When you think of Facebook services, high-speed connectivity is not the first thing that comes to mind. But the social media giant is doing just that, offering high-capacity fiber-optic routes to sell unused capacity between its data centers for third parties.
Facebook has created a subsidiary called Middle Mile Infrastructure to sell excess capacity on its fiber, starting with new fiber routes between its data center campuses in Virginia, Ohio, and North Carolina. The company made the announcement in a blog post by Kevin Salvadori, director of network investments.
“We intend to allow third parties — including local and regional providers — to purchase excess capacity on our fiber,” he wrote. “This capacity could provide additional network infrastructure to existing and emerging providers, helping them extend service to many parts of the country, and particularly in underserved rural areas near our long-haul fiber builds.”
To be clear, the company will not be providing fiber services to consumers the way Google Fiber works. This is for carriers and operators. Construction is set to begin this year and take approximately 18 to 24 months.
One of the routes will connect a data center cluster in Ashburn, Virginia, with a new campus in New Albany, Ohio. That means the fiber will run through the state of West Virginia, a state that has struggled with broadband access because the carriers and operators have been slow to support rural states.
One thing the big hyperscalers (Amazon, Microsoft, Google, and Facebook) have in common is they often have to build supporting solutions because existing products don’t quite cut it. The Open Compute Project was borne out of dissatisfaction with existing hardware, for example.
And the big hyperscalers have also started building their own fiber backbones to connect their data centers, which are the size of malls and have to move immense amounts of data. Facebook recently introduced the Fabric Aggregator, custom networking gear that was driven out of necessity because its traffic outgrew what Cisco hardware could handle.
Pros and cons
My view on this is a bit mixed. On the one hand, it is welcomed that Facebook is monetizing excess capacity. That’s how Amazon Web Services (AWS) was born, after all. Amazon had all this idle capacity it had bought and installed for the Christmas crush sitting idle, so it did a little experiment renting its excess capacity. Boom, EC2 and S3 were born, and the rest is history.
On the other hand, this industry is legendary for overbuilding and being left with excess inventory. GPU makers are sitting on a ton of video cards right now because of the mad rush for cards for cryptomining. Well, the cryptocurrency market has since collapsed, and suddenly people aren’t buying a dozen top-of-the-line GPUs anymore. And guess who is stuck with the inventory?
Likewise, there is such a glut of DRAM and NAND flash memory that prices are set to drop 20 percent or more this year.
I don’t doubt for a second that data center capacity is any different. Google alone will spend $13 billion on data center expansion this year, while Research and Markets projects the global data center construction market will grow to $22.73 billion this year, which should do wonders for sucking up that excess memory supply.
But what happens if all that construction grinds to a halt, for whatever reason? Everything is cyclical. Facebook faces numerous headwinds, in particular. What if its popularity tanks?
The nation is already dotted with dead malls. When will we see the first dead data center?
