- The author calls ASTS’s FLY deal a steal: $400m upfront cash plus up to 14m ASTS shares for a perpetual Miranda/Vira licence, followed by $500m over three years.
- Access to Firefly’s engines addresses the hardest launch bottleneck; five-satellite Blacklift flights could reduce the 248-satellite constellation’s launch cost from $6.1bn on Falcon 9 to $2.1–2.8bn.
- Adding $1–2bn for the airframe, recovery, factory and pad moves estimated breakeven to roughly 100–150 satellites, with $2–3bn remaining savings; ASTS already has external launches for 100+ satellites through 2028.
- The capital burden remains unresolved: the author expects carrier-partner investments, while management targets a late-2028 first launch and the author assumes Q1 2029.
Investment pitches
filter:
83 79 - OpenAI’s $50–70bn revenue debate may miss the issue: revenue can grow rapidly while frontier-lab valuations fail to produce shareholder value.
- Training is expensive, but low switching costs and model-routing platforms let workloads move to the cheapest model that reliably completes each task.
- Meta, Google and Microsoft control distribution, user context and monetization; their platforms can subsidize inference while standalone labs struggle to match ads or token pricing.
- The author would not invest in standalone labs at current valuations, preferring AI platforms, inference providers and semiconductor suppliers; specialized bio lock-in remains a possible Anthropic exception.