From 2022 to 2023, the market for industrial gas turbines quietly collapsed. Manufacturers cut workforce, walked back expansion plans, and reduced capacity (all rational responses to genuine demand weakness). Then AI arrived, and those capacity cuts had locked in a shortage no amount of fresh capital can quickly reverse. GE Vernova's orderbook now runs deep into 2027. US turbine orders in 2024 hit their highest level since 2002. Crusoe has already contracted roughly a gigawatt of aeroderivative turbines for a campus in Abilene, Texas. SpaceXAI, formed by the April 2026 SpaceX-xAI merger, wants to control enough of this supply to outrun everyone else.

The panel's conclusion, after two rounds of analysis: the race probably has too many competitors for any single monopoly winner, and the actual choke point in the system sits upstream of the turbine.

Where the panel puts its weight

The most likely single outcome (33%) is a field of early movers each capturing real power advantage, with no single actor achieving monopoly allocation. The second-most-likely path (27%) is that the gambit runs into a constraint nobody in the public conversation is discussing: not the turbine factory, but the metal the factory depends on.

The history that explains the shortage

The sharpest disagreement

Complexity Economics (W. Brian Arthur, 2014) argues that the "no monopoly" outcome is unstable over time. Increasing returns to early compute access should push the field toward concentration rather than sustained competition. The counter is that Crusoe's gigawatt, the Williams/Blackstone pipeline, and existing GEV customer relationships may each represent a self-reinforcing position already locked in by different actors. Multiple partial monopolies, not a single winner.

Theory of Constraints (Eliyahu M. Goldratt, 1990) offers a different critique: the gambit targets the wrong part of the system. Gas turbine blades require nickel superalloys produced by a handful of specialized foundries with capacity locked up through at least 2025. Buying turbine manufacturing slots does not buy the metal those slots require. Controlling turbine assembly is a local optimization error when the actual binding limit is upstream.

What to watch

Three dates are particularly diagnostic. By September 2026: does SpaceXAI announce a turbine supply contract above 500 MW or a formal OEM acquisition? By October 2026: does SpaceXAI disclose an upstream investment in nickel superalloy production capacity? By November 2026: do two or more non-SpaceXAI hyperscalers announce comparable BTM deployments above 500 MW?

Yes to the first, silence on the second: the consolidation case (22%) gains weight, but the upstream warning remains live. Yes to the second: the consolidation scenario earns its probability with both legs. Silence on both: the race with multiple competitors (33%) extends its lead, and the path to structural monopoly narrows further.

No resolved calls yet. This is the panel's opening position.

In April 2026, Amazon bought Globalstar, the satellite company that carries iPhones' Emergency SOS calls. A hyperscaler now sits one layer below the emergency button on hundreds of millions of phones. That choice, who controls the screen rather than who owns the satellite, frames everything happening in satellite connectivity right now.

The weighted answer:

The most likely outcome (40%) is that nothing resolves cleanly before end-2028. Two or three satellite providers continue in carrier-specific arrangements; coverage improves, but wholesale pricing stays negotiated and opaque. In the second-most-likely future (28%), the margin migrates to Apple and Amazon, the companies that own the phone and the billing relationship. Across three of the four futures, it does not go to the carriers.

The one big disagreement

The sharpest split in the panel is between Disruptive Innovation (Christensen) and the fragmentation camp. Disruptive Innovation argues the satellite technology S-curve is in its steepest phase since 2022, and performance adequacy for routine, non-emergency use arrives in 2026-27. Once it does, disruption theory says fragmented markets tend to break. The fragmentation scenario (40%) bets that the inflection stalls. That is the most consequential uncertainty embedded in the top scenario.

How we got here

In 2020, AST SpaceMobile's BlueWalker 3 prototype proved a satellite could talk directly to an unmodified phone. Two years of regulatory waiting followed. Then 2022 brought two models in quick succession: SpaceX partnered with T-Mobile (satellite operator supplies coverage; carrier supplies the customer), and Apple launched Emergency SOS via Globalstar (no carrier; OEM owns the experience). The structural tension between those two arrangements has not resolved.

What to watch

• By September 2026: Does Apple announce satellite features beyond emergency SOS, routine messaging or location sharing? That is the clearest leading signal of OEM Vertical Capture (28%) accelerating.
• By October 2026: Does SpaceX renegotiate its T-Mobile deal, or launch a direct subscription product? A move here would push the 17% scenario sharply higher.
• By January 2027: Does a first open D2D wholesale tariff appear? A public tariff card is the single clearest sign the coverage layer has crossed from Product to Commodity.

This is the panel's opening position. No calls have resolved. The track record is empty, and should be read that way.

In September 2024, Microsoft signed a deal to restart Three Mile Island, a nuclear plant shuttered since 2019. The press called it an energy story. It was an infrastructure story: Microsoft needed power for data centers and could not get it from the grid fast enough.

That gap between when AI infrastructure needs power and when the grid can supply it is the constraint shaping everything right now. Not chips. Not money. The wire from the substation to the building.

The numbers make the picture concrete. ERCOT, the Texas grid, has 233-plus gigawatts of large-load requests queued for interconnection. Projects entering PJM, the largest U.S. grid operator, wait more than seven years on average before getting power. Globally, the IEA counts 2,500-plus gigawatts of generation stalled in interconnection backlogs. Total U.S. generation capacity is roughly 1,200 gigawatts. The queue is twice the grid.

What the panel sees: Three of four analytical frameworks predict that power and grid interconnection becomes the single binding constraint on AI deployment by 2027-28 (35% probability). Another 25% says memory friction persists alongside the grid problem, a two-front bind. Together, those paths carry a 60% probability that power is part of the constraint picture regardless of how chips resolve. The remaining 40% splits between an efficiency-driven demand slowdown (20%) and an architecture or geopolitical shock that reshuffles the stack (20%).

No calls have resolved. This is the panel's opening position.

How the constraint got here

Figure 1: The constraint migration, 2020-2026.

The constraint on AI infrastructure has been migrating up the stack since 2020. Foundry capacity first, then packaging and memory, then grid and power. Each time one layer cleared, it accelerated arrivals at the next. The cascade has not resolved. It has compounded.

The dissent worth taking seriously

Wardley Mapping (Wardley) argues that when a mature utility service is constrained, markets find workarounds. Hyperscalers have already started: Microsoft restarted a nuclear plant, Meta is building a 6.6-gigawatt nuclear portfolio, and brownfield industrial sites with existing power are being bought up. Wardley's prediction is that these mechanisms will partially circumvent the grid constraint before it reaches maximum severity, making the clean single-bottleneck outcome less decisive than the other three frameworks suggest.

What to watch

Two early checks matter. By October 2026, HBM lead times should show signs of normalizing (target: below 26 weeks by Q1 2027), and CoWoS should be tracking toward a supply gap below 5% (target: Q2 2027). If both clear on schedule, the freed capacity flows straight into the grid queue and the 35% scenario gains weight. If HBM stays stuck, the 25% two-front scenario becomes more plausible.

The grid check comes later: FERC/ISO median approval times falling below three years would be the first sign of real institutional relief, with a first read in January 2027.

The queue is twice the grid. That sentence is the forecast.