Ex sell-side research analyst/ ex event-driven hedge fund guy. All commentary is my opinion and does not constitute investment advice or a recommendation.
Reposting my take on the potential for a US govt AI capex "backstop" following speculation that the Trump admin is discussing a stake in @OpenAI $NVDA $CBRS $CRWV $NBIS $GOOGL
Thinking about the bear case for AI capex and this @PegasusFund thread below sparked a Munger-style “invert, always invert” moment: Could there be a US govt level “put” on AI capex if the current administration sees AI as an existential “Cold War” (or greater) conflict vs China and uses fiscal and regulatory policy along with pressure on monetary policy to sustain AI capex no matter what the ROIC? Potentially a “run it hot at any cost” approach to fiscal/monetary policy if it stabilizes and supports AI capex? The fiscal/monetary/defense spending regime of the 1960s and the ensuing 1970s inflation crisis might even be a decent if imperfect analogue.
What could this fiscal/monetary regime look like? Export controls/subsidies have long been a policy tool in the semi space and the tension between strategic restrictive controls on $NVDA or other chips versus the need to buttress private sector cashflow/balance sheets will probably be a key dynamic. Additional features might be:
1) Heavy executive branch influence to loosen monetary policy or let real rates drift lower to support private sector credit for datacenter/GPU capex (potentially reigniting inflation). Trump is clearly willing to intervene and push on monetary policy.
2) Tilting fiscal policy towards large, supportive govt contracts in intelligence/defense especially if there is slack compute that might derail the capex cycle (probably also benefits $PLTR $LDOS et al)
3) Additional strong fiscal incentives for AI capex (grants or tax credits, maybe even more favorable depreciation treatment)
4) Further expansion in regulatory scrutiny/export controls as above
5) Major fiscal support for power generation (nuclear, etc., even willingness to relax environmental controls for strategic AI goals)
To me, there’s no obvious finish line for AI for China or the US so, assuming no major political change in the US, the end results could easily be 1) reaccelerating inflation 2) further upward pressure on long-term yields due to fiscal expansion 3) potentially less risk to AI-related equities in the near-term. The 1961-1966 period is an interesting analogy as the US had increased defense and strategic spending (Cold War, Vietnam, space race) along with a Fed that was acknowledging strategic national goals (LBJ pressured the Fed not to tighten, Fed chair acknowledging that “monetary policy that must adapt itself to the hard facts of the budget” in 1965. Whether 1970s style inflation would ensue under this scenario is an interesting question. $NVDA $AMD $GOOG $QQQ $ORCL $MSFT
@invertedfragil1 I don’t think it has a major impact on earnings ultimately but historically there is usually a bit of a bid for lodging names in the run up to a World Cup. Plus sector rotation we saw today
Without seeing the specifics of the Citi note it's hard to say how they're defining "frothy" currently but I would still say equities were "frothy" but not bubble territory in 2007/early 2008 because the bubble in credit meant financials (the largest weight of the S&P500 at that time) were massively over-earning. Valuation wise, It was more an issue of the "E" being overstated than the "P"--and then we had huge collapse in those earnings.
Definitely true. I was there as a sell-side tech analyst at a bulge bracket bank in 2000. I'd say:
1) conversations invoking Dutch tulips or the South Sea Co were routine and frequent
2) buy side institutions had a mix of skeptics, pragmatists and a few "true believers" but by and large the "mode" was an analyst/PM who was pragmatically resigned to being exposed without taking excessive risk on the more toxic sludge...and certainly not one to take career risk by calling a top.
3) A few hedge funds were early to being short...their biggest hunting grounds were the companies that had excessive or mispriced leverage risk (the mkt actually forgot that some companies had big converts issued at the top that were just assumed to convert to equity (stocks go up right?). Some sell-side analysts literally didn't even include these hybrids in their models as they were so equity focussed.
4) Unsurprisingly the biggest cheerleaders lacking skepticism (at least outwardly) were those in the origination pipeline (investment bankers, equity capital markets, etc). There were a few cheerleader sell-side analysts too but skeptical sell-siders probably felt they could never issue a warning (career risk). Most of the sell-side analyst cheerleaders were gone by 2001 as the majority knew nothing about actually analyzing and valuing companies IMHO.
5) The backlash that happened against sell-side equity research (Spitzer settlement) etc after the bubble burst was understandable but anyone can look around today and see why there was reluctance to issue a warning (get called a "doomer" etc). Imagine the abuse an analyst might get for calling a top or putting a "Sell" on AI complex darlings right now. It was a rational strategy for most of the sell-side to remain complacent.
The abuse of the term "alpha" on this app is never-ending. All the time I see the furu posts: "I delivered 200% in alpha today." No, you were long calls on a single stock with a beta of 2.5 on a day the index went up 2%...that's levered beta not alpha. Show me your excess return after you hedged your calls with a delta-equivalent, beta-adjusted amount of SPX puts (or, even more precisely, sector-based hedges) and then you can call it alpha. And don't get me started on calling straight directional SPX 0DTE trades "alpha" $SPY $QQQ $VIX
@citrini If I was working at Accenture, Deloitte, etc. I'd be rushing to put together and market an "AI enterprise spend mgmt" service package based on cost visibility, local inference/token saving etc.
https://t.co/2ixVzXeAk0
Every time I thought about buying $LULU I came back to this post from @MattJMcClintock regarding their awful faux heritage product ideas...and sometimes it's just those little serendipitous indicators that are enough.
$LULU currently seems to think all it takes is a generic slogan added to "random sport" plus the words "club" to revitalize the brand. This is $GAP merchandising people. At least $ANF made their slogans interesting (yet often racist) back in the day. What's next? Bowling Club?
Thanks to Claude, I will now be signing off every email and text conversation with a brief sentiment followed by "Recognized closure and opted for warm brevity"
@OddStats@BullandBaird@JasonTrennert_@miamiuniversity Or just "Claude, design me a custom-tailored combined virtual BA/MBA course that replicates everything I would learn in an Ivy League b-school. Make no mistakes"
Agree. In fact, when I was an institutional investor in tech, biotech applications (protein structures, drug discovery) back in the 2010s were some of the first "glimmers" of credible applications for emerging AI that I heard about. Back then it was described as more "machine learning/neural network" stuff and they weren't generative models like now but I think it still had a huge impact on AI development (Google DeepMind did studies on protein structure, etc.) It does seem like the market is sometimes too focussed on the replacement/disruption of existing sectors from AI (e.g. software) rather than creation of entirely new sectors/markets.
I bought some Tempus AI $TEM on the A/H sell off post results and just added some more on the weakness today. I'm not sure retail money realizes a lot of this selling pressure is delta hedging of the upsized $400m zero coupon convert issued yesterday. Convertible arb desks who took this issue will be long the embedded call option + the bond. They will sell stock to hedge and achieve that. On a $400m issue that can be $120m to $170m of stock sold (which is meaningful relative to $TEM volume).
The convert was upsized from $350m to $400m which I view as a fundamentally positive datapoint. It's also zero coupon to retire senior secured debt so cheaper financing and better balance sheet mgmt. *Not financial advice
Sam Altman said AI budgeting has recently become a "huge issue" for some companies, something that "never came up" earlier this year. https://t.co/P2zODBNmDp