Bloomberg Surveillance · · 40 min

AI Stocks Rebound as Fed Hike Bets Return

With Amanda Lynam, David Malpass, Ed Hirs, Gautam Mukunda

Summary

On Bloomberg Surveillance, Amanda Lynam argued that hyperscaler AI debt is becoming a paradigm shift for credit, with jumbo, correlated financings arriving in quick succession and likely to keep widening spreads, while dismissing CCC stress as mostly company-specific and high yield as larger and higher quality than before. Ed Hirs described ERCOT's price-minimizing, scarcity-rewarding market as ill-suited to data center demand, expected data centers to bring their own power, pointed to solar, wind and batteries as the fast-growing supply, and called refining a spectacular business amid global capacity shortages. Gautam Mukunda was bearish on Nvidia over export-control evasion and lobbying, saying more revelations and investigations are likely, and bullish on Anthropic's compute efficiency and cost leadership. David Malpass warned that US supply chains cannot easily decouple from China and urged permitting reform to expand power supply.

Key takeaways

  • David Malpass on Power demand & the grid: US needs far more electricity generation, but permitting is blocking it and electricity prices are rising.
  • Amanda Lynam on AI infrastructure & data centers: Debt-funded AI buildout is a credit paradigm shift that needs all markets and keeps widening spreads.
  • Ed Hirs on Power demand & the grid: ERCOT's market design rewards scarcity over reliability, and data centers are driving up wholesale power prices.
  • Gautam Mukunda (bearish on NVIDIA): Nvidia faces serious legal and regulatory fallout from alleged chip evasion to China, not a minor scandal.
  • Gautam Mukunda (bullish on Anthropic): Anthropic uses its internal models to dominate rivals on cost-performance, not just push the frontier; massively undercovered.

What was said, by company and theme

15 views from 4 speakers. Every quote is checked against the episode's transcript (how).

Anthropic

Across the shows: 112 speakers in other episodes, 152 bullish and 91 bearish calls, past 30 days

  • Gautam Mukunda · Yale School of ManagementNeutral

    Anthropic's IPO priority is cash to cover huge compute commitments, not valuation or first-day pop.

    Mukunda says founders' control makes them largely immune to stockholder pressure, so valuation and the first-day pop matter less to them. What matters most is building a cash reserve as a cushion against the very large compute commitments the company has made. He notes that its funding strategy has been rounds with money kept in the bank.

    “what is most important to them is getting enough cash reserve that they can have a cushion against the gigantic commitments for compute that they've made.”at 37:17
  • Gautam Mukunda · Yale School of ManagementBullish

    Anthropic leads on compute efficiency and cost-performance, an under-covered story versus rivals.

    Mukunda says Anthropic has been raising rounds and banking the money, and seems much more computationally efficient than everyone else. He argues it is using its internal models, which it believes are better than rivals', not just to push the performance frontier. It is using them to be Pareto dominant at every performance level on cost, even against open models. He calls this a massively under-covered story.

    “They seem to be much more computationally efficient than everyone else. And this is a really interesting, I think massively undercovered story with Anthropic... what they've been doing is using their internal models... to make themselves Pareto dominant at all levels of performance across cost, even when compared to the open models.”at 37:25
  • Gautam Mukunda · Yale School of ManagementBullish

    Anthropic uses its internal models to dominate rivals on cost-performance, not just push the frontier; massively undercovered.

    Mukunda argues that over the last few years Anthropic has raised funding in rounds and kept the money in the bank rather than burning it, and that it is far more computationally efficient than its rivals. His central claim is that Anthropic uses its internal models, which it believes are better than everyone else's, not only to push the performance frontier forward but to make itself Pareto dominant at every performance level on cost, even against open models. He calls this a massively undercovered story.

    “using their internal models, which they believe are better than everybody else's. Not just to push the performance frontier forward. But to make themselves Pareto dominant at all levels of performance across cost, even when compared to the open models”at 37:40

NVIDIA

Across the shows: 103 speakers in other episodes, 86 bullish and 57 bearish calls, past 30 days

  • Gautam Mukunda · Yale School of ManagementBearish

    Nvidia faces serious legal and regulatory fallout from alleged chip evasion to China, not a minor scandal.

    Mukunda separates three scandals. The least serious is Huang's relationship with an alleged Chinese intelligence agent, which American intelligence knew about without alerting the company. More serious is Nvidia's lobbying and a $25 million donation to the Trump ballroom, which he says aimed to get the administration to look away from export controls. The prohibited-entity list has not been updated in over a year. Third, reports suggest up to 20% of Nvidia's sales went to China through evasion, which he says implies knowledge was pervasive inside the company. He expects more revelations, Democratic congressional investigations, possible prosecutions and a company full of lawyers for a long time.

    “where there are accusations that could be as much as 20% of all NVIDIA sales suggests this isn't just a thing of like, oh, one or two people were kinda making some sales to make their quota.”at 31:04

Power demand & the grid

  • David Malpass · American Global Strategies

    US needs far more electricity generation, but permitting is blocking it and electricity prices are rising.

    Malpass cites the latest inflation data, which showed electricity prices rising. He argues the country needs a lot more energy focused on making electricity. Permitting has been a real obstacle to that buildout. He suggests a reconciliation bill could pass permitting reform by year end, since it can get through with a simple majority.

    “And what the country needs is lots more energy focused on making electricity, but the permitting has been a real problem for that. So maybe they could get that done here by year end.”at 5:28
  • Ed Hirs · University of Houston

    ERCOT's market design rewards scarcity over reliability, and data centers are driving up wholesale power prices.

    Hirs says ERCOT's minute-by-minute price-minimizing model does not look ahead, and its market rewards scarcity rather than reliability. As the market tightens, prices spike. Tens of billions of dollars have gone to bonuses and buybacks rather than new power plants. After the 2021 freeze, the state needs to rebuild the market. Data centers already in Texas have pushed up wholesale electricity prices for everyone, and the state is still debating whether to scrap the market.

    “And it's a market that rewards scarcity, not reliability. And so as the market gets really tight, the price goes way up.”at 19:51
  • Ed Hirs · University of Houston

    Gas turbine lead times of 3–5 years make self-powered gas plants for data centers hard within two years.

    Asked whether data centers will need their own power plants within a year or two, Hirs says gas-fired plants will be very difficult to build in that window. There is a three-to-five-year lag to line up new turbines. Some entrepreneurs are repurposing jet engines from the desert boneyard in Arizona to fill the gap. He expects data centers to bring their own power plants, and points to solar, wind and batteries as the fast-growing supply.

    “It's gonna be very difficult if they're using gas fired power plants because there's a... Now a three to five year lag time in lining up the new turbines.”at 22:00

US–China relations

  • David Malpass · American Global Strategies

    Reshoring supply chains from China is a hard financial and engineering challenge given China's scale.

    Malpass argues that China is fully embedded in world supply chains and will be for a long time. The US wants deeper domestic supply chains, but China has economies of scale in making these goods. He calls building those US supply chains a real financial and engineering challenge. He hopes the Xi–Trump summit focuses on stability in the relationship rather than on China's aggression in the South China Sea and the Arctic.

    “China's fully in a part of the world supply chains. So for, I think quite a bit of time into the future now, there's got to be some trade off of we would like to get deeper supply chains in The US at a time when China has an economy of scale making this stuff. So that's a real financial and engineering challenge for The US.”at 6:02
  • Gautam Mukunda · Yale School of Management

    Chinese espionage likely won't derail the Xi–Trump summit absent something egregious.

    Mukunda does not expect the Nvidia espionage issue to be front and center at the November summit. In a normal administration, spying is simply part of the game, and both sides spy on each other. Everybody knows it is already happening, so absent something really egregious it will not disrupt the negotiations.

    “So I actually don't expect that this would be front and center... absent something really egregious, it's not gonna disrupt these negotiations because everybody knows it's already happening.”at 34:57

Credit markets

  • Amanda Lynam · Goldman Sachs

    High yield is larger, more diverse and higher quality than before the pandemic and financial crisis.

    Lynam says high yield spreads at 320 basis points are tight in absolute terms, but high yield is trading at a larger discount to investment grade than its ratio would suggest. She argues the bigger picture is that the high yield market is much larger, more diverse and higher quality than before the pandemic or the financial crisis. This year's average high yield bond deal has been above $1 billion, so issuers are large, scaled companies.

    “the high yield market is just larger, more diverse and higher quality than even pre pandemic and certainly pre financial crisis.”at 14:17
  • Amanda Lynam · Goldman Sachs

    Triple-C credit stress is idiosyncratic and adverse selection, not a sign of broader credit weakness.

    Lynam says she does not see the CCC cohort as representative of the broader credit backdrop. CCCs have been widening since May and their interest coverage was about 0.8x in June, so they lacked earnings to cover interest. Rates have sold off since, making it harder to grow into those capital structures. Triple-Cs are about 10% of high yield, and most of their problems are company-specific. Some firms have graduated to single-B, so what remains is adverse selection, the left tail of laggards that never grew into their capital structures.

    “I do not view the triple c cohort as emblematic of kind of broader credit backdrops. In fact, triple c's have been widening since May.”at 14:57

AI infrastructure & data centers

  • Amanda Lynam · Goldman Sachs

    Debt-funded AI buildout is a credit paradigm shift that needs all markets and keeps widening spreads.

    Jumbo investment-grade deals usually come for M&A and are one-and-done, with companies then deleveraging. Hyperscaler AI financings are arriving in quick succession, all correlated to the AI theme, which makes them a paradigm shift for credit. Lynam expects the buildout to need a range of markets, structures and currencies, an all-hands-on-deck effort. She expects ongoing spread widening so the debt can clear at a higher discount. She adds that CoreWeave's bullish earnings helped sentiment, and that as long as the unlevered IRR is accretive versus a higher cost of capital, the trend keeps momentum.

    “Generally, we see these jumbo financings in the IG credit market, it's for M&A. It's kind of a one and done situation... the manner and the scale and the concentration of this is a bit of a paradigm shift for the credit market.”at 11:49

Renewables & clean energy

  • Ed Hirs · University of Houston

    Solar and batteries are expanding fast and driving Texas electricity growth.

    Hirs says solar is expanding dramatically and rapidly, and batteries are fabulous. ERCOT now has almost 40 gigawatts of solar farms, almost 40 gigawatts of wind farms and 21 gigawatts of short-term battery storage. He says this is what is driving growth in the Texas electricity market.

    “Solar is expanding dramatically and rapidly, and batteries are fabulous.”at 22:23

Oil & fuels

  • Ed Hirs · University of Houston

    Refining tightness keeps the oil patch and refiners strong; oil would not fall below $70.75 even with peace.

    Hirs says there is a global refining capacity shortage, though not in the US, which exports about 1.5 million barrels a day of diesel. Europe has shut many refineries, Asia is not getting enough crude, and Iran has taken competitors' refineries offline. He calls it a spectacular time for refiners. Even if peace breaks out and the strait reopens, he expects oil to drop significantly but not below $70.75. He is watching the continued constriction in refined products, and he fears a recession could eventually push commodity prices down.

    “This is a great time for the oil patch and a spectacular time for refiners. ... even if peace breaks out today and the street is reopened, the price of oil will drop significantly, but not below $70.75.”at 24:38

US–China AI & chip race

  • Gautam Mukunda · Yale School of Management

    If chip smuggling is real and export controls are enforced, China's AI progress faces major economic setbacks.

    Mukunda notes speculation that China's AI industry was built on domestic chips and did not need American compute. If the reported evasion is true, that story was a smokescreen by Chinese intelligence to hide its success at getting high-end chips from Nvidia and other companies. If that is true and the US enforces export controls seriously, it would have significant economic implications for China.

    “If that is true, and The United States now actually decides to take export control seriously, that's going to have some pretty significant economic implications for the Chinese.”at 35:56

Across the shows

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

  • Anthropic112 speakers · 152 bullish, 91 bearish calls
  • NVIDIA103 speakers · 86 bullish, 57 bearish calls

Topics

  • Anthropic IPO
  • CCC high yield risk
  • ERCOT market design
  • Xi–Trump summit
  • chip smuggling and China's AI progress
  • compute efficiency and cost leadership
  • data center power supply
  • export control evasion
  • high yield market quality
  • hyperscaler debt financing
  • model cost-performance efficiency
  • permitting reform and power buildout
  • refining margins and oil price floor
  • solar and battery buildout
  • supply chain decoupling

More from Bloomberg Surveillance

All Bloomberg Surveillance 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.

Setup guide

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