A $5 Billion bet by Blackstone says the AI bottleneck is not software, but hardware.

Blackstone and Google on Monday announced a collaboration to create a new US-based entity selling access to Google's custom AI chips (known as Tensor Processing Units) along with data center capacity, cooling, networking and operations. It's basically an AI cloud, one that someone else will run for you so that you dont have to build one on your own.

Blackstone's original equity commitment is $5 billion. Including debt financing, the Wall Street Journal said the total investment could be up to $25 billion. Capacity of the first 500 megawatts is expected to come into operation in 2027. Blackstone, which manages over $1.3 trillion in assets and is the largest data center operator globally, will be a majority owner of the three companies.

Why Google is doing this

Google is building its own Tensor Processing Units for 10 years. These are not generic chips. They are built from the ground up to train and run the kind of AI models that companies want to deploy. They power Gemini. They fuel the AI products Google serves to billions of users. In some workloads, they're competitive with Nvidia's H100s, often significantly more efficient.

The problem has been access. But if you had wanted to score Google TPUs, you were stuck with using Google Cloud. That meant surrendering those to Google's pricing, Google's contracts, and Google's ecosystem. That is a dependency risk concentration most large enterprise customers, especially those in financial services are not willing to accept. You are opting out of the full Google relationship but wanting the chips.

The Blackstone venture solves that. This gives rise to an neutral, independently operated company with Google TPUs that are not part of Google. It looks more akin to traditional infrastructure investment than cloud subscription, and gives customers access to one of the best AI computing platforms.

Why Blackstone is doing this

The business of Blackstone is buying things that throw off predictable cash flows. Toll roads. Office buildings. Warehouses. The company has been steadily encroaching into data centers for the last several years, as a result of exactly that: long contracts, stable tenants and an ongoing need that doesn't stop during a recession.

This was different from your standard data center investment, however. Blackstone will not just own buildings that other companies fill with servers. Part of co-owning the computing product those buildings will sell. You would be paying the customers by dividing this compute time, represented by how much AI work they can run and how long. That is a software-type business on top of a physical infrastructure.

"This is a generational opportunity," Jon Gray, the president of Blackstone, told CNBC in an interview. That is the language of Blackstone, when it thinks it's buying a structural trend, not a cycle.

Where you think AI investment money is going vs where it actually is

By 2026, the five largest US tech companies are expected to spend around $800 billion on AI infrastructure. The majority of that cash is spent on land, power, chips and cooling. Not models. Not software. The physical stuff.

Larry Fink made that argument last week at the Milken Institute conference: compute scarcity is the real AI bottleneck, and concretely will turn into a financial asset class as oil or electricity did.

The Blackstone-Google deal is the first major transaction that appears to resemble the kind of deal that Fink was describing. You can bundle computing power as a financial product, and sell it to enterprise customers that require the services at a scale commensurate with institutional capital returns: $5 billion worth of this proposition.

If they are correct, compute-as-a-service market will be one of the most relevant investment categories in the coming decade. They still have 500 MW of high-end data center grade capacity with Google TPUs inside, and for being early the consolation price isn't too bad at all.

SEC set to release tokenized stock exemption this week as DTCC targets July production trades

The SEC’s tokenized stock exemption could land as early as this week, opening a regulated path for digital versions of publicly listed securities to trade on decentralized crypto platforms. The exemption has been in development since SEC Chair Paul Atkins launched Project Crypto in July 2025 and told the Economic Club of Washington on April 21 that the agency was “on the verge” of releasing it.

• The SEC is preparing to let stocks trade on the blockchain.

• The SEC is set to release its so-called "innovation exemption" for tokenized stocks which will pave the path for trading digital versions of securities, per Bloomberg.

•  NVIDIA CEO Jensen Huang: “Everybody will have to use AI because if you don’t use AI you’re gonna lose your job to somebody who does.”

• Kevin O’Leary says Gen Z is financially cooked when people making $70K a year are spending $28 on lunch

• You can now create more with Claude Design.

Join the Conversation!

We'd love to hear your thoughts and comments. Join our community and stay updated with the latest trends and discussions in crypto.

Twitter | Instagram | Telegram Channel | Linkedin | Facebook