When you hear “data center,” what do you picture? If the image is a humming, GPU-stuffed hangar training the next large language model, you’re only picturing roughly 14% of the building. The headlines all suggest the opposite.
Every new data center announcement is framed as an AI build, and every electricity forecast leads with chatbots. The reality is our demand for this infrastructure is woven deeply into our daily life online, which flows through data centers, and almost none of it touches generative AI.
According to Goldman Sachs Research, today’s global data center capacity breaks down this way: cloud computing 54%, traditional business workloads like email and storage 32% and AI 14%. Add the first two together, and the punchline is clear: 86% of what data centers do has nothing to do with AI.
The single largest source of internet traffic isn’t a chatbot; it’s video. Sandvine puts YouTube at 16% of all internet traffic and Netflix at 12%. Add Disney+, Spotify, Twitch, gaming downloads and TikTok, and video apps alone account for roughly 39% of fixed download traffic. Every stream is cached and delivered by content delivery networks. Cloudflare and Akamai now move more than 80% of internet traffic through edge servers inside data centers. When a Netflix episode loads in three seconds, a copy of it is sitting in a data center 50 miles from your couch.
Then there’s the office. Email, calendar, documents, Slack, Zoom, Salesforce, Dropbox. Every tool that “just works” from any browser is a data center workload. The Radicati Group estimates 376 billion emails were sent and received every day in 2025, routed through 4.6 billion mailboxes. Global cloud infrastructure spending will exceed $400 billion in 2025, growing 25% year over year. Amazon, Microsoft and Google together hold 63% of that market.
It’s there again, at the heart of the economy. Every time a card touches a reader, a data center wakes up. Visa alone processes about 322 billion transactions a year, roughly $16 trillion in payments, with peak capacity north of 65,000 transactions per second. McKinsey finds more than half of banks now run mature cloud programs and over 70% are past the pilot stage. Every balance check, online order and brokerage trade is data center work.
The least photogenic workloads might be the most essential. Domain name lookups, security filtering, identity checks, backups, software updates, fraud detection. None of it makes the news, but all of it runs in data centers. Synergy Research counts 1,360 hyperscale data centers operating worldwide, accounting for 48% of global capacity. The rest sits across colocation facilities and on-premise enterprise rooms.
As the newest technology, AI’s share of data center usage is the fastest-growing slice, with IEA reporting data center electricity use surging 17% in 2025, reaching 485 terawatt-hours, and demand at AI-focused facilities growing 50%. AI racks pull 60 to 160 kW versus 5 to 10 kW for a standard server rack. Data center electricity is projected to roughly double to 945 TWh by 2030, about 3% of global power, and while AI’s share of that capacity could climb as high as 50%, our global demand will continue to stay anchored in our total digital demand. A data center isn’t an AI factory. It’s the crucial infrastructure behind your morning email, your lunchtime card tap, the Zoom meeting you took, the show you streamed last night while resting your eyes, and yes, now the AI planning session you had ahead of the big meeting.
Data centers are our fastest-growing infrastructure because with every digital breath we take, we demand it. Legitimate concerns about land use, water and energy are real policy questions, but we have to recognize that we already voted for this construction boom with our devices. We can either choose to guide how the future is delivered to us, or we can be paved over as it passes us by.