Meta’s new deal with Google for AI cloud hosting comes with a big price tag – the deal is reportedly worth $10 billion. That’s the equivalent of $666 million large pizzas – or two for each person in the U.S. It’s also equal to the GDP of a small country. So what’s behind this monolithic business agreement?
Here are some interesting aspects of this newly announced tech partnership.
Meta’s Big Game Plan
The $10 billion deal with Google is part of a Meta plan to invest a total of over $100 billion in AI. Zuckerberg has repeatedly talked about this kind of scaling up, as the company lays off significant numbers of human workers, moves the survivors to departments that are heavier on AI, and automates business processes like it’s going out of style. The proclaimed “Year of Efficiency” in 2023 saw a lot of this action, and Meta doesn’t seem to be slowing down.
In Sickness and in Health
The deal also lasts until 2031, by which time we might have superhuman agents, godlike AGI, and brand new national economies built on crypto or some other weirder foundation. So you could be forgiven for imagining that, for all practical purposes, Google and Meta are essentially going to the altar, and not just making a short-term deal when it comes to tech allegiances.
Build vs. Buy
Another thing to know about this deal is that it is, in some ways and to some extent, a stopgap solution, and part of the greater debate for large companies like Meta about “build vs. buy,” that is, whether it is better to utilize an outside vendor’s resources, or build them internally. That extends to models as well as data centers: as LLMs became more ubiquitous, companies had to decide whether to run their AI applications on someone else’s model (cough cough, OpenAI) or build one themselves. And to be clear, OpenAI does have competitors, but in the grand scheme of things, it doesn’t have that many.
In terms of data center planning, Meta already has fairly concrete plans for Prometheus and Hyperion, which are projected to provide 1 gigawatt and 5 gigawatts, respectively, over time. So after that six years, we could see a very different landscape emerging.
Deal Bolsters Google Against Top Competition
In offering services in the cloud AI market, Google goes up against a formidable competitor. AWS has been the household brand for cloud practically since the beginning, and by now, the firm has launched AWS Bedrock, where users can actually choose from top models, from OpenAI, DeepSeek, or others including Anthropic (AWS was an early backer of Claude) and more.
But the new agreement gives Google a boost in AI cloud, and reportedly led to a 32% increase in revenue for the company since the deal was announced.
Will It Last?
Another part of the context is how customers, particularly smaller ones, feel about Google’s cloud services, including the new AI provisioning. Presumably, Meta would not be worried about a rug pull in terms of deprecation, but notes in places like Reddit show that others ponder whether Google cloud is “going to last” in the competitive environment, and alternately, whether Google will choose to deprecate services. Google’s documentation reveals that Google services are subject to deprecation, so that’s something to keep in mind, but doesn’t necessarily mean that something like that is in the cards, or that the company will be “churning” services and rocking smaller customers out of the boat.
Diversity of Services
Regardless of its deal with Meta, Google isn’t just in the business of offering bulk. Its GCP also offers feature services like BigQuery, a serverless data warehouse, and TensorFlow, an open-source ML framework enabling multi-dimensional arrays. There’s a Kubernetes engine, object storage, and more, helping the company to rival AWS’s ecosystem approach.
Here are some more of Google’s AI offerings:
Vertex AI – Google Cloud’s unified platform for building, training, deploying, and managing machine learning models with integrated MLOps tools.
Generative AI Studio – Vertex AI tool for creating, fine-tuning, and deploying generative AI models using prompts and customization.
Speech-to-Text / Text-to-Speech – prebuilt AI APIs.
Vision AI – cloud-based service that analyzes images and videos, detecting objects, text, faces, and visual patterns using machine learning.
All of this is going to help the company to remain a powerhouse in AI cloud along with AWS and Microsoft.
That’s a little bit about a deal that’s notable, partly because of its size, and partly because it involves not one but two of the companies included in the former acronym FAANG, which represented the five top tech companies in America. That list has shifted a little bit, notably, with the emergence of Nvidia as king of the stock market, but Meta and Google are still two titans. Stay tuned for more.
