What changed
Companies building AI data centers are pulling jet engines from retired aircraft, converting them to run on natural gas, and using them to generate electricity because the wait for grid connections now stretches seven years in major markets like Northern Virginia. Elon Musk's xAI installed eight converted jet turbines in Mississippi to power its Colossus 2 data center. The engines come from companies like ProEnergy, which rebuilds Boeing 747 engines for power generation. A facility using jet turbines can go live in about 24 months, versus seven years to hook up 100 megawatts to the grid in Northern Virginia, according to industry analysts. The trade-off: converted jet engines are less efficient, pollute more, and require more units to produce the same power as purpose-built gas turbines.
What you can do with it now
This doesn't change what your team does Monday, but it explains the price and delivery time on every cloud AI service you're quoted this quarter. If the data centers running ChatGPT, Claude, and Gemini can't get power from the grid for seven years, they're either paying for expensive shortcuts like jet turbines or they're telling customers the service isn't available in your region yet. Watch for both. When you price a new transcription service, a voice bot, or any tool that says it runs on frontier AI, ask: is this running on shared cloud infrastructure or dedicated capacity? If it's shared, check whether the vendor has a waiting list or a regional rollout—that's a signal the power isn't there yet. If the vendor says capacity is guaranteed, get the SLA in writing and confirm what happens to your price if they have to add emergency power later.
What to keep in your own hands
Don't lock your core workflows into a tool whose pricing depends on infrastructure that doesn't exist yet. If the data center can't get power for seven years, the vendor will either raise your price, cut your access, or sell the contract to someone who can. Before you commit to a two-year AI contract, confirm: does the vendor own the data center, lease it, or resell capacity from a hyperscaler? If they're reselling, what happens to your service if the hyperscaler raises rates or rations access? Can you export your data and move to a different provider in under 30 days? The jet engines are a symptom: the AI infrastructure build is moving faster than the power grid can support it. Your contract should assume that gap won't close on schedule.