Creators → adopters: the handoff that decides whether capex normalization is a bear market

One-line summary: The AI trade's creators and enablers drove a hyper-concentrated market; the claim now is that adopters start capturing the productivity — and whether that handoff happens is the difference between "capex moderates and the S&P 493 out-earns" and "capex moderates and it's 2022 again."

The insight

Two independent July-2026 sources frame the same handoff, and it is the most consequential open question in the book because it determines the sign of capex normalization rather than its timing.

The setup. jonathan-thomas (CEO, American Century Investments, ~$350B AUM): "for a while I'll call it, the creators. The creators and the enablers of AI were just soaring. And now what's starting to happen is the adopters are starting to receive the benefit from it. So that concentration of the creators at the beginning drove the market like crazy. Now the adopters are getting the benefit. And I think ultimately the winners and losers will be determined by who adopt it and who's deploying it correctly."

The stakes, stated precisely. josh-brown: "what happens when capex plans start to moderate and all this earnings growth. What's the — is there a handoff where we don't have to automatically just have a bear market? Like we don't just automatically have to have a 2022, another year of efficiency. And I think that's the — I don't know if it's gonna work. That is the handoff where the S&P493 start to outearn... because of all these investments they're making in AI start to result in higher earnings and more productivity. Like that is the way the market gets out of this hyper concentration amongst seven giant stocks."

Note Brown's honesty: "I don't know if it's gonna work." This is a hypothesis about the future, not an observation.

The mechanism Thomas names is productivity, and it is testable. "If the adopters realize the expected benefit, productivity, because productivity is what drives margin, profits, gdp, the economy, everything. If they realize those productivity benefits that I think are out there, this will tick back up." And Brown names the observable that would confirm it: "You could have 500 companies in the financial sector and the healthcare sector say something on an earnings call. Like, we credit the beat to these investments we made a year ago in updating our workflows and processes to the AI era." That is a falsifiable prediction with a clear check — count the earnings calls.

The breadth data is real, and it is the one part that isn't a forecast. sean-russo: "the 493 is absolutely trouncing the Mag 7 so far this year." jonathan-thomas: "for the first time in I don't know how long, 50 some odd percent of the S&P is up over 10%... And that broadening out is now happening. And you see it across small cap, you see it across emerging market. It's really, it's live and it's real." Earnings support it: 23.2% expected YoY S&P growth, 20.9% excluding the Mag 7, with 10 of 11 sectors growing.

But the same episode supplies the counter, and it is important. A Forward Guidance co-host's read on the identical tape is that the rotation is mechanical, not fundamental — see Contradictions. Two sophisticated observers looking at the same week's rotation and disagreeing about whether it is a handoff or a squeeze is exactly the tension worth recording.

The independent-forcing-function version. tyler-neville arrives at the same destination from the cost side: "what if the models get cheaper and cheaper and now value businesses actually can grow margins and it switches the game? That's sort of what I'm thinking about here. I don't think anyone's really talking about that." If open-weight models collapse the cost of intelligence, the adopter's margin gain arrives without the frontier labs capturing it — which is open-source-share-shift-bullish-for-compute pointed at the app layer instead of the compute layer.

Evidence

The rotation thesis — all from 2026-07-17-podcast-the-compound-and-friends-you-re-about-to-see-the-real-ai-winners-stand-up:

• jonathan-thomas: "the creators and the enablers of AI were just soaring. And now what's starting to happen is the adopters are starting to receive the benefit from it... ultimately the winners and losers will be determined by who adopt it and who's deploying it correctly."
• jonathan-thomas: "If the adopters realize the expected benefit, productivity, because productivity is what drives margin, profits, gdp, the economy, everything. If they realize those productivity benefits that I think are out there, this will tick back up."
• josh-brown: "That is the handoff where the S&P493 start to outearn... that is the way the market gets out of this hyper concentration amongst seven giant stocks." With the caveat: "I don't know if it's gonna work."
• josh-brown on where to look: "healthcare and financials arguably have companies that could be among the biggest beneficiaries of AI as users, as adopters. Like what? Think about insurance underwriting... if AI lives up to the promise... You could have 500 companies in the financial sector and the healthcare sector say something on an earnings call. Like, we credit the beat to these investments we made a year ago."

The breadth data (realized, not forecast):

• sean-russo: "the 493 is absolutely trouncing the Mag 7 so far this year."
• jonathan-thomas: "50 some odd percent of the S&P is up over 10%... you see it across small cap, you see it across emerging market. It's really, it's live and it's real."
• josh-brown: analysts expect 23.2% YoY S&P earnings growth; excluding the Mag 7, 20.9%; 10 of 11 sectors will grow earnings; tech's contribution is largest at 65% growth.
• Base rate on the forecasts — sean-russo: "generally, outside of recessions, analysts nail it. They nail earnings. Unless it's 2001, 2002, 2009, and then 2020." Brown's caveat: "The verticality is the thing that bothers me. Of the next year's estimates."

A live adopter datapoint from an adopter's own call:

• patrick-conway in 2026-07-16-earnings-unh-q2-fy2026 (⚠ partial/condensed source — only quoted strings are verbatim): ambient-listening AI available to "70% of employed Optum Health providers", targeting "exceed 90% by year-end"; Value Connect AI credited with pharmacy cost savings of "17% for early clients." stephen-hemsley reaffirms a "13% to 16%" long-term annual earnings growth target with technology-driven productivity a named contributor. This is close to the Brown test — a healthcare company crediting AI deployment on an earnings call — but note it is a deployment statistic and a growth target, not "we credit the beat to AI." The test is not yet met.

