Google is trying to use its cash pile to do what it couldn't do with engineering alone: break Nvidia's stranglehold on the AI chip market.

According to The Wall Street Journal, Google is in talks to invest around $100 million in cloud startup Fluidstack, valuing the company at $7.5 billion. The goal? Get more data centers using Google's homegrown AI chips—called Tensor Processing Units, or TPUs—instead of defaulting to Nvidia's GPUs like everyone else does.

Here's the problem Google is facing: its TPU chips are technically impressive and gaining adoption with companies like Anthropic, but the broader cloud market doesn't care. The biggest buyers of AI chips are Google's competitors—Amazon Web Services, Microsoft Azure—and they're not exactly lining up to buy Google's hardware. Meanwhile, manufacturing bottlenecks and limited interest from cloud rivals have kept TPU adoption in the minor leagues.

So Google's strategy is essentially this: if hyperscalers won't buy our chips, we'll fund the companies that will. Fluidstack is one of a new breed of "neocloud" providers that offer GPU-as-a-service to AI companies. The WSJ reports that Google has also backed crypto-mining converts like Hut 8, Cipher Mining, and TeraWulf, all of which are pivoting to AI data centers.

The Journal also reports that some Google Cloud managers are pushing to spin out the TPU division into a standalone unit. That could unlock outside capital and give Google more flexibility to invest—but it would also be a tacit admission that TPUs can't win inside Google's existing structure.

So what does this mean for investors? If you own Nvidia (NVDA), this isn't exactly a five-alarm fire. Google throwing $100 million at a startup isn't going to dent a company with an AI chip monopoly generating tens of billions in revenue. But it does show that Big Tech is getting serious about ending their dependency on Nvidia—and that could create margin pressure down the line.

For Google (GOOG) shareholders, the question is simpler: is this a smart use of capital, or is Google just throwing money at a problem that can't be solved with checkbooks? The TPU has been around for years. If it was going to win on merit, it would have won by now. Subsidizing adoption feels like the kind of thing you do when you know your product isn't compelling enough on its own.

Bottom line: Nvidia's moat is still intact. But Google's desperation to crack it is worth watching.