liquidity pathways are becoming increasingly important.
There are now several different routes for shareholders to eventually realize value:
- Acquisition, where another company buys the business
- IPO, where the company enters public markets
- Secondary sales, where existing shareholders sell their positions
- Tender offers, which create liquidity for selected shareholders before an IPO
- Strategic investments, which can provide capital without changing control
Each structure can produce a very different outcome for shareholders.
That means looking only at a company’s latest valuation does not tell the whole story.
The next question should be:
Who can actually sell, when can they sell, and what kind of transaction could create that liquidity?
The bigger picture
The AI market is entering a phase where ownership, infrastructure and liquidity are becoming increasingly intertwined.
Nvidia is reportedly looking to own a major piece of the open-source AI ecosystem through Hugging Face.
Anthropic is preparing for a potential public listing while exploring ways to manage shareholder liquidity.
The company has also explored acquiring chip capabilities through MatX.
SpaceX’s experience shows what can happen when large volumes of locked-up shares begin entering the public market.
OpenAI is creating liquidity for employees before a potential IPO.
And investors continue to put billions into robotics, autonomous driving and other AI-related businesses.
For the private market, this creates a much more interesting environment than simply tracking the next funding round.
The real question is increasingly how these companies move from private valuations to actual liquidity events.
And as the AI industry matures, those pathways could become just as important as the headline valuations themselves.