The cost-side route to the same place:

• tyler-neville in 2026-07-17-podcast-forward-guidance-the-ai-unwind-is-forcing-a-historic-market: "what if the models get cheaper and cheaper and now value businesses actually can grow margins and it switches the game?"
• jack-farley in the same source: "I can get 80% of the capability of Fable 5, but in this open model that costs 10%" — the cost collapse that would make adopter margin gains arrive fast.

Implications

• The falsifiable test is Brown's: count healthcare/financial-sector earnings calls that credit a beat to AI-era workflow investment. That is checkable each quarter and is the cleanest confirmation available. UNH's Q2 is a deployment datapoint, not yet an attribution one.
• This is the sign-determining question for the whole AI-capex cluster. If the handoff happens, capex moderation is a rotation (mega-issuance-peak-to-ai-capex-derate is a hedge that never pays). If it doesn't, capex moderation is a de-rate. The book is currently positioned mostly at the bottom of the ai-cake-layer-framework, which is the losing side of that coin.
• The book has no clean adopter expression. RSP (equal-weight S&P) is the crude version of "the 493 out-earn." Named candidate sectors are healthcare and financials as users.
• Thomas's dot-com analogue is the cautionary version: the internet's value did accrue above the infrastructure, but he dates full realization to 2020 — twenty years after the capex bubble. A handoff that takes two decades is not a trade.

Contradictions / tensions

• 🔴 Direct disagreement on what the rotation is. From 2026-07-17-podcast-forward-guidance-the-ai-unwind-is-forcing-a-historic-market — a co-host (diarized only as "Felix", a first-name label this ingest declined to resolve into a person entity) reads the same tape as mechanical, not fundamental: "I see a lot of reading into this as idiosyncratic strength due to some reason behind the hyperscalers which definitely has some truth to it... But a lot of this I would attribute to the rotational mechanical flows of a positioning unwind and squeeze" — and, on the momentum move generally, "it was also the craziest run up in momentum in the same period and it was by design with the manipulation in the market." Thomas sees a fundamental handoff beginning; the Forward Guidance read is a crowded-trade unwind wearing a handoff's clothes. Both are looking at the same week. Recorded, not adjudicated.
• The productivity gain is asserted, not measured. Thomas: "those productivity benefits that I think are out there." No source in this ingest measures realized adopter productivity. This sits in direct tension with ai-roi-reckoning, where Chamath cites realized enterprise ROI of "0 to 2%" — the adopter benefit this concept depends on is the exact thing that concept says is not showing up.
• Interested parties. Thomas runs an asset manager whose Avantis products (value/small-cap tilts) benefit directly from a broadening market; the episode is substantially an interview about those products. Brown runs an RIA. Neither is neutral on "the 493 out-earn."
• Brown himself doesn't believe it yet — "I don't know if it's gonna work" — and notes he has been burned before: "I've also been burned too many times of thinking, here comes the value rotation."
• The breadth data is YTD price, not earnings delivery. "The 493 trouncing the Mag 7" is consistent with the handoff and with a positioning unwind. It does not discriminate.

Open questions

• Do healthcare/financial earnings calls start attributing beats to AI deployment? The Brown test. Q3 2026 is the next read.
• Is realized adopter productivity measurable anywhere, or does it stay an assertion? (See the Cintrini report mentioned by Thomas as the doom-side counterpart.)
• Can the "mechanical unwind" and "fundamental handoff" readings be discriminated — is there an observable that separates them?
• Is there a better adopter expression than RSP?

Valuation snapshot

Last refreshed 2026-07-20 (pre-open; the main-table marks are the Friday 2026-07-17 close, markets closed over the weekend). Every price fill tagged twelvedata. Mkt cap / Fwd P/E are not in the Twelve Data free tier and were not re-sourced (index proxies, not single names). Note: the single-session comparison below is a dated 2026-07-16 read and is left intact as the historical evidence for the concept — it is not re-marked.

2026-07-16 was, on its face, this concept's thesis printing in a single session — and the page's own contradiction section is the reason to be careful about saying so.

The discriminating comparison the day offers — creators vs the rest, same session:

Forward-looking outcomes (12-month)

Bull case — the handoff is real: adopters convert AI deployment into attributable margin, and capex moderation becomes a rotation rather than a de-rate: jonathan-thomas names the mechanism and it is testable — "If the adopters realize the expected benefit, productivity, because productivity is what drives margin, profits, gdp, the economy, everything. If they realize those productivity benefits that I think are out there, this will tick back up." josh-brown names the stakes: "That is the handoff where the S&P493 start to outearn... that is the way the market gets out of this hyper concentration amongst seven giant stocks." The breadth data supporting it is realized, not forecast — sean-russo: "the 493 is absolutely trouncing the Mag 7 so far this year"; jonathan-thomas: "50 some odd percent of the S&P is up over 10%... it's live and it's real"; and 23.2% expected YoY S&P earnings growth, 20.9% excluding the Mag 7, with 10 of 11 sectors growing. Implied price: RSP +12–20% and, more to the point, RSP outperforming SPY by 5–10pts. Cited: 2026-07-17-podcast-the-compound-and-friends-you-re-about-to-see-the-real-ai-winners-stand-up.

Base case — breadth persists but attribution never arrives, so it stays a rotation without a thesis: the 493 keeps outperforming on flows, positioning and defensive rotation rather than on demonstrated AI-driven productivity. Nobody ever satisfies the Brown test. RSP grinds up with the market; the concentration unwinds slowly for ordinary mean-reversion reasons rather than for the reason this page argues. Implied price: RSP +5–10%, roughly in line with SPY. Cited: 2026-07-17-podcast-the-compound-and-friends-you-re-about-to-see-the-real-ai-winners-stand-up.

Bear case — the rotation is mechanical, and the productivity it assumes is measurably absent: 🔴 the same week's tape reads the opposite way to a different desk. Per 2026-07-17-podcast-forward-guidance-the-