Starship Reached Orbit. Here's the $1B-Revenue-Per-Launch Math | The Brainstorm 152

Space Exploration Technologies

Across the shows: 1 speaker in other episodes, 0 bullish and 1 bearish calls, past 30 days

• Brett Winton · ARK InvestBullish

  Reusable Starship could make each launch worth about $1B a year in incremental revenue.

  Brett describes Starship's first orbital success with the v3 satellites, which are the first revenue-generating assets launched this way. He expects full reusability to bring launch costs below $100 per kilogram, versus $22,000 to $3,000 per kilogram at today's market price. Once the rocket is packed with Starlink satellites, each launch could yield something approaching $1B of annual incremental revenue, rising to $1B–$2B per launch if SpaceX catches up on AI and monetizes it. Even a crashed rocket, costing about $200 million, would pay back through five years of satellite revenue. Brett calls it probably the most valuable business artifact created.

  “Once they're fully packing it full of Starlink satellites, we think each launch could yield something approaching a billion dollars in annual incremental revenue to SpaceX for each launch as they scale.”at 19:38

• Sam Korus · ARK InvestBullish

  SpaceX's reusable-rocket lead is at least a decade long and not being encroached upon.

  Sam argues SpaceX's lead in reusable rockets is at least a decade. Competitors are mostly replicating the Falcon 9 approach. Stoke is trying a fully reusable rocket, but it is building a smaller vehicle, so SpaceX's lead is not under threat for now. He also says the enterprise-driven revenue per launch could be steadier and hold at a high price point, while the AI-compute decision (renting versus monetizing it internally) creates fluctuations in revenue per launch.

  “So I do think, you know, SpaceX's decade plus lead in reusable is not being encroached upon.”at 20:38

• Brett Winton · ARK InvestBullish

  Starlink's incremental satellite capacity will monetize less well than its first capacity.

  Brett argues that each marginal terabit per second SpaceX adds through Starlink will monetize less well than the initial capacity, although there is still room to grow. He points to enterprise connectivity demand, including United, which he expects to have Starlink on 100% of its flights by year-end 2027, as evidence that markets outside consumers can be an important demand driver. He also suggests Musk is pushing Starlink adoption among airlines in a way that hurts rivals such as Delta. His view is that demand is real but returns on new capacity will diminish.

  “Marginal terabits per second they put up will monetize less well than the immediate terabits per second, albeit with still a lot of room to grow.”at 22:06

• Brett Winton · ARK InvestBullish

  SpaceX will likely find the compute endpoint; capital is its only constraint.

  Brett is 'ultra bullish' because Elon Musk is the only person willing to make decade-long plans. Musk is building his own fab because he expects to run out of chips, and buying a turbine maker to secure power on the ground. Brett expects SpaceX to discover the ultimate compute endpoint, possibly orbital compute, and says capital is the only thing holding the company back in his model. SpaceX could raise equity if it is satisfied with its pricing, and accelerate faster than the aggressive numbers his team has modelled for the early 2030s.

  “I'm ultra bullish right now. ... Wherever the endpoint is, SpaceX will almost certainly be the company to discover it because Elon seems to be the only person who's willing to make decade long plans.”at 29:07

OpenAI

Across the shows: 11 speakers in other episodes, 3 bullish and 14 bearish calls, past 30 days

• Brett Winton · ARK InvestBullish

  Ultra-fast inference tier shows enterprises and traders will pay a premium for speed.

  Brett argues speed is becoming its own competitive axis alongside intelligence. OpenAI's ultra-fast tier offers about 8x the tokens for 6x the price per second, and reportedly its fastest capacity is sold to Jane Street at a reported $200 billion per gigawatt, far above what Anthropic pays to rent SpaceX capacity. In high-frequency trading, a marginal speed edge is worth a lot of money, so the firm pays up. Brett expects enterprises more broadly to pay for speed, which he sees as a source of monetization for these labs. He also notes the monetization rate of the off-the-shelf products is unknown, so the size of the opportunity is uncertain.

  “Why would Jane Street do such a thing? It's because when you're in a high frequency trading game, having an ultra intelligent thing that is faster than your competitor is a license to mint money.”at 2:51

• Nick Grous · ARK InvestBearish

  OpenAI will likely favour enterprise compute over consumer, risking consumer lock-in to Meta and Alphabet.

  Nick argues OpenAI faces a trade-off between selling scarce compute to enterprises and serving consumers. If the consumer business cannot be monetized for years, a buyer willing to pay $200,000,000,000 per gigawatt (Jane Street) makes the choice easy. He thinks the long-term question is whether to take near-term revenue or pursue a larger consumer market. If Meta and Alphabet give consumer AI away free, they could lock in users, making it hard for OpenAI to re-enter the consumer space later. His view is bearish on OpenAI's consumer position, while noting this is a major open question.

  “If you're sitting there going, well, we can't monetize a consumer for x amount of years, but Jane Street's willing to pay us $200,000,000,000 for a gigawatt. Like, the choice is very easy.”at 6:15

Robinhood Markets

Across the shows: 1 speaker in other episodes, 1 bullish and 0 bearish calls, past 30 days

• Nick Grous · ARK InvestBullish

  Robinhood's agentic trading tools and social features could be powerful for retail traders.

  Nick praises Robinhood for letting retail users connect their own AI agents (Claude, OpenAI or others) to their accounts to co-develop, backtest and run trading strategies. Robinhood also added data connectors for data that hedge funds use, such as weather data. Traditionally, creating a fund or strategy has been reserved for the elite on Wall Street. Pairing this with Robinhood Social, which is now generally available, could let users publish strategies, have others copy them, and eventually monetize them with fees, in the way fund managers do with ETFs. He sees this democratization of fund strategy as an important development, though the number of people building high-touch strategies is still uncertain.

  “Now Robinhood handing this over to everyday traders, I do think will be very powerful, especially when they pair it with social.”at 12:38

AI infrastructure & data centers

• Brett Winton · ARK Invest

  AI compute is short for years; a later glut could break financing and end the cycle.

  Brett says the evidence is very clear that compute is short today. He expects demand, including productivity-driven demand across large markets, to keep compute constrained over the medium term. Friction points like TSMC's production rate, and the huge funding needs that make financiers less aggressive than the returns justify, slow the buildout. He thinks at least a couple of years must pass before skeptics are proven wrong or demand is fully accepted. He also warns that flooding the market with compute two years ahead of demand would dry up financing, cause bankruptcies and consolidation, and mark the end of the cycle. Consumer AI use cases will likely stay slightly starved for compute for at least the next few years.

  “But all of the evidence right now... It's very, very clear, on an immediate basis where it's short compute.”at 26:34

• Nick Grous · ARK Invest

  AI compute will stay short for about a decade, supporting the market through several bottlenecks.

  Nick says compute will stay short for roughly a decade because the constraint is not one bottleneck but several across the supply chain. Lifting one constraint exposes the next. He also argues the shortage is good for stocks, since demand is growing faster than supply can come online, keeping the supply-demand gap in line. He allows that this could go wrong, but says the current balance supports the market. Sam and Brett add nuance: there may be shorter oversupply pockets, but the long-run view is that compute stays scarce.

  “I think we'll be short of compute for, like, a decade. Like, I think it's gonna take a really long... Because to your point, it's not just one bottleneck.”at 28:12