Two sites can offer the same AI data center two very different kinds of “cheap.”
Site A offers electricity at $50/MWh.
Site B charges $75/MWh.
If that were the whole story, Site A would look obviously better.
But Site A cannot connect to the grid for 48 months.
Site B can connect in 18 months.
Now the question changes.
The cheapest electricity is not necessarily the cheapest site. Power price tells you what it costs to run. Time-to-power tells you when you can start.
By the end of this article, you will be able to compare a cheap-but-slow site with an expensive-but-fast site, calculate the break-even value of delay, and see when a low power price stops being the better business choice.
The Two-Site Problem
We will compare the same 100 MW IT data center in two locations.
| Input | Site A: cheap, slow | Site B: costly, fast |
|---|---|---|
| Electricity price | $50/MWh | $75/MWh |
| Grid connection delay | 48 months | 18 months |
| Difference | Site A saves $25/MWh but waits 30 months longer. | |
The data-center model is the same one used in the earlier articles:
- 100 MW IT capacity
- 85% average load factor
- PUE 1.30
- 967,980 MWh annual electricity use
The question is not simply:
Which site has the lower electricity price?
The useful question is:
How much is the 30-month head start worth?
Step 1: Calculate the Annual Electricity Advantage
At Site A:
967,980 MWh × $50/MWh = $48.399 million/year
At Site B:
967,980 MWh × $75/MWh = $72.5985 million/year
So Site A saves:
$72.5985M − $48.399M = $24.1995 million per operating year
If we stop here, Site A looks much better.
But it also starts 30 months later.
The Contexta Price-vs-Time Split
There are two different economic questions hiding inside “cheap power.”
| Price question | Time question |
|---|---|
| How much does each MWh cost after the site opens? | When can useful compute and revenue begin? |
| Recurring operating-cost effect | Schedule / opportunity-cost effect |
| Can accumulate over many operating years | Can dominate when capacity is scarce or equipment depreciates quickly |
A site decision needs both columns.
Step 2: Put a Scenario Value on One Month of Delay
Waiting for power can create several costs:
- land and buildings sit underused
- financing costs continue
- staff and development teams keep working
- ordered equipment may age before operation
- newer accelerators may arrive before the site opens
- customers may sign elsewhere
- revenue starts later
For a first screening model, we can compress those effects into one visible input:
monthly value of delay
Suppose we use $2 million per month.
This is not a market average.
It is an assumption the reader should replace.
| Screening item | Site A | Site B |
|---|---|---|
| One year of electricity | $48.4M | $72.6M |
| Illustrative delay value | 48 × $2M = $96M | 18 × $2M = $36M |
| One-year screening score | $144.4M | $108.6M |
Under this deliberately simple convention, Site B has the lower screening score by about $35.8 million.
The expensive electricity is offset by the earlier connection.
Step 3: Find the Decision-Flip Point
Instead of arguing about whether $2 million per month is “right,” we can solve for the threshold.
The annual electricity advantage is $24.1995 million.
The schedule gap is 30 months.
So the one-year screening break-even is:
$24.1995M ÷ 30 ≈ $806,650 per month
This gives a cleaner rule:
- if one month of delay is worth less than about $806,650, Site A has the lower one-year screening score
- if one month of delay is worth more than about $806,650, Site B has the lower one-year screening score
The threshold belongs to this screening convention. Change the energy price, delay, PUE, load factor, or economic value of time and the decision moves.
The Contexta Decision Flip Test
The general screening equation is:
Break-even monthly delay value = annual power-cost gap ÷ connection-delay gap
This is useful because it turns a vague argument—“fast power is valuable”—into a number the decision team can challenge.
The value can then be compared with:
- financing carry
- expected gross margin from scarce compute
- customer-contract value
- equipment depreciation
- staff and lease carry
- strategic cost of entering a market late
But One-Year Screening Is Not Project Economics
The break-even test is intentionally simple.
It compares one year of recurring electricity-cost advantage with the extra delay.
A real project may run for ten, fifteen, or twenty years.
That creates a second question:
How long does the cheap site stay cheap after it finally opens?
If Site A continues saving about $24.2 million every operating year, those savings can accumulate over a long project life.
But Site B receives a 30-month head start.
A full model therefore needs:
- discounted cash flow
- revenue ramp
- equipment replacement cycles
- power-price escalation
- taxes and incentives
- grid-upgrade payments
- probability of further connection slippage
- terminal value or common comparison horizon
The $806,650 threshold is not a universal site-selection rule. It is a transparent first screen that shows which assumption needs a deeper financial model.
The Contexta Horizon Test
When comparing a cheap/slow site with an expensive/fast site, ask which economic horizon dominates.
| If this dominates... | Then this may matter more... |
|---|---|
| scarce near-term AI capacity and rapid equipment turnover | earlier energization |
| long, stable operating life | recurring electricity price |
| large uncertainty in connection date | schedule risk, not just expected delay |
| future expansion to 500 MW or 1 GW | headroom and grid upgrade path |
There is no universal “cheap site.”
There is only a site whose total economics fit the workload and time horizon.
Why Time-to-Power Is Now a Real Site-Selection Variable
The grid-delay problem is not theoretical.
The IEA estimates that grid constraints could delay around 20% of global data-center capacity planned for construction by 2030.[1]
In the European Union, connection waits can range from roughly two to ten years, while average queues in major FLAP-D data-center hubs can reach roughly seven to ten years.[2]
The IEA also notes that new data centers can be built in roughly 1–3 years while new grid infrastructure can take roughly 5–15 years.[3]
That timing mismatch is why “available power” and “cheap power” are not the same product.
