Google Cloud in 2026
This executive summary was written on the research I completed with Stanford GSB professor Raj Joshi. It boils down my research and analysis of Google Cloud’s evolution from 2024-2026 into a concise overview.
Currently Google might look like it’s falling off in the AI race. Despite having all the data an AI lab could dream of, Gemini is lacking when you compare them to other frontier AI models on the benchmarks. A closer look, however, tells a different story. While their models aren’t at the forefront of the rankings, they’re positioning within the AI boom is perfect.
Introduction
In 2019, when Thomas Kurian joined Google Cloud, he was given an under-performing company and tasked with turning it into—at the very least—a viable business. To say he over delivered is an understatement. By investing in the full stack, building on an open ecosystem, and fostering an employee culture of success, Kurian grew Google Cloud from $5.8B in revenue in 2018 to a conservatively estimated $94B in 2026.1
Throughout this overview, the three major overarching explanations for Google's unimaginably immense growth will be covered. While the sections are organized into the full stack, customer growth, and their open culture, they do still overlap and support each other.
Full Stack Strategy
To control quality over each layer, Kurian had made sure Google Cloud invested in owning the full stack. That included the models, the compute, and specifically, the chips. That bet would pay off in the coming years as they would have control over every layer in the AI boom. They could sell the models, sell their compute to run other models, and sell the TPU’s to those who needed control over the compute themselves. The business model was further complicated as they needed to work with entities that were simultaneously buying Google’s compute and chips but also competing with Google as they trained, ran and sold their own models. More on this in the Open Culture portion.
Because the AI boom created such demand for each layer, Google was able to capture a large portion of AI spending and it’s largely paying off, showing up in revenue and margin growth across the business. The demand itself, however, had outgrown their supply, and Google Cloud agreements just continued to pile up. This backlog partly explained the increase in revenue over the past few years, but to understand it better, we need to see the full picture—the customers.
Customer Growth
The most interesting thing about Google Cloud’s growth over the past few years was, although they have already grown exponentially, that it did not seem to be decelerating. While the underlying reason is their investment in the full stack, the revenue acceleration was seen through the parallel growth in customers and how much they were buying.
The reason Google Cloud was seeing such a backlog was due to two things. First, existing customers who were trying to get in on the AI race were spending more than they had previously.
That increase in customer spending had contributed to revenue growth already, but the boom had created a massive wave of new customers. These new companies were, even if not fully, also using Google Cloud. In a Goldman Sachs interview, Kurian shared that nine of the top ten frontier labs were using Google Cloud. While he didn’t specifically mention it, I believe the only reason they did not have all ten was due to geographic restrictions; Google Cloud does not work with DeepSeek which, to be fair, does not use Azure or AWS either.
The increase in existing and new customers had created a backlog that Google simply could not keep up with. Every couple of months, Google Cloud's new supply, funded by CapEx roughly doubling year over year, released a portion of their $514B backlog, converting it into revenue. But that customer growth, however, would probably not be as magnificent if it hadn’t been for the open culture which Kurian originally bet on.
Continuing The Open Culture
In an effort to let Google Cloud have a chance of competing with the big dogs in the cloud sector, Kurian had built it up with a culture of not locking customers into their own products. That philosophy has continued on and expanded into all of Google Cloud's products.
The first example of this doctrine could be seen through the products they had built on top of Gemini. While businesses might be hosting their data on other cloud platforms, storing was only a part of the problem for any business. To make any data or information useful, analyzing it and acting upon it was a necessity. Google’s investment into the full stack allowed them to provide tools for that. Since models were included in that stack, Google had been able to build solutions upon their Gemini suite which companies were free to use regardless of the cloud provider.
The more obvious instance though, was who they’re selling compute/TPU to. I touched on this briefly in the Full Stack Strategy section, but there is a lot to unpack here. In the AI boom, there have been two sides. The gold miners and the people selling shovels. Google has been playing both sides, and it has been working for now. On one side, they have been developing their own family of AI products (apps, API, models, etc) via Google DeepMind. On the other side, they have been selling the pickaxes too, and just like the pickaxe sellers, they have been winning more than the former. There’s a balance to keep though. They aren’t in a position like Nvidia where they can, or should, go completely into compute infrastructure, but they also aren’t in a position where they can, or should, become only a frontier lab like OpenAI or Anthropic. While it might seem counter intuitive to supply the competitor, it makes a lot of sense once you follow the flow of money. Currently, the gold miners aren’t making any profit, only the pickaxe sellers are. So if you’re able to fund the gold mining yourself without breathing, eating, and drinking venture capital money, you completely should. That is exactly what Google is doing now, and it’s working pretty well. The only question is where will this go and if it’s sustainable.
I personally don’t see a world where this goes wrong unless the sun explodes. Google has positioned itself in the most successful way possible. It’s no surprise they are one of the top three most valuable companies in the world. They are providing the two most difficult solutions to provide. Creating these models requires the greatest talent, infinite money, and all the data in the world. Google has all of that. Compute infrastructure on the other hand, even with the money, is hard to break into when Nvidia, TSMC, and established cloud infrastructure companies already own the market. Even if the AI race leads to locally hosted open weight models, they still win as these models would either way run on chips, which Google sells. It also can’t be left out that Google is one of the most prominent leaders in the open-weight AI space.
The most beautiful part of this whole story is that Google helped lay the foundation for the AI boom it now profits from. Google researchers developed the transformer architecture years before it became mainstream with its use in ChatGPT and other large language models. Without Google’s breakthrough, November 30th2 wouldn’t have happened.