When puzzle pieces are out of order and spread across the table, it’s easy to miss the big picture. But there’s a reason they’re all in the same package. The same could be said of the wide range of announcements coming out of Google over the past year.
Allow me to connect some of the puzzle pieces for you. Google’s recent updates are all part of a plan to secure infrastructure dominance.
Notably, the company appears to be focusing its efforts across three main areas: compute, communications and power. These actions could position Google as a key player in what may become a unified infrastructure market, but if nothing else, they will secure the hyperscaler’s future as critical utility assets become increasingly interdependent and scarce.
Compute
Google may be No. 3 by market share in the cloud market, but it’s hot on the heels of its competitors when it comes to data center investments. The company is planning $175 billion to $185 billion capex in 2026, just shy of the $200 billion AWS is planning to spend but well above the $120 billion Microsoft has forecast for its fiscal year.
Already this year, Google has put in proposals to build data centers in Nebraska, Minnesota, North Carolina and Virginia, adding to previously announced projects spread across the U.S., Europe and Asia.
But Google isn’t stopping with terrestrial data centers. In November, it also announced Project Suncatcher, an initiative aimed at studying how best to scale AI compute in space.
It’s not alone in looking up. SpaceX has proposed a patently ludicrous constellation of 1 million data center satellites to expand AI processing power. AWS is also headed into orbit via Starcloud’s deployments of Outpost, though AWS CEO Matt Garman doesn’t seem to think orbital assets will become a near-term reality.
Google, though, is planning to “launch two prototype satellites by early 2027” to see if its Tensor Processing Units (TPUs) can withstand the sun’s radiation. Those will be on satellites in a sun-synchronous low earth orbit to maximize solar power and connect back to earth with free-space optical links.
The company’s rationale is clear. With land and power now priced at a premium, Google is looking for an out. And, believe it or not, yeeting servers into space may not end up costing too much more than building data centers to house them here on earth.
“Our analysis of historical and projected launch pricing data suggests that with a sustained learning rate, prices may fall to less than $200/kg by the mid-2030s,” Google wrote in a blog detailing the project. “At that price point, the cost of launching and operating a space-based data center could become roughly comparable to the reported energy costs of an equivalent terrestrial data center on a per-kilowatt/year basis.”
Power
All this talk about the price of power brings us to the next puzzle piece. Google has been hustling hard to hedge its energy bets in a market where demand is rising faster than electrical utilities (and, frankly, interconnection processes) can keep up.
The company in recent weeks has stepped up its focus on procuring renewable energy, coming out with back-to-back agreements with Xcel Energy for 1.9 GW in Minnesota and DTE Energy for 2.7 GW in Michigan.
But it’s also going beyond partnerships to become an owner of power generation assets itself. Google notably scooped up energy company Intersect for a cool $4.75 billion at the end of last year, netting itself solar and battery assets in Texas.
It’s also aiming to weave its technology into the electrical grid at large. In April 2025, the company announced its moonshot project Tapestry was working with grid operator PJM Interconnection to see how AI can be used to manage and optimize interconnecting power generation across PJM’s 13-state territory. Tapestry, of course, runs on Google Cloud and Google DeepMind.
While Google pitched the move as a way to help make electricity more “reliable and affordable” for citizens in PJM’s territory, it could also have the self-serving benefit of helping PJM connect new energy sources to the grid “much faster.”
Given Google now owns Intersect and is eyeing small modular reactor (SMR) deployments with Kairos Power to fuels its data centers with new nuclear generation, the throughline of where this is all headed is clear.
Communications
But physical infrastructure and the power to light it all up means nothing without communications pipelines to move data back and forth. Luckily, Google already has plenty of those.
The hyperscaler may have just sold off its residential fiber business, but that doesn’t mean it’s getting out of the communications game. The company owns and operates a much, much larger private fiber transport network.
As of last year, the company said its fiber infrastructure included a total of 2 million miles of lit fiber – including 33 subsea cables – that connect not only its data centers but more than 200 edge locations and 3,000 content delivery network locations.
Interestingly, it likely runs this network with significantly fewer people than would be needed for a legacy telecom network.
In April 2025, Google Cloud’s VP and GM for Network and Security Muninder Sambi told Fierce the company was well on the road from “autonomation to autonomous” and was working toward a network run by AI agents “with no manual intervention.”
Again, Google isn’t alone here. AWS' Matt Rehder told Fierce in January its fiber backbone is similarly mostly autonomous.
But Google also working to prepare its network for the realities of AI at scale with a new fundamental architecture. Specifically, it’s developed what it calls a “multi-shard network,” which – for lack of a better comparison – is somewhat akin to the wireless world’s network slicing.
Each shard has its control, data and management plane to limit the spread of issues across shards. Google can both expand individual shards and add additional shards to the network. More nerdy goodness about this architecture can be found here.
The big picture
With the puzzle pieces organized, the picture is clear: Google is setting itself up for infrastructure dominance across all utility sectors that matter. There are few companies that have the financial resources to compete, and those that do don’t have nearly as many ducks in a row.
Only time will tell if they have the will – or time – to catch up.
Opinion pieces from industry experts, analysts or our editorial staff do not represent the opinions of Fierce Network. Send us your opinion pieces and we'll take a look. Email fiercenetwork@questex.com.