A Current 2026 Example: Power Delay Can Reach the Contract
In September 2026, Reuters reported that Oracle issued a force-majeure notice connected to potential power delays at Project Jupiter in New Mexico. Oracle said the project remained on schedule, but the episode showed that power timing can affect lease payments, financing structure, and risk allocation in a very large AI infrastructure project.[4]
The lesson is not that every delayed grid connection creates the same financial outcome.
The lesson is simpler:
time-to-power can move from an engineering schedule into the contract and financing model.
Cheap Power Can Also Hide a Complicated Tariff
A headline electricity price such as $50/MWh may not describe the full service economics.
Berkeley Lab’s August 2026 review of large-load tariffs shows that utilities and regulators are using combinations of rate structures and electric-service agreements to manage risks such as insufficient supply, underused grid investment, and cost shifts to other customers.[5]
So a site team should separate:
- energy price
- demand or capacity-related charges
- minimum bills or commitments
- grid-upgrade contributions
- contract term
- escalation
- exit or delay provisions
“$50/MWh” can be a useful model input.
It is not automatically the all-in delivered cost.
Why Developers Are Sometimes Paying More for Faster Power
Reuters reported on September 29, 2026 that U.S. data-center developers are increasingly considering smaller behind-the-meter gas turbines because they can sometimes be deployed faster than waiting for larger grid or generation projects.[6]
Those systems can have higher lifetime energy costs than large combined-cycle plants.
Yet some developers still consider them.
That is exactly the economic trade-off in this article:
A higher cost per unit of electricity can still make business sense if it creates valuable months or years of earlier operation.
The Contexta Time-to-Power Checklist
Before choosing a data-center site, ask at least these seven questions.
- What is the all-in electricity cost?
Not only the headline $/MWh. - What MW can actually be delivered?
Initial phase and full buildout. - What is the firm connection date?
And what still has to happen before that date? - What is one month of delay worth?
Use a range, not one magic number. - How likely is the date to slip again?
Expected delay and schedule risk are different. - Can the site expand?
A fast 100 MW path may be weak if the business needs 500 MW later. - What happens if faster power costs more?
Run a break-even test instead of assuming lower $/MWh wins.
Three Numbers to Put on the First Page
If the site-selection deck has room for only three numbers, start here:
All-in power cost + firm time-to-power + monthly value of delay
Then add expansion headroom and reliability.
This turns “cheap power” from a slogan into a decision model.
What Should You Watch Next?
- Connection certainty: signed agreement or early queue estimate?
- Large-load tariff design: how much cost is fixed, variable, or contingent?
- Behind-the-meter power: temporary bridge or long-term strategy?
- Transformer and substation milestones: ordered, permitted, under construction?
- Power-price escalation: does the cheap site stay cheap over the project life?
- Revenue urgency: how valuable is one earlier month of compute capacity?
- Expansion path: can 100 MW become 500 MW without restarting the grid process?
The Main Idea
The cheapest electricity price is not enough to choose a data-center site.
You need at least two clocks:
the operating-cost clock and the time-to-power clock.
In this illustrative example, Site A saves about $24.2 million per operating year but opens 30 months later.
Under the one-year screening convention, the decision flips at about $806,650 of value per month of delay.
Cheap power has value only after power can reach the site. The better site is the one that creates the stronger total outcome after price, time, risk, and expansion are included.
Continue Reading
- What Must Be Built to Power the AI Data Center Boom? — see the infrastructure chain behind time-to-power.
- AI Data Center Power Cost Calculator in Python — change the price, delay, PUE, and load assumptions yourself.
- The Grid Bottleneck Behind the AI Boom — follow the long-lead equipment and construction chain.
Key Terms
- time-to-power: the time before a site can receive enough dependable electricity to operate its planned load
- grid connection delay: the gap between the desired energization date and the date sufficient grid service becomes available
- break-even value: the threshold where two modeled choices produce the same screening outcome
- delay value: an assumed economic value for one month of waiting, including opportunity and carrying costs
- all-in power cost: total electricity-service economics beyond a simple energy price
- large-load tariff: utility rate and service structure designed for very large electricity customers
- connection certainty: confidence that the promised MW and energization date can actually be delivered
- schedule risk: the possibility that a planned connection or opening date slips further
- expansion headroom: additional power capacity that can support future campus growth
- behind-the-meter generation: onsite or colocated generation serving a customer without relying entirely on the normal grid path
Sources
- IEA — AI and Energy Security — estimate that grid constraints could delay around 20% of planned global data-center capacity through 2030.
- IEA — Overcoming Energy Constraints Is Key to Delivering on Europe’s Data Centre Goals — European connection waits and FLAP-D queues.
- IEA — Electricity 2026: Grids — grid build times, connection queues, and large-load bottlenecks.
- Reuters — Oracle power-delay force-majeure report, September 24, 2026.
- Lawrence Berkeley National Laboratory — Electricity Rate Designs for Large Loads: 2026 Update.
- Reuters — Dash for small gas turbines set to impact data center costs, September 29, 2026.
Status checked September 30, 2026. The Price-vs-Time Split, Decision Flip Test, Horizon Test, and Time-to-Power Checklist are The Contexta analytical frameworks. The Site A/B values and $806,650/month threshold are illustrative screening results, not universal market benchmarks or a full project-finance model.