Marathon Asset Management CEO Bruce Richards Talks Anticipated Anthropic IPO

Anthropic

Across the shows: 47 speakers in other episodes, 27 bullish and 38 bearish calls, past 30 days

• Bruce Richards · CVC MarathonBullish

  Anthropic's IPO will succeed, since it is a fast-growing company that public markets can fund like other trillion-dollar companies.

  Richards says Anthropic will do just fine in its IPO, pointing to SpaceX's trillion-dollar listing as precedent. He calls Anthropic a solid company with rapid growth and a strong revenue trajectory, so it can finance itself and the capital markets will fund it as they funded other trillion-dollar companies. He is generally constructive on Anthropic, while noting the IPO is roughly a $100 billion raise.

  “Anthropic will do just fine. The last time we saw a trillion dollar market cap company come to market, referring to SpaceX, did just fine.”at 3:00

• Bruce Richards · CVC MarathonBullish

  Anthropic will capture most AI model profits because closed-weight models earn the profits, as in software.

  Richards argues the open versus closed weight debate plays out as it did in software. Android and Linux are successful open platforms, but profits accrue to closed providers such as Apple and Microsoft. He expects Anthropic to capture a lot of profitability from its closed model, and says open and closed models will coexist. He backs this with a broad shortage of compute and strong demand for AI.

  “profits accrue to the close weights or the close models, and so Apple, Microsoft, and in this case Anthropic, I think, is going to capture a lot of profitability in their model in general.”at 3:50

CoreWeave

Across the shows: 48 speakers in other episodes, 12 bullish and 43 bearish calls, past 30 days

• Bruce Richards · CVC MarathonBearish

  CoreWeave's unsecured debt is too risky at about 15% yield and 1,100bp CDS; lend only on a secured basis.

  Richards calls CoreWeave the biggest neocloud, with CDS trading around 1,100 basis points and unsecured bonds yielding about 15%, versus where its secured debt trades. The market will fund it only on a secured basis. Its market cap is just under $30 billion, which he says is not enough for the debt needed for gigawatt builds. Neoclouds issued unsecured debt earlier in the year, but it now trades wide, so the market has stopped lending to them on an unsecured basis. He says it is good to be secured and not so good to be unsecured in this space.

  “You look at the biggest [neocloud] of them all and you see their CDS trading at 1,100 spread, and you see their unsecured trading at a 15% yield versus where the secured debt trades.”at 6:24

AI infrastructure & data centers

• Bruce Richards · CVC Marathon

  Most neoclouds won't survive as independents by 2030; only the strongest will, so unsecured neocloud debt is risky.

  Richards cites a Gartner report that 80% of neoclouds will not exist as independent companies by 2030. They will be absorbed by stronger players needing their compute and assets, or disappear, so only the top 20% will make it. He compares this to an 80/20 rule. Because the market only lends to neoclouds on a secured basis, he advises secured lending. He agrees with Elon Musk that debt markets have shut for companies without substantial revenue outside AI data centers. He warns some neoclouds may stop being businesses because they cannot raise debt.

  “[Gartner] Group is a very reputable technology consulting firm... where they say that 80% of the neo clouds by the year 2030 won't be able to exist as independent companies”at 5:38

• Bruce Richards · CVC Marathon

  Richards would avoid financing data center construction, where banks hold unpriced construction risk and projects can fail.

  Richards says data centers are hard to permit and power, given regulatory approval, land, water and power (grid or behind-the-meter) requirements. Amazon wrote off several hundred million dollars in Loudoun County, Virginia, after failing to secure power. He argues banks hold much of the construction risk for data centers still in flux, and that this exposure is under-discussed. He cites a Mexico project with a $10 to $20 billion delayed-draw facility, partly tapped on a failing project, whose loans trade at 70s to 80s. He expects more of this to unravel and possible bank write-offs.

  “I'd say data centers I'd lean away from in terms of financing, certainly with respect to the construction of it.”at 8:19

• Bruce Richards · CVC Marathon

  GPU-backed secured financing is a good play, especially with investment-grade offtake and full principal repayment.

  Richards favours secured lending against AI picks and shovels such as GPUs, because the lender holds collateral. He says GPU financing with investment-grade offtake should return all principal plus a nice multiple over the loan term. This assumes zero residual value after four to five years. He contrasts this with unsecured neocloud debt, which he avoids.

