- Google inked a deal to create a new neocloud company with Blackstone called N1, which will use Google's TPU chips
- It's not clear what kinds of customers the new company will target
- Using TPUs could help N1 offer services at a lower price point and could help Google reap the benefits of increased manufacturing scale
Google is going neocloud. The tech giant just inked a $5 billion deal with private equity firm Blackstone to build a new U.S. based data center business based on its TPU chips, with the goal of bringing 500 MW of capacity online next year. But why, exactly?
In many ways, it feels like the joint venture with Blackstone (dubbed N1) is just a drop in an ever-growing bucket. Google is already one of the leading global cloud providers and is planning to spend up to $190 billion this year to expand its data center footprint.
But even at neocloud scale, the deal is on the small side.
“The neocloud space is kind of beyond this particular sizing,” Moor Insights and Strategy’s Jason Andersen told Fierce. “CoreWeave has already crossed the 1GW line and is forecasting four more [gigawatts] in the next three years. So, a half-gigawatt in the next year isn't likely to shake the competition.”
Is this Blackstone's first foray into the data center market?
This isn’t Blackstone’s first foray into the data center market, either. The firm took major data center operator QTS private back in 2021.
So, what exactly is going on here?
The question is a tricky one to answer, especially given it is not exactly clear who N1’s target customer base will be. As Andersen pointed out, it’s entirely possible that Google itself could end up being N1’s primary customer.
But J.Gold Associates Founder and Principal Analyst Jack Gold offered up two theories.
First, he said Blackstone could be looking to expand its presence in the data center market. Teaming up with Google to get its TPUs is a smart way to do that, because Nvidia chips are hard to come by and the TPUs offer a cheaper alternative with sufficient performance.
That, Gold said, will allow N1 to charge its customers less, which could help it draw in smaller and mid-market enterprises for which using Nvidia chips is cost-prohibitive. TPUs also make more sense than top of the line GPUs in a post-AI training world where inference becomes the dominant workload. That’s because they use less power and are generally more cost effective for inference at scale.
Second, Gold said Google could be doing this to help boost its position in the chip manufacturing line.
“This helps them because it’s a volume commitment,” he explained. “When Google goes to TSMC or whoever else and instead of 1,000, it’s 100,000 or 100 million, that helps them in getting further up in the queue. So, all of that becomes part of the play.”
N1’s 500 MW of capacity will likely take more than a million Google TPUs. Couple that with the fact that Anthropic recently signed up to buy up to 1 million Google TPUs as well, and well, Google is starting to cook.
Check out al of our neocloud coverage here.