Alright, let's get straight to it, because yesterday was one of those days that almost seemed unbelievable.
An American startup was hit by a cyberattack. And when it tried to defend itself using the most advanced American AI models, they couldn't help. So it turned to a Chinese model instead. Yes, you read that correctly.
On the very same day, Nvidia announced a new AI security alliance, invested $5 billion into a startup that doesn't even sell products yet, and is reportedly discussing providing $250 billion in guarantees to support OpenAI.
The company targeted by the attack was Hugging Face. The issue was that leading American frontier AI models include safety guardrails that don't distinguish between an attacker and a defender. As a result, the company ended up using a Chinese open-weight model running on its own servers, without those restrictions.
"So what exactly is an open-weight model?" you might be wondering.
Simply put, these are AI models that you can download, modify, and run on your own hardware. That's very different from closed models like those from OpenAI or Anthropic, where access is limited to the company's own infrastructure.
Shortly after the incident, Nvidia, together with Microsoft, SpaceX, Palantir, and dozens of companies from the United States and Europe, announced the Open Secure AI Alliance. The goal is to develop and openly share AI security tools.
Nvidia summed it up perfectly:
"The recent Hugging Face incident was a clear reminder. Defenders need open, advanced systems to protect themselves."
There is, however, a major catch.
In Washington, policymakers are debating the exact opposite approach. Treasury Secretary Scott Bessent warned that Chinese companies could face sanctions over AI distillation, the process of extracting knowledge from a more capable model. As Chris McGuire of the Council on Foreign Relations put it:
"In Washington, this isn't a debate about open versus closed source. It's a debate about whether we'll tolerate Chinese theft of intellectual property."
The problem?
Most of today's most capable open models come from China.
Now let's move to the second part of the story, which is arguably even more remarkable.
The scale of investment is staggering.
First, Nvidia invested $5 billion in Ilya Sutskever's Safe Superintelligence. Sutskever was one of the key architects behind OpenAI's technology. The deal also includes access to Nvidia's Vera Rubin platform, which is expected to increase the company's computing power by roughly tenfold.
What's especially interesting is that Safe Superintelligence has publicly stated that it doesn't intend to sell products anytime soon, despite being valued at $32 billion last year.
Second, Nvidia is reportedly discussing providing $250 billion in guarantees to support OpenAI's lease of a 10 gigawatt data center in southern Ohio, being developed by a SoftBank subsidiary. The total cost of the project could exceed $500 billion, making it the largest data center ever announced.
Third, Nvidia announced a partnership worth more than $500 billion with South Korea's SK Group to develop AI data centers and next generation memory technologies. It also invested $1 billion for a 4.5% stake in Naver.
As if that weren't enough, Samsung signed a memorandum of understanding with Broadcom covering more than $200 billion worth of memory and semiconductor projects through 2030.
Altogether, South Korea announced $950 billion in new AI initiatives.
Jensen Huang described it this way:
"Ilya has opened fundamental paths in modern AI, beginning with AlexNet."
Sutskever responded:
"We have research that's worth scaling, and access to a massive Nvidia computing platform will make that possible."
With all of that positive news, you would probably expect Wall Street to celebrate.
Instead, Nvidia fell roughly 5%.
The Philadelphia Semiconductor Index declined 3.9%.
Micron lost 2.3%, Sandisk plunged 11%, Western Digital fell 4.2%, AMD dropped nearly 5.2%, despite Wedbush saying its data center revenue is growing faster than expected.
ASML lost 8.4%, although that move was also influenced by reports that a state backed company in Shanghai has begun developing lithography equipment.
So why did the entire sector sell off?
The concern is something investors call circular financing.
The idea is that Nvidia funds startups, those startups buy Nvidia chips, and Nvidia records the resulting revenue. Investors are increasingly asking where the truly independent demand is coming from.
Personally, I think the market is once again overreacting.
We've discussed this before. Investors want proof that these enormous investments will generate meaningful returns, and that's perfectly reasonable.
But Nvidia isn't borrowing money to fund these projects.
It's deploying the enormous free cash flow it already generates.
And consider this.
A company willing to back $250 billion in guarantees while committing to more than $500 billion in strategic partnerships clearly doesn't believe the AI infrastructure buildout is slowing anytime soon.
The same is true for AMD, which recently signed a similar agreement with Anthropic.
This isn't one company acting irrationally.
It's an entire industry making the same long term bet.
