Jeremie Eliahou Ontiveros

Jeremie Eliahou Ontiveros × Reyk Knuhtsen × Jordan Nanos

SemiAnalysis argues SpaceX could bring 10 GW of AI capacity online in 2027, selling scarce “emergency megawatts” for roughly $50 million per MW-year.Modeled frontier inference near $100 million per MW-year could repay GPUs in under a year and make NVIDIA financing plausible.Microsoft’s late-2027–28 capacity gap supports demand, while permitting, chips, uptime and political restrictions remain risks.

Jordan Nanos × Jeremie Eliahou Ontiveros × Nicolas Bontigui × Haroon Inam

AI rack density is pulling data centers toward 800 V DC, which can deliver almost triple the power through the same copper while reducing conductor weight, cost, and I²R losses.DG Matrix’s multi-port SST roadmap targets 400 kW units today, 6 MW in 2027, and 10 MW in 2028, but adoption timing, mixed hardware, protection requirements, grid bottlenecks, and cybersecurity remain material execution risks.

Andrew Walker × Jeremie Eliahou Ontiveros

Modular on-site gas is becoming the fastest route to energizing AI data centers, with 3-40 MW systems and credible hyperscaler projects targeted for 2026-27 while displacing diesel backup demand.SemiAnalysis expects hyperscaler capex to grow 35%-40% in 2026 and 25%-30% in 2027, but CoreWeave and Oracle face 15-year infrastructure obligations against roughly five-year GPU contracts, alongside GPU depreciation risk.

Andrew Walker × Jeremie Eliahou Ontiveros

AI has broken the data-center growth curve: 2024 NVIDIA hardware added about 5 GW of demand versus roughly 4 GW of global annual IT-capacity growth from 2019 through 2023.That shortage supports infrastructure suppliers and NVIDIA’s networking-led inference moat, but the cycle ultimately depends on traffic, OpenAI and perhaps Claude revenue indicators, and whether users pay $20 or $200 monthly.