Google and Blackstone Just Fired a Shot at Nvidia — Is the AI Cloud Bubble Starting to Crack?

Alphabet (GOOGL) and Blackstone (BX) announced one of the most important AI infrastructure deals of the year Monday evening, unveiling plans to create a massive new AI cloud company designed to provide compute capacity powered by Google’s proprietary Tensor Processing Units, or TPUs. The announcement immediately sent shockwaves through the AI infrastructure space, pressuring shares of CoreWeave (CRWV) and Nebius Group (NBIS) while also raising new questions about the long-term dominance of NVIDIA (NVDA) within AI compute. The timing is also notable, arriving just ahead of Google I/O on Tuesday and Nvidia’s highly anticipated earnings report Wednesday after the close, two events that could significantly shape sentiment around AI infrastructure demand, competition, and valuation.

Under the terms of the agreement, Google and Blackstone will create a new U.S.-based AI cloud company that will offer data center capacity, networking, operations, and TPU-powered compute-as-a-service offerings. Blackstone is committing an initial $5 billion in equity capital to the venture, which is expected to bring approximately 500 megawatts of compute capacity online by 2027 — roughly enough electricity to power a midsize city. The companies also indicated they intend to scale significantly beyond that initial footprint over time.

The venture is essentially Google’s answer to the rise of neocloud providers such as CoreWeave and Nebius, companies that rent AI compute infrastructure to customers training and running advanced artificial intelligence models. Until now, most of those neocloud providers have depended heavily on Nvidia GPUs, which remain the dominant AI accelerators globally. What makes this deal important is that Google is now aggressively commercializing its own TPU ecosystem externally rather than primarily keeping the technology in-house for Gemini and internal AI products.

For years, investors and industry observers questioned whether Google would eventually monetize its TPUs more aggressively. The chips have been in development for roughly a decade and already power many of Google’s internal AI services, including Gemini. Google has also quietly expanded TPU partnerships with companies like Anthropic and Meta Platforms, but this new venture represents the company’s largest external commercialization effort yet.

The move is significant because the AI compute market is evolving rapidly. Early in the AI boom, companies simply needed access to any available GPUs as demand exploded following the launch of ChatGPT in late 2022. That created massive pricing power for companies like CoreWeave, which effectively became middlemen renting scarce Nvidia compute capacity at enormous premiums. Investors rewarded that scarcity model aggressively, helping fuel huge rallies in neocloud stocks over the past two years.

Now, however, the market appears to be shifting into a more institutional phase. Blackstone brings enormous financial firepower and infrastructure expertise to the table through assets like QTS and AirTrunk, while Google contributes one of the deepest AI technology stacks in the world. Together, they potentially create a vertically integrated AI compute platform capable of competing directly against Nvidia-powered neocloud providers.

This is where the implications for CoreWeave and Nebius become especially important.

CoreWeave’s business model has largely revolved around securing Nvidia GPU allocations and renting them to enterprises and AI developers. Nebius operates similarly. But the Blackstone-Google partnership introduces a competitor that does not need to wait for Nvidia allocations because Google already owns the silicon stack through TPUs. Several investors and analysts immediately interpreted the move as a direct threat to the neocloud model. Shares of CoreWeave fell sharply in premarket trading following the announcement, while Nebius also traded lower as investors worried that AI infrastructure margins could compress if TPU-powered alternatives scale aggressively.

The technical backdrop adds another layer of pressure. Shares of CoreWeave are now testing a major convergence zone around the psychologically important $100 level, which also coincides with both the 50-day and 200-day moving averages. For momentum-driven AI infrastructure stocks, those kinds of technical levels often become critical battlegrounds between bulls and bears. A decisive break below that zone could accelerate concerns that the neocloud trade is losing momentum after one of the strongest runs anywhere in the market.

The implications for Nvidia are more nuanced.

At first glance, the deal appears negative because it introduces another large-scale AI compute ecosystem not fully dependent on Nvidia GPUs. Google has spent years trying to reduce its reliance on Nvidia by developing proprietary silicon, similar to efforts underway at Amazon and other hyperscalers. The venture could eventually divert some inference and training workloads away from Nvidia hardware and toward TPU-based systems.

However, Nvidia’s position likely remains extremely strong for now. Nvidia still dominates the broader AI ecosystem through CUDA software, networking, developer tools, and its enormous installed base. In fact, Google itself still uses Nvidia GPUs throughout portions of its cloud infrastructure. The broader takeaway may be less about Nvidia losing dominance and more about AI compute becoming increasingly fragmented as hyperscalers seek greater control over their own infrastructure economics.

Importantly, this announcement arrives during a fragile moment for AI stocks overall. After a massive rally over the past two years, investors have started reassessing the sector’s elevated valuations and growing execution risks. The AI trade remains incredibly powerful fundamentally, but the market is becoming far more selective. Investors are beginning to differentiate between companies with durable competitive moats and those that may simply have benefited from temporary supply shortages or speculative enthusiasm.

That dynamic is especially relevant heading into Google I/O and Nvidia earnings. Google I/O could provide additional updates on Gemini, TPUs, agentic AI, and inference infrastructure, potentially reinforcing Google’s ambition to become a much larger AI platform player. Nvidia earnings, meanwhile, will likely determine whether investors remain comfortable paying premium valuations across the broader AI infrastructure complex. Any signs of slowing hyperscaler demand, tighter margins, or increasing competition from custom silicon could pressure the entire group further.

Ultimately, the Google-Blackstone partnership represents more than just another AI infrastructure deal. It signals the next stage of the AI compute arms race, where hyperscalers, private equity firms, and infrastructure giants are all competing to control the backbone of artificial intelligence itself. The “wild west” phase of AI cloud computing may now be evolving into something far more competitive — and far less forgiving for companies relying solely on access to Nvidia hardware as their core moat.