Aaron Ross and Aaron Dillon open on SpaceX's first public earnings report, and Dillon's position is that the historical quarter is close to irrelevant: the large compute and data center commitments landed too late in Q2 to move reported revenue in any material way, which means the entire information content of the event sits in the forward guide and in whatever the CFO — not Musk — is willing to say about how compute converts to earnings. Sell-side consensus is scattered, and the scatter is diagnostic: analysts don't have visibility into capex on new data center capacity, so the range of outcomes on the guide is wider than the range of outcomes on the print. Dillon's long-run framing is a $5T+ market cap, held independently of any single quarter. From there the conversation moves to the Boring Company's reported ~$20B raise, where the house view is considerably more cautious — tunneling is a physical-world business with a physical-world throughput ceiling, and the revenue model is structurally government-driven, which caps the TAM in a way software doesn't. The Las Vegas Loop was built for roughly $48.7M, and the interesting wrinkle is per-rider variable revenue on top of the build contract, a toll-road-like annuity that keeps paying — but the commercial applications (utility conduit, fiber runs, anything that doesn't require a municipal procurement cycle) are where the multiple would have to come from, and those aren't visible yet. Conviction is lower here than elsewhere in the book, and entry-point discipline argues for waiting. On prediction markets: Polymarket is reportedly marking toward $20B after an April round at $15B, on annualized revenue above $1B, with Kalshi rumored to be raising at a substantially higher mark and Robinhood disclosing that prediction markets now out-earn equities trading. The structural read is the one that gets missed — roughly 0.04% of users capture about 70% of profits, and something on the order of 84% of participants lose money, which makes this a quant-extraction venue wearing information-market clothing. The genuinely interesting call option is institutional: a hedging layer where a restaurant hedges tomato input costs, or any SMB hedges a commodity exposure it currently has no instrument for. And the demand driver almost nobody underwrites correctly is non-US access — the Interactive Brokers precedent, where the majority of accounts are foreign nationals opening US brokerage access, is the right analogue, and it's the same Starlink dynamic: the revenue isn't the American who already had DraftKings, it's the four billion people who never had anything. Closing on Base Power — Zach Dell's residential battery storage business, $1B raised at a $13B valuation, 23,000 units installed across Texas and the Chicago area at roughly 100 installs a day — buying grid power at trough pricing overnight and discharging at peak. With PJM capacity clearing prices reportedly up several hundred percent year over year, distributed storage is running the same cost-compression playbook against electricity that Baseten and the inference layer are running against tokens, and we would watch the pairing of home storage, rooftop solar, and eventually local compute as a single residential utility stack.
Ross Pre-IPO | AGDillon & Co.
Institutional-grade research on private and pre-IPO companies.
Nothing in this video is investment advice. Valuations and figures discussed are as reported at time of recording.