ARK Invest (FYI podcast, In the Know video) · · 29 min

The AI Shakeup Has A Message | The Brainstorm 158

ChorusBring the market's conversation to your AI. Learn more

Summary

In this ARK Invest episode, Brett Winton and Nick Grous discuss how Jensen Huang's departure as Nvidia's central figure could reshape the AI and semiconductor landscape. Brett is bullish on SpaceX, arguing the successful Starship upper-stage landing and faster refurbishment could cut launch costs toward $100/kg, making orbital data centers viable, while he is bearish on Nvidia's share of the cutting edge as its balance-sheet-driven ecosystem funding (neoclouds, open-weight developers) loses its backer under tighter US export rules. Brett also argues AI spending will follow Jevons paradox, with cheaper intelligence driving higher total spend, countering Salesforce-style skepticism about Anthropic's slowing revenue onboarding. Nick is more cautious, expecting a conservative new CFO to slow Nvidia's chip orders, cuts in capex forecasts (Semianalysis reportedly down about 28%), and willingness to pay for AI to gate growth longer than the market expects. Both agree the next two quarters of earnings, and GPU rental pricing, will show which view holds.

Key takeaways

  • Brett Winton (bullish on Space Exploration Technologies): A reusable Starship upper stage could cut SpaceX launch cost toward $100/kg from $600-700/kg on Falcon 9.
  • Brett Winton (bullish on Space Exploration Technologies): Launch costs below $100/kg make orbital data centers viable, and SpaceX has cornered scaled space compute.
  • Brett Winton (bearish on NVIDIA): Huang's exit weakens Nvidia's ecosystem funding and its share of the cutting edge over time.
  • Nick Grous (bearish on NVIDIA): A more conservative CFO would cut Nvidia's large chip orders, quickly changing its market-share outlook.
  • Nick Grous on Enterprise AI adoption & AI software: Companies will decide what they'll actually pay for AI; AI spend can't outgrow revenue for long.

What was said, by company and theme

9 views from 2 speakers. Every quote is checked against the episode's transcript (how).

Space Exploration Technologies

Across the shows: 55 speakers in other episodes, 64 bullish and 27 bearish calls, past 30 days

Where everyone stands on Space Exploration Technologies

  • Brett Winton · ARK InvestBullish

    A reusable Starship upper stage could cut SpaceX launch cost toward $100/kg from $600-700/kg on Falcon 9.

    The upper stage landing was historic, roughly three times the Space Shuttle's payload by volume. Reusability was not assumed to be possible. Brett's team thinks a reusable top stage could lower SpaceX's cost per launch to around $100/kg from roughly $600-700/kg on Falcon 9, and about $500/kg even without reuse. The ship returned in much better shape than ARK expected. Musk said it could be refurbished in weeks rather than the roughly 90 days ARK had modelled, and ARK had expected six commercial launches before a top-stage landing would succeed. Lower blended launch cost reduces the capital SpaceX must commit up front to launch satellites that generate revenue. Faster refurbishment also accelerates data gathering and iteration.

    “We think that with a reusable top stage, which we weren't sure was actually going to be possible, they could lower their cost per launch to around $100 per kilogram from... Right now, we think on the Falcon nine, they're at, you know, 6 to $700 per kilogram.”at 1:19
  • Brett Winton · ARK InvestBullish

    Launch costs below $100/kg make orbital data centers viable, and SpaceX has cornered scaled space compute.

    Brett's team had previously argued that data centers in orbit begin to pencil once launch cost falls to about $300-400/kg. Other inputs still matter, such as satellite manufacturing and chip costs. Once launch cost drops to $100/kg or below, launch cost falls out of the equation, and launch vehicle availability is no longer the constraint on space compute. SpaceX has built a manufacturing factory for Starship at a scale nobody has achieved in rocketry. Reusing those vehicles should make its launch inventory grow sharply. Compared with rivals building data centers on the ground, which must identify sites and negotiate with municipalities, SpaceX looks well placed to deploy compute at scale quickly. Brett says the market is less certain that demand for compute is huge.

    “then kind of like SpaceX definitely has cornered the market on being able to do compute at scale. As in, it's highly likely that they'll be able to do a lot more and continue to grow it relative to themselves and others trying to operate on the ground”at 9:13

NVIDIA

Across the shows: 174 speakers in other episodes, 184 bullish and 164 bearish calls, past 30 days

Where everyone stands on NVIDIA

  • Brett Winton · ARK InvestBearish

    Huang's exit weakens Nvidia's ecosystem funding and its share of the cutting edge over time.

    Huang used Nvidia's balance sheet aggressively to create demand for its chips across the stack. He seeded neoclouds that buy accelerators, invested in open-weight model developers as a check on closed frontier labs, and funded biology startups as a possible source of chip demand. A more measured CFO would likely still make these strategically rational moves, but at a slower pace. The US government reportedly pushed him out over concern that Chinese firms could reach next-generation chips through cloud access. That tightens a channel that let Nvidia sell chips to willing buyers who then built open-weight models. Less venture funding, slower neocloud financing, and a weaker supply-chain alignment built around Nvidia all make its share of the cutting edge more likely to shrink over the medium term. Google and AMD are positioning as more reliable packaging partners, and US buyers such as SpaceX may gain leverage.

    “it probably does make it more likely over the medium term that NVIDIA's share of kind of, like, the cutting edge diminishes.”at 17:18
  • Nick Grous · ARK InvestBearish

    A more conservative CFO would cut Nvidia's large chip orders, quickly changing its market-share outlook.

    Huang's approach was to tell TSMC to build very large volumes so Nvidia would hold about 80% of the market. A CFO who says Nvidia only needs a smaller amount because of what they are seeing would change that picture very quickly. The new CFO appears likely to take a more conservative approach to AI spending than Huang did. Nick thinks this could change the outlook within a short time, though he is uncertain how the market will price it.

    “and saying, here's what we need, and, you know, making it a huge number so that they remain 80% of the market. Now, if you have a CFO step in and say, well, we actually only need this amount”at 19:16

AI infrastructure & data centers

Where everyone stands on AI infrastructure & data centers

  • Nick Grous · ARK Invest

    Uncertainty after Huang's exit will delay data center spend and how companies approach AI.

    Nick thinks Huang's aggressive investment approach was what pushed AI scaling forward, and that the reflex from losing that central figure will be uncertainty. As evidence, Semianalysis reportedly cut its CapEx spend forecasts by about 28%, which could flow through to industry ARR. He sees a more conservative CFO taking a different stance on AI. The market is waiting for earnings to show how much uncertainty will delay data center spending and how people approach AI more broadly.

    “me, when I look at this, I say, you know, there's gonna be some uncertainty in the market, and that is gonna cause, you know, delays in not just, you know, spend for data centers, but how people approach AI generally.”at 12:05

Open-source AI models

Where everyone stands on Open-source AI models

  • Brett Winton · ARK Invest

    Huang's exit removes a funder of open-weight models that checks frontier labs; Chinese labs may fall behind.

    Huang invested heavily in enabling open-source and open-weight models, which give a competitive layer against core frontier labs such as OpenAI and Anthropic. That investment came partly through infrastructure access to chips for Chinese developers. With Huang gone and US restrictions limiting Chinese access to next-generation chips, Brett expects Chinese companies to struggle to keep open-weight models close to the frontier. Possible offsets are US companies. OpenAI already has an open-weight model, and Meta may see an opening to return to open-weight, though it has pulled back to some degree. Without Huang's push, the open-weight competitive layer loses its main backer.

    “and he was investing heavily in enabling open source or open weight models because that provides a competitive kind of layer against those core frontier labs at that part of the chip stack”at 14:07

AI chips & semiconductors

Where everyone stands on AI chips & semiconductors

  • Nick Grous · ARK Invest

    Chips no longer bound for China should stay in the US, so GPU rental prices will reveal true demand.

    With the export channel to China restricted, chips that were headed there should be sold to willing buyers in the US. That should give many companies the allocation they had been seeking. Nick thinks this will produce a faster answer on true GPU demand. Outside possible power bottlenecks, more capacity should come online in the US. GPU rental pricing is the best indicator to watch. If rental prices hold up, that shows true demand. In a period of scarcity, buyers over-order, and the market will find out where true demand sits once the main figurehead is no longer steering orders.

    “You might get a faster answer into what, like, true demand is there for GPUs. I think one thing to watch is GPU rental pricing.”at 18:13

AI pricing & unit economics

Where everyone stands on AI pricing & unit economics

  • Brett Winton · ARK Invest

    Cheaper AI intelligence drives more total spend, so industry revenue reaches high single-digit trillions by 2030.

    Brett uses ramp data on an intelligence-adjusted, blended token basis. The cost of intelligence to these firms fell about 97% year over year. Across that data, firms spend more in aggregate as intelligence gets cheaper, with roughly a 1.2% rise in total spend for every 1% fall in intelligence cost. That is Jevons paradox: cheaper capability means more uses and more spending. Skeptics point to Anthropic's revenue onboarding slowing from Q2 to Q3. Brett rejects a Salesforce-style model in which each sales cohort lands smaller revenue, because AI sells on usefulness rather than utilisation. Under the Salesforce model, industry spend would be less than $1 trillion by 2030. Revenue per unit of compute has improved as models deliver more intelligence per unit of compute, and with the planned compute coming online, that supports high single-digit trillions in revenue by 2030. Brett expects Q4 results and commentary to show which view is right.

    “These firms tend to spend more in aggregate, not less blended across... for every 1% drop in basically intelligence costs, you end up with a 1.2% increase in the total spent.”at 21:22

Enterprise AI adoption & AI software

Where everyone stands on Enterprise AI adoption & AI software

  • Nick Grous · ARK Invest

    Companies will decide what they'll actually pay for AI; AI spend can't outgrow revenue for long.

    Nick says AI diffused extremely quickly into intuitive use cases such as coding, where adoption was fast. The next phase is whether companies will pay top dollar. They will keep using AI across marketing, legal and other functions, but the question is how much they will actually pay. AI spend cannot keep outgrowing revenue growth indefinitely, so companies will offset AI spend with cuts elsewhere. Nick thinks that gating factor will take longer than the market expects, with bumps along the way. He expects the next two quarters to show whether his cautious view or Brett's Jevons view is right, and he is more cautious than Brett on ARR and demand over the next year.

    “you're going to figure out whether or not it's actually worth to spend money on it. As in, you may still be using across your marketing, and legal, and x, y, and z, but how much are you actually willing to pay for it is the big question.”at 25:00

Across the shows

What other episodes Chorus covered said about the same companies in the 30 days up to this one.

Ask your AI who they are and what they said

Topics

  • GPU demand visibility
  • Jevons paradox in AI spend
  • chip market share
  • chip orders
  • data center capex
  • launch cost per kilogram
  • open-weight model development
  • orbital data centers
  • willingness to pay for AI

More from ARK Invest (FYI podcast, In the Know video)

All ARK Invest (FYI podcast, In the Know video) episodes

Ask your AI about this episode

Chorus gives Claude, ChatGPT and other assistants every view from this episode and every other show Chorus follows. Ask who agrees, who changed their mind, and what else was said about the same companies.

A weekly email with the calls that mattered from the most important voices in investing. Unsubscribe any time. Privacy Policy

Chorus reports what speakers said. It isn't investment advice. How Chorus works · About Chorus