Artificial intelligence spending has entered a new phase. The industry’s biggest technology companies are no longer debating whether to build more infrastructure — they’re competing to build it faster than everyone else. Meta Platforms (NASDAQ:META | META Price Prediction) is among the most aggressive, with capital expenditures expected to reach between $125 billion and $145 billion this year, according to the company’s latest guidance.
Last week, reports suggested Meta was preparing a new business selling excess AI computing capacity. This week, it announced a new $10 billion Canadian data center. At first glance, those headlines seem to point in opposite directions, but they just might reveal the same long-term strategy.
New Data Center Doesn’t Contradict Compute Strategy
Reports last week indicated Meta wants to create a new revenue stream by renting unused GPU capacity to outside customers, much like cloud providers already do. The idea is simple: if Meta has computing resources sitting idle between AI training cycles, why not monetize them instead of letting expensive hardware go unused?
Then came the announcement of a 1-gigawatt, roughly $10 billion data center in Alberta, Canada — its first major facility in the country. The timing prompted some understandable skepticism. If Meta expects to have excess compute available to rent, why is it adding another massive data center?
Surprisingly, that’s exactly the point. Hyperscale data centers take years to construct, while AI demand rises in bursts rather than a straight line. Meta isn’t building for today’s workloads. It’s building for where it believes AI demand will be in 2028 and beyond. Any temporary excess capacity becomes inventory that can generate revenue instead of remaining an idle cost.
Rather than undermining the cloud strategy, the Canadian facility expands the amount of compute Meta can potentially monetize.
Why Investors Are Rolling Their Eyes
Some investors are giving Meta a “facepalm” reaction to the announcement due to growing AI spending fatigue. Every major hyperscaler is spending tens of billions on AI infrastructure before corresponding revenue has fully materialized. That fuels concerns that the industry could create a compute glut, where everyone builds more capacity than customers ultimately need.
Meta also carries baggage. Investors still remember the tens of billions spent — and lost — on Reality Labs and the metaverse with little financial payoff. Adding another $10 billion project naturally invites comparisons, even though AI has far clearer commercial applications than virtual reality ever did.
Environmental concerns add another layer. Alberta officials have emphasized projects capable of securing their own power supplies, but critics note large AI data centers consume enormous amounts of electricity, with this project expected to rely heavily on natural gas generation.
The Bigger Bet Is Utilization
Meta isn’t expecting every GPU to remain busy with its own applications every hour of every day. That’s inefficient. Instead, management appears to be treating compute much like airlines treat empty seats or hotels treat vacant rooms. Internal AI projects receive priority. Excess capacity becomes a product.
Granted, skeptics argue every hyperscaler now seems to be planning the same thing — build more infrastructure than needed and rent the surplus to someone else. If every company follows that strategy, pricing pressure could emerge.
That said, Meta has one advantage many rivals lack. Its advertising business continues generating tens of billions in quarterly operating cash flow, giving it sufficient financial flexibility to absorb years of infrastructure investment while new revenue streams mature.
Key Takeaway
In short, Meta’s Canadian data center doesn’t undermine its plan to sell excess compute — it strengthens it. The company isn’t building because it already has too much capacity. It’s building enough infrastructure to satisfy its own AI ambitions while creating a cloud-like business capable of generating additional revenue from any unused compute.
Ultimately, the real question isn’t whether Meta is building too much. It’s whether utilization rates remain high enough to justify the investment. If management succeeds in keeping both internal AI projects and outside customers filling those servers, today’s $10 billion spending announcement could eventually look less like overbuilding and more like laying the foundation for an entirely new business. For long-term investors, that’s the metric worth watching — not simply the size of the construction bill.
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks