Saxo Market Call
Saxo Market Call

Nov 13, 2026 · 56 min

AI’s infrastructure boom meets its deflationary reckoning

Nvidia and the AI trade, featuring guest Peter Garnry

The episode tests whether enormous AI investment can earn durable returns as model costs fall, risks concentrate, and physical constraints tighten.

3 key takeaways
  1. 1NVIDIA faces regulatory uncertainty and financial-engineering risks as its ecosystem expands beyond chip sales.
  2. 2Falling model costs and open-weight competition could shift AI’s profits from frontier labs toward specialized applications.
  3. 3AI concentration, cybersecurity demand, Google’s search exposure, and data-center power limits reshape portfolio risk.

Don't miss

Peter Garnry argues that falling AI model costs could redirect value away from frontier laboratories toward specialized models and application-layer businesses.

The brief

John Hardy opens with rising U.S. Treasury yields, a stronger dollar, and a week ahead packed with rate, earnings, and NVIDIA risks.

Peter Garnry examines NVIDIA’s alleged chip-routing violations, Jensen Huang’s responsibility, and the uncertainty facing management and long-term investors.

The conversation widens from NVIDIA’s guarantees and GPU collateral to whether data-center financing can produce adequate returns as infrastructure bottlenecks grow.

Falling token prices and improving open-weight models challenge frontier AI laboratories, potentially moving value toward domain-specific models and application businesses.

Garnry frames AI exposure as a market-wide risk factor, while cybersecurity demand, Google’s search vulnerability, and power constraints complicate the investment case.

The standout tension is simple: AI may become dramatically cheaper to use just as the infrastructure required to deliver it becomes harder and costlier to build.

What was said on this episode

30 statements · 4 positive · 21 negative · 3 mixed · 2 neutral

  1. guest Peter Garnryon Nvidia CorporationNegative12:22

    Knowledge of chip-routing circumvention could constitute a serious U.S. law violation.

    “if someone in NVIDIA, either Jensen Wang or people close to him, knew about this, then it's a very, very serious violation of U.S. laws.”

    Listen at 12:22

  2. guest Peter Garnryon Nvidia CorporationNegative12:42

    NVIDIA will face substantial uncertainty in coming weeks and months.

    “there will be a lot of uncertainty around NVIDIA in the coming weeks and months.”

    Listen at 12:42

  3. guest Peter Garnryon Nvidia CorporationNegative12:53

    Long-term NVIDIA investors face serious questions and risks.

    “for the longer-term investor, I think there's some serious questions and risks you need to consider now.”

    Listen at 12:53

  4. guest Peter Garnryon Nvidia CorporationNeutral14:53

    Some AI-ecosystem companies cannot obtain credit without NVIDIA guarantees or backstops.

    “there are companies in this ecosystem that cannot get credit financing unless you have guarantees or backstops coming from NVIDIA.”

    Listen at 14:53

  5. guest Peter Garnryon NVIDIA business and balance-sheet supportMixed15:12

    NVIDIA may restructure its business and balance-sheet support risks.

    “will we see a rearrangement of how NVIDIA conducts its business and the risks that they're taking on their balance sheet to support all this? That is the fundamental question, I think.”

    Listen at 15:12

  6. guest Peter Garnryon AI data centersNegative17:10

    Data-center construction, grid connection, and GPU deployment are severe bottlenecks.

    “getting data centers built and plug them in and get the GPUs running, that is where the severe bottleneck is.”

    Listen at 17:10

  7. guest Peter Garnryon GPUNegative18:30

    GPUs technologically depreciate quickly, reducing their collateral value.

    “GPUs actually degrade pretty fast in terms of technology and their collateral value”

    Listen at 18:30

  8. guest Peter Garnryon AI compute-rental businessesNegative19:33

    Commodity compute-rental businesses likely cannot earn 15% returns on capital.

    “I don't think you can get 15% return on capital for a commodity business.”

    Listen at 19:33

  9. guest Peter Garnryon AI compute-cloud businessesNegative19:43

    AI compute-cloud businesses will face severe pressure in coming years.

    “these neural clouds are going to be squeezed very hard in the years to come.”

    Listen at 19:43

  10. guest Peter Garnryon GPU rental businessesNegative22:24

    Large-scale GPU rental remains a very low-margin business despite higher prices.

    “it's still very low margin business despite of the rising gpu prices”

    Listen at 22:24

  11. guest Peter Garnryon GPU rental businessesNegative22:47

    Investing in GPU-rental businesses is a dangerous trade.

    “I think it's a dangerous trade if you engage yourself in that from the investor side.”

    Listen at 22:47

  12. guest Peter Garnryon AI model token pricesNegative24:04

    Token prices have depreciated 70%–80% across model generations.

