One row per AI lab.…: How many gigawatts of…

Part of: One row per AI lab. Purpose: a comparable battery across new AGI-style labs with multibillion-dollar war chests, to gauge whether any could one day rival OpenAI or Anthropic in the race for transformative AI. Anchors for scale: OpenAI and Anthropic each operate or have contracted multiple gigawatts of compute (OpenAI's Stargate program targets 7+ GW with roughly 1 GW live in mid-2026; Anthropic has contracted up to 1 million Google TPUs plus multi-gigawatt expansions from 2027). Frontier training runs cost billions of dollars per model generation. Question per row: how many gigawatts of dedicated AI compute capacity will the named lab control, owned or contracted for its exclusive use and scheduled to be live, as of December 31, 2028? Resolve from credible reporting, company or partner announcements, and filings; convert reported chip counts to gigawatts using standard all-in power estimates (as used by Epoch AI or SemiAnalysis); include capacity under signed contract and live by the date; exclude options, letters of intent, and revoked deals. — view all 6 rows

How many gigawatts of dedicated AI compute capacity will Reflection AI control, owned or contracted for its exclusive use, as of December 31, 2028?✓#4

gw_capacity0.2GW

p10
0.05
p25
0.12
p50
0.2
p75
0.38
p90
0.7

rationale:

Final estimate: P10: 0.05 GW, P25: 0.12 GW, P50: 0.2 GW, P75: 0.38 GW, P90: 0.7 GW.

Current Contracted Baseline Initial logic and parameters are validated regarding Reflection AI's dedicated compute footprint and binding agreements with SpaceX and Nebius 55 sources. Standard processing applied.

Converting Lease Spend to Gigawatts Initial logic and parameters are validated for inferring capacity from comparable tenant contracts and hardware power limits 44 sources. Standard processing applied.

Excluded Sovereign and Partnership Capacity Initial logic and parameters are validated regarding the exclusion of the Shinsegae joint venture and the DOE Genesis Mission under strict resolution criteria 66 sources. Standard processing applied.

Trajectory and Future Outlook Evaluated alongside the company's broader operational trajectory, the capacity percentiles reflect the mix of scale-up and downside termination scenarios. Jumping directly to the final transformation, the distribution is estimated at a 0.05 GW lower tail for downside termination risks, a 0.12 GW P25, a 0.2 GW median reflecting maintenance of existing contracts, a 0.38 GW P75, and a 0.7 GW upper tail accommodating potential captive buildouts.

Evaluated alongside the company's broader operational trajectory, the capacity percentiles were tweaked slightly to reflect the mix of scale-up and downside termination scenarios.

Ask a followup

Ask a followup