Power agreements are reshaping the AI fiber opportunity

THE DEVELOPMENT

Two deals make the path to power more concrete

On October 6, Google and Constellation announced a 20-year agreement supporting 890 megawatts of additional output from upgrades to 11 existing nuclear reactors in Illinois, Pennsylvania and New Jersey. Constellation plans more than $4.3 billion of investment. The first upgrade is expected in 2028; Google expects the full additional capacity before the end of 2032.

A separate 15-year agreement covers 2,700 megawatts from operating generation. Together, the agreements total 3,590 megawatts, or 3.59 gigawatts. One gigawatt (GW) equals 1,000 megawatts (MW). Only the 890 MW component adds new generating capacity. A regional supply contract also does not establish when electricity will be available at any particular data-center campus.

The agreements serve PJM Interconnection, the regional organization that coordinates the high-voltage electricity grid and wholesale power markets across all or parts of 13 states and Washington, D.C. PJM manages the shared system; it is distinct from the local utility connecting a campus.

On October 5, TeraWulf announced an amended Kentucky Power agreement increasing future contracted capacity at its Muskie Data Campus in eastern Kentucky from 500 MW to 1 GW. The first 500 MW remains targeted to begin ramping in 2028. Planned delivery of the second 500 MW advances from 2030 to 2029, subject to Kentucky Public Service Commission approval and Kentucky Power’s construction schedule.

Kentucky Power is developing a high-voltage substation to serve Muskie. TeraWulf is also evaluating an eventual 2 GW campus, which would require further planning, infrastructure and agreements. Its announcement identifies no Muskie tenant. The expansion improves the development opportunity, while committed customer demand remains to be demonstrated.

PREMIA VIEW

Follow credible power plans, then stage the fiber investment

Premia’s assessment is that contracted power increasingly distinguishes the most credible AI infrastructure projects. Google’s approach illustrates how a major customer can support investment in the electricity system itself. For fiber operators, network engineering firms and infrastructure investors, this strengthens the case for evaluating power readiness alongside customer demand.

Give greater planning priority to campuses with named power suppliers, signed agreements, funded infrastructure and realistic delivery schedules. A campus relying on future grid expansion needs stronger evidence before it warrants comparable construction spending. Even a credible power contract leaves local connection work, permits, financing and execution to verify.

Use a staged investment sequence: establish contracted power and its funding; confirm regulatory and construction milestones; validate the computing customer’s contractual commitment; finalize network design; then release major fiber construction spending. Early route studies and customer discussions can run in parallel. Each larger spending decision should require stronger evidence.

For the developments covered here, the immediate opportunity is route planning, engineering and customer development. Building too early can leave a network unused while power or tenants arrive later than expected, tying up capital and delaying revenue.

WHAT REMAINS UNCERTAIN

The announcements do not prove that a particular campus has an unconditional opening date or that every required local connection is funded and ready. Muskie’s utility approval and construction schedule remain material conditions; its customer interest should not be treated as a signed tenant commitment.

DECISION QUESTIONS

Three questions stakeholders should answer now.

01
When can this campus actually use the electricity?
Verify the supplier, committed capacity, required approvals, local connection and construction schedule. Identify who is accountable for each remaining milestone.
02
Who will pay for the network—and for delays?
Check the customer contract, minimum payments, construction contributions and delay protections. Confirm the amount of capital exposed before revenue begins.
03
What evidence releases the next round of spending?
Separate modest planning expenditure from major construction. Tie each commitment to documented progress on power delivery, campus readiness and customer obligations.

WHAT TO WATCH

Delivery milestones and signed customers will determine the opportunity

Watch the reactor upgrade schedules and the local utility commitments that connect regional supply to individual campuses. At Muskie, follow the regulatory decision, substation progress and any announced tenant agreement.

Premia’s practical test is whether power readiness, customer commitments and the network delivery schedule support one another. The investment case strengthens as those pieces become verifiable.

SOURCES

Reporting and primary materials.

Premia’s analysis reflects information available as of the publication date. Reported facts are attributed to the linked sources; the Premia View and decision questions are our analysis.