Investment Gems

Not long ago the market looked like it wanted to finance the solution to every bottleneck on earth. AI semis, networking, power, electrical infrastructure. In the first week of October we could see NVDA, AMD, MRVL, MTSI, ALAB, LITE, APH, power names, all strong. It looked like AI was escaping the GPU rack and turning into a broader physical capital cycle.

Then the market changed its mind fast. By October 13-14 the message became forget building it, own whatever already exists. Refiners, tankers, oil, pipelines. On my strict October 14 momentum screen, roughly 40% of the qualifying names were some form of energy/scarcity exposure while the physical AI cluster basically disappeared. That was our Scarcity Crowding-Out moment: oil/products/logistics stay expensive, inflation stays annoying, rates stay stupid, and suddenly the marginal project gets squeezed from both directions. Building costs more while the hurdle rate also goes up. Scarcity stops encouraging new investment and starts preventing it.

But then something even more interesting happened. The scarcity trade itself started losing monopoly leadership. October 15 suddenly had 176 stocks clearing the stricter momentum screen, and by October 16 leadership had rotated heavily into software, healthcare, consumer/services, payments and other businesses that don’t need to light gigantic piles of money on fire to grow. Refiners and tankers lost much of their dominance. And still no proper return of the AI physical buildout.

So I think the market is slowly reaching a consensus and it isn’t really bullish or bearish in the usual sense. It looks more like: demand probably survives, scarcity probably stops getting exponentially worse, capital stays expensive, so own businesses that can make more money from assets already built instead of requiring another mountain of physical investment.

Basically: Monetization Over Buildout.

This is not AI is dead. Might actually be the opposite. The market may finally be asking the question everyone should have been asking six months ago: cool, we spent a trillion dollars building intelligence, who can actually monetize the damn thing?

That is why software suddenly makes more sense. BOX, DOCU, NOW, TEAM, ZETA, VEEV, ACN, PCTY, PAYC, RELX and a bunch of other capital-light software/services names are showing up in leadership. Paysign is actually a good example of what the bull case can look like for these capital light stocks in the age of AI. They don’t need another 500 MW campus to sell the next dollar. They need all the compute, cloud capacity and enterprise data already sitting there to get used harder.

AI agents generate more workflows. More workflows create more data. More data creates more governance, security and observability requirements. More automation creates more software interactions. You don’t necessarily need AI capex to accelerate forever. You need the capex already spent to start creating economic activity. For instance, Both Twist Biosciences and 10X Genomics are good examples of that in the biotech industry.

The consumer cluster makes this even weirder. If this were just recession incoming, I’d expect utilities, staples, pharma and canned beans. Instead we have SG, FIGS, COTY, ELF, CAKE, ASO, URBN, MAR, HLT, ULTA, IHG and others alongside software and healthcare. That doesn’t scream depression. It looks more like nominal demand is still there, but capital itself has become scarce. Consumers keep spending. Hotels fill rooms. Software monetizes users. What the market doesn’t want right now is a business model saying give us another $20B and maybe we’ll earn something interesting on it in 2030.

Financials are saying the same thing. Mastercard, Visa, MSCI, insurers, specialty finance. Those work. But we still don’t have the JPM/BGC/IBKR/PJT cluster we wanted for a genuine capital boom. Again: monetize financial activity that already exists is better than finance huge amounts of new productive capacity.

Even the industrial names starting to appear fit the same pattern. EMR, GWW, ATKR, ZBRA, EXPD. Automation, distribution, electrical products, logistics, efficiency. Not yet CAT/DE build-everything industrialization. The market isn’t abandoning productivity. It is demanding more productivity per dollar of capital.

So what is the market actually trying to price? My guess is the acute scarcity shock peaks without the economy falling apart. Oil/products/logistics remain structurally tighter than the old 2010s world, but investors stop extrapolating ever-rising scarcity rents. Inflation stays annoying rather than catastrophic. Rates stay higher than the old world. Consumer demand survives. AI keeps spreading. But cheap money physical-capex insanity doesn’t come back immediately.

Capital therefore gets rationed toward businesses with high incremental margins, low incremental capital requirements, existing customers and distribution, and the ability to monetize infrastructure already installed. The market is moving from fearing scarcity will kill demand toward believing scarcity will mostly kill bad capital allocation.

That is actually pretty bullish for certain companies. Just not necessarily the ones everyone spent the last year calling AI picks-and-shovels.

META is probably still one of my favorite expressions of this even though it isn’t currently confirmed by the screen. If AI infrastructure works, Meta monetizes it through ads, recommendations, agents, commerce, Muse and eventually wearables. If the market moves from build AI toward use AI, arguably even better. And if AI monetization keeps working while Meta eventually becomes more disciplined on capex, FCF itself becomes the bull case. META can potentially win from successful AI capex or eventual AI capex discipline. That’s a much nicer payoff tree than owning a supplier whose customer simply stops ordering.

BOX is cleaner and is actually showing up persistently. Enterprise AI needs permissioned access to enterprise data. Agents become more useful, but they also create a giant need for governance, workflow and secure corporate content. No new factory. No tanker. No 500 MW campus. Just more intelligence trying to interact with corporate information. Very clean monetize-the-installed-base exposure.

ZETA is the higher-gamma version. If AI starts moving from infrastructure spending into measurable marketing productivity, ZETA can have serious operating leverage. Small enough for the numerator to matter, high-margin enough for incremental revenue to matter, and much less dependent on whether hyperscalers build campus number 847 next year.

VEEV is the quality version. AI + biological/clinical data + regulated workflows. The life-sciences cluster keeps appearing and I don’t think we should ignore it. If AI improves drug development and commercialization productivity, the softwarelayer can benefit without taking molecule level binary risk. AI utilization with healthcare ballast.

EXPD might be the weirdest convex name in the bunch. If scarcity/logistics remain messy, global supply chains get more complicated and freight intermediaries become more valuable. If scarcity normalizes and physical activity comes back, volume can accelerate. Unlike a tanker, EXPD doesn’t need Hormuz to stay broken forever. It can win from complexity or volume.

MA and V aren’t rockets, but they are very clean ways to own the idea that nominal activity survives. They don’t care whether someone builds another steel mill. They care whether economic value keeps moving through the system. In a capital rationing economy, that is pretty useful ballast.

EMR, GWW and ATKR are interesting as bridge stocks. Right now the market prefers optimization/automation/distribution over giant new capital projects. Fine. But if scarcity keeps easing and physical capital formation starts returning, these are the kind of names that could move before CAT/DE finally show up. They let us own a little optionality on Monetization Over Buildout eventually becoming Monetization plus Buildout.

I would still keep smaller exposure to MRVL, MTSI and maybe MXL. Not as core holdings anymore. More like second leg options. Right now they lack leadership confirmation. That’s fine. The attraction is what happens if oil/products/freight fall, the 10-year eases, software stays strong, MSFT/META/GOOG capex remains intact, and suddenly MRVL/MTSI/ALAB/optical start making highs again. Then the market is basically saying the hurdle rate fell before the demand curve broke.

That is when I would want more AI infrastructure again.

So the portfolio logic becomes pretty simple. Core monetization: META, BOX, ZETA, VEEV, maybe MA/V. Macro bridge: EXPD, EMR, GWW, selectively ATKR. Second-leg AI rockets: MRVL, MTSI, MXL sized smaller. I would be much less interested in chasing tankers and refiners after the giant move because they are basically short the normalization catalyst we now want.

The market spent early October saying build more. Then it screamed own what’s already scarce. Now it increasingly seems to be settling on fine, the economy isn’t dying, just make more money from what we’ve already built while capital is expensive.

That is the consensus I think is forming.

And if that’s right, the key ratio for the next phase isn’t AI exposure. It’s:

incremental earnings / incremental capital required.

The really bullish transition comes later. If the market goes from monetize what we built to monetize what we built AND start building again, that’s when the second AI/capital-formation leg probably starts.

And if we catch that cluster rotation early, that is probably where the outsized returns are.