  “Hyperscalers make sure you're secured, super secured with good assets, and kind of the GPU financing, probably a pretty good play, especially if you have IG off take and you're going to make a nice return and get all your principal back.”at 8:35

• Bruce Richards · CVC Marathon

  AI infrastructure spending will reach about $30 trillion over the next decade or more, driven by compute demand.

  Richards builds the number from 500 gigawatts at $60 billion per gigawatt, which gives $30 trillion over a decade or more. He calls the 500 gigawatts ridiculously high. He says the spend is based on the demand needed to fully build out AI, and that the demand for compute is there because AI is the most important technological advance. He cites Morgan Stanley's estimate of $10 trillion of US spend over the next four years, and a huge shortage of compute. He compares $30 trillion to a $32 trillion US GDP and calls it exponential growth for the industry.

  “So I think in the next ten to fifteen years, the spend is actually based upon the demand that's required to fully build out AI close to 30,000,000,000,000.”at 10:16

Private credit

• Bruce Richards · CVC Marathon

  Credit is cheap at wider spreads; opportunistic lending will be the best vintage in a generation.

  Richards says investors can move up in credit at much wider spreads and higher base yields than a few months ago. High yield started the year yielding about 7%, and a performing portfolio now yields 8.5% to 9%. Marathon lends to growth companies growing revenue 8 to 15% a year with protected margins, and to companies transitioning through difficult periods as credit is rationed. He expects the best vintage for opportunistic credit in a generation because of the capital structure fixes corporate America, Europe and the globe need. Marathon is already doing this for clients.

  “I think this is going to be the best vintage for opportunistic credit in a generation because of all the capital structure fixes and the capital structure needs that corporate America and the Europeans and the globe needs.”at 15:27

• Bruce Richards · CVC Marathon

  Private lenders extending maturities and waiving covenants is effectively default, lowering long-run recoveries.

  Richards argues that when a company's financials have deteriorated, extending its maturity is effectively default, especially when the lender is the sole or small-group lender. Deferring interest or waiving covenants, which he calls amend and pretend, delays a hard default. For most companies it lowers the terminal recovery rate. He says this happens more in private credit, where lenders control the deal, than in the broadly syndicated loan and high yield markets. The 22% of BDC loans maturing within two and a half years, about $117 billion, adds to this risk as floating-rate coverage gets harder.

  “as far as I'm concerned when the company's financials have deteriorated and you're extending a maturity, and you're the sole lender or group lender in a very small group, that's effectively default”at 16:59

AI: overall outlook

• Bruce Richards · CVC Marathon

  AI is inflationary in the short run from spending, but deflationary over the long term.

  Richards says the inflationary effect of AI spending is short-run. Big AI spending adds to GDP and recirculates money through the economy, which creates a bump. Over the long term he thinks AI is deflationary. He does not give a numerical basis for the long-run view in this segment.

  “The inflationary part of it is the short run. Long term, I think AI is deflationary.”at 9:40

Credit markets

• Bruce Richards · CVC Marathon

  CCC credit stress is severe and will drive a step-up in defaults over the coming year.

  Richards says BB high yield is fine, with spreads near 226 basis points and tighter than average. CCC yields are about 18%, with spreads above 1,000 basis points. The BB-CCC gap has been this wide only in the GFC and the early months of COVID, even though GDP grows about 3% and S&P earnings grow 25 to 30%. He says tighter financial conditions are hitting less creditworthy companies across high yield, broadly syndicated loans and direct lending. He expects a big step-up in defaults and painful restructurings in the coming year.

  “And so you're going to see a big step function higher in default rates in the coming year as the markets have tightened financial conditions for companies that are less credit worthy.”at 13:52

US rates & the Fed

• Bruce Richards · CVC Marathon

  The Fed will hold this week but most likely tighten in December, raising floating-rate borrowing costs.

  Richards expects the Fed to hold at this week's meeting but most likely tighten in December. The front-end rate, which floating-rate assets are based on, will rise. This makes interest coverage harder for companies on the margin, especially those with floating-rate debt maturing into a higher-rate environment. This is a near-term risk for weaker borrowers.

  “they're certainly gonna hold, but they're gonna tighten most likely in December. And so the rate trend at the front end, which all these floating rate assets are based off, will go up”at 16:16