    “the token price depreciation, it's somewhere between 70% and 80% across the different vintages of models.”

    Listen at 24:04

  13. guest Peter Garnryon AI model providersNegative24:16

    AI model providers need continual strong demand growth merely to maintain revenue.

    “you need a very, very, very high demand acceleration constantly to just stand still, basically, in terms of revenue.”

    Listen at 24:16

  14. guest Peter Garnryon Local and open-weight AI modelsPositive26:12

    Local and open-weight AI models will become more prevalent.

    “Going forward, we'll see more local models, more open-weight models.”

    Listen at 26:12

  15. guest Peter Garnryon frontier AI labsNegative26:16

    Open-weight competition will put frontier AI labs under pressure unless they enter applications.

    “it puts these frontier labs in a very difficult position over time unless they can sort of branch into the application layer.”

    Listen at 26:16

  16. guest Peter Garnryon AI risk factorNegative28:36

    AI exposure now affects nearly half of industry groups among largest companies.

    “the AI risk have morphed into almost half of all the industry groups when you look at the very largest companies.”

    Listen at 28:36

  17. guest Peter Garnryon Investment cyclesNegative29:09

    Investment booms eventually end and produce losses afterward.

    “investment cycles always come to an end and there's always some bleeding on the other side of when it ends”

    Listen at 29:09

  18. guest Peter Garnryon Equal-weight indicesPositive30:29

    Investors should consider equal-weight rather than cap-weighted passive indices.

    “I really recommend people to think about things like this. Is cap-weighted index and passive the right way? You can still be passive, but you can do it in an equal weight. I think it's really worth considering.”

    Listen at 30:29

  19. guest Peter Garnryon S&P 500Neutral34:05

    During the AI trade, 150 S&P 500 stocks have negative beta.

    “This time around during this AI trade, we have 150 stocks in that category.”

    Listen at 34:05

  20. guest Peter Garnryon CybersecurityMixed37:43

    AI has permanently changed cybersecurity.

    “cyber security has changed forever because of ai”

    Listen at 37:43

  21. guest Peter Garnryon Cybersecurity industryPositive38:25

    Cybersecurity could grow 10%–20% annually over the next decade.

    “it could be somewhere between 10 and 20 percent compounded growth rate”

    Listen at 38:25

  22. guest Peter Garnryon CybersecurityNegative38:50

    Cybersecurity-industry valuations are currently stretched.

    “the valuations are getting quite stretched, to say the least, in this industry.”

    Listen at 38:50

  23. guest Peter Garnryon Google AdWordsNegative42:22

    The inability to place ads in chatbots will increasingly threaten Google.

    “Nobody has cracked this, how to place ads in our chatbots. And I think increasingly it will be a problem for Google.”

    Listen at 42:22

  24. guest Peter Garnryon Google search business modelNegative43:47

    Edge AI could pose a larger threat to Google’s search business model.

    “that would create an even bigger severity for a bigger threat to the whole Google business model on the search side.”

    Listen at 43:47

  25. guest Peter Garnryon Shopify Inc.Mixed44:34

    Shopify captures over half of current U.S. e-commerce growth while Amazon loses share.

    “more than 50% of the growth right now is actually going to Shopify. So Amazon is actually losing out.”

    Listen at 44:34

  26. guest Peter Garnryon Niche e-commerce sites versus AmazonPositive45:10

    Niche category sites are growing two to four times faster than Amazon.

    “right now you're seeing 2x, 3x, 4x. larger growth in niche category sites versus Amazon.”

    Listen at 45:10

  27. guest Peter Garnryon Large technology companiesNegative45:39

    Large technology companies may be disrupted by AI in unexpected ways.

    “they will be disrupted in very, very... this way that we didn't think imaginable just four years ago.”

    Listen at 45:39

  28. guest Peter Garnryon SoftBank Group Corp.Negative47:56

    SoftBank’s write-down evidences infrastructure as the AI bottleneck.

    “the softbank write down was just a very concrete evidence where the bottleneck is”

    Listen at 47:56

  29. guest Peter Garnryon AI data-center expansionNegative49:17

    Permitting, environmental, and community requirements will further slow data-center expansion.

    “I think we'll see a different dynamic and that will slow down things even further.”

    Listen at 49:17

  30. guest Peter Garnryon SoftBank Group Corp.Negative49:25

    Further AI infrastructure delays pose a key risk to SoftBank investors.

    “how many dollars are on the line for SoftBank if things slow down even more.”

    Listen at 49:25

Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.

Books & mentions

Some links are affiliate links — PodLume may earn a commission if you buy.

Weekly email

State of the Pod, every Monday

The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.

Listen to the full episode and explore every guest, topic, and moment on PodLume.