Google Collaborates with Blackstone to Establish AI Cloud Firm, Threatening NVIDIA and CoreWeave
Alina Collins
Alphabet and Blackstone are forming an AI cloud company targeting 500 MW of compute by 2027, powered mainly by Google's own TPU chips — a direct challenge to Nvidia's data-center dominance and CoreWeave's head start.
What exactly is this new venture?
Alphabet and private-equity giant Blackstone are setting up a joint AI cloud company, targeting 500 megawatts of compute capacity before 2027.
The critical detail: the venture will run primarily on Google's in-house Tensor Processing Units (TPUs), not Nvidia GPUs.
This means → Google is pushing its custom chips from an internal tool to an externally sold service — a commercial ambition upgrade.
Why should Nvidia be nervous?
Nvidia currently dominates the data-center market. Its chip designs account for 86% of data-center revenue.
Latest earnings: Nvidia posted $81.6 billion in Q1 FY2027 revenue, up 85% year-over-year, with non-GAAP EPS of $1.87, up 140%.
In plain terms = Nvidia's business has never been better, but Google, Apple, Amazon, and Microsoft are all building custom chips — the biggest customers are becoming competitors.
Where does Alphabet's $185 billion capex go?
Alphabet's capital expenditure plan for this year totals $185 billion, mostly earmarked for data-center construction.
As TPU deployment ramps up, part of that spend will likely shift away from Nvidia processors toward in-house chips.
This reflects a structural trend: tech giants no longer want to just buy someone else's chips — they want to build, use, and resell their own.
How long can CoreWeave's first-mover advantage hold?
CoreWeave currently has a head start, and Nvidia recently invested $2 billion in the company.
But Bernstein analyst Madison Rezaei warns that Google's entry could become "more aggressive competition," squeezing CoreWeave's pricing power and margins on new contracts.
Put simply = Google doesn't need to win the whole market. Just offering an alternative gives customers leverage at the negotiating table — CoreWeave's bargaining position gets diluted.
High debt meets a heavyweight rival — where's the risk?
CoreWeave plans to scale AI compute to over 5 gigawatts by 2030, taking on significant debt to do so.
The AI cloud market is projected to reach roughly $400 billion by 2031 — a large enough pie, but a major new player just showed up.
This means → For a company expanding on borrowed money, compressed pricing power is the most dangerous signal — revenue expectations fall, but the debt doesn't.
Google enters AI cloud — are Nvidia and CoreWeave really in danger?
BULL
The moat still holds
Nvidia accounts for 86% of data-center revenue; ecosystem lock-in is hard to break short-term.
TPU is unproven externally
Google's custom chips have been mostly internal; commercial deployment is still early-stage.
BEAR
Customers turning competitors
Google, Apple, Amazon, and Microsoft are all building custom chips — the biggest buyers are de-risking away from Nvidia.
CoreWeave's leverage problem
When pricing power shrinks, highly leveraged companies get hurt the most.
In plain terms = Nvidia's near-term numbers are untouchable, but its biggest customers are all building alternatives — this isn't a one-quarter story, it's the industry structure shifting.
Content is for reference only, not financial advice.