Back to Forecasting
TLNActive

Will AWS draw 960+ MW from TLN's Susquehanna campus before 2029?

Resolves March 31, 2029(1107d)
IG: 0.48

Current Prediction

43%
Likely No
Model Agreement78%
Predictions9 runs
Last UpdatedMarch 19, 2026

Why This Question Matters

AWS ramp acceleration is the upside catalyst the narrative depends on. The $18B PPA revenue stream is contracted but timing-dependent. Faster draw validates that AI data center power demand is materializing, not theoretical. This market tests the gap between the AI narrative (DIVERGING per Myth Meter) and operational reality.

COMPETITIVE_POSITIONNARRATIVE_REALITY_GAP

Prediction Distribution

0%25%50%75%100%
opus
sonnet
haiku
Range: 38%47%Aggregate: 43%
Individual Predictions(9 runs)
opusRun 1
42%

This is a 3-year horizon question with high uncertainty. The PPA allows acceleration, but Amazon must actually build and commission data center capacity at the campus. Physical construction, permitting, equipment delivery, and staffing all constrain the pace. While AI demand is strong, going from 480 MW to 960 MW requires doubling infrastructure. Demand exists but physical delivery constraints create genuine uncertainty. Below 50% reflects long timeline risk despite favorable demand environment.

Physical construction constraints3-year horizon uncertaintyAI demand supports but doesn't guarantee paceDoubling infrastructure is a major undertaking
opusRun 2
38%

The original ramp schedule targets full 1920 MW by 2031/32, which means 960 MW by roughly 2028/29 on the original timeline. Hitting 960 MW before 2029 would require acceleration. While Amazon has the capital and motivation, data center construction timelines are physical constraints — you can't speed up concrete curing or transformer delivery. The campus is actively under construction but scale-up from 480 to 960 MW is a massive jump.

Original schedule has 960 MW around 2028/29Physical construction speed limitsAmazon has capital but faces supply constraintsMassive infrastructure jump
opusRun 3
45%

The AI capacity race is unprecedented in its intensity. Amazon, Microsoft, and Google are all racing to build infrastructure. This creates strong demand-side pull. Amazon has exercised acceleration options before (the 2.0 PPA doubled the original ISA). But the question is whether physical buildout can match the demand signal. I give a slightly higher probability because Amazon's stated urgency and the contractual acceleration mechanism create a genuine pathway.

AI capacity race intensityAmazon has used acceleration beforeContractual mechanism existsPhysical buildout is the constraint
sonnetRun 1
40%

The ramp schedule targets full capacity by 2031/32. 960 MW by 2029 would mean reaching half-capacity 3 years early. While demand supports acceleration, data center construction involves long lead times for power transformers, cooling systems, and networking. Amazon may want to accelerate but physical constraints limit the pace. 40% reflects genuine uncertainty with a slight lean toward not achieving this ahead of schedule.

3 years ahead of midpoint targetLong lead time equipmentPhysical construction constraintsDemand supports but physics limits
sonnetRun 2
47%

I'm slightly more optimistic. Amazon has been building data centers at an unprecedented pace globally. The Susquehanna campus has existing infrastructure (the campus is already energized). Power is not the constraint here — Talen has the generation capacity ready. The constraint is Amazon's data center construction speed. Given Amazon's massive capex ($200B+ announced), they have every incentive to build as fast as possible on a site with guaranteed power.

Amazon building at unprecedented paceSite infrastructure already energizedPower supply is not the bottleneckMassive capex commitment
sonnetRun 3
43%

Splitting the difference. The demand signal is real but 960 MW before 2029 is an aggressive timeline. Historical data center buildout timelines suggest 2-3 years per phase. Going from 480 MW operational to 960 MW requires completing one full additional phase. Possible but not the base case. Myth Meter's DIVERGING classification suggests the narrative may be ahead of execution reality.

2-3 year phase construction normOne full phase requiredPossible but not base caseNarrative may outpace reality
haikuRun 1
44%

Long-dated question with high uncertainty. AI demand strong but physical construction sets the pace. Slightly below 50% due to construction timeline constraints.

Long horizonStrong demandConstruction constraints
haikuRun 2
38%

Original schedule targets this around 2028/29. Meaningful acceleration would be needed. Equipment supply chains are constrained globally. Below 40%.

Original timeline around 2028/29Supply chain constraintsAcceleration needed
haikuRun 3
45%

Amazon has the motivation and capital. The acceleration clause exists. But physical construction is the binding constraint. Slightly below coin flip.

Amazon has motivationAcceleration clausePhysical constraints bind

Resolution Criteria

Resolves YES if Talen Energy discloses that Amazon's Susquehanna campus power draw has reached or exceeded 960 MW before January 1, 2029. Resolves NO if the 960 MW threshold is not reached by that date.

Resolution Source

Talen Energy earnings disclosure or 10-K/10-Q filing referencing AWS campus capacity draw

Source Trigger

AWS campus ramp acceleration — Track via earnings commentary. Any indication that Amazon is drawing capacity faster than the stated ramp schedule would be a positive catalyst. Threshold: 960 MW drawn before 2029.

moat-mapperCOMPETITIVE_POSITIONMEDIUM
View TLN Analysis

Full multi-lens equity analysis