A highly detailed, month-by-month financial model for xAI/SpaceX’s AI data center leasing business. This highlights three existing major leasing deals already in place worth $27.8 billion annualized. Google deal ramping to nearly $1 billion per month (fully online by October) and predicts 1–2 additional major leasing deals will be announced before the end of the year (2026). This could push SpaceX toward an $82 billion annual run rate (roughly $7 billion per month).

The analysis is built on power capacity at Colossus 2 approaching 2 GW (using mobile gas turbines from APR Energy, Caterpillar JV, and upcoming larger permanent units)
Chip count estimates from Epic AI showing ~440,000 chips already working (330k B300s + 110k B200s).

Rapid power deployment solving the #1 bottleneck (energy) that is holding back hyperscalers like Microsoft and Google.

Larry Goldberg, a highly experienced venture capitalist who has co-founded multiple companies, was genuinely impressed by the depth, granularity, and grounded nature of the analysis. He calls the work “incredible” and “fantastic,” noting how rare it is to see this level of month-by-month modeling instead of high-level year-by-year guesses.

His key assessment of the core of the analysis was that it holds up very well and is directionally correct. Larry specifically highlighted that the demand side is real and urgent — hyperscalers have massive signed deals they can’t monetize because they lack available, working GPUs and power. He agreed that xAI/SpaceX is uniquely positioned to capture this because they can actually deliver monetizable capacity in months rather than years. Larry viewed the power and chip buildout numbers (especially the 2 GW and 440k chip estimates) as the most important part of the thesis, and he found the accounting treatment of capex/depreciation realistic. Overall, he described the company as “one of the most exciting companies on Earth right now” and said most of the assumptions felt “pretty darn fair” after reviewing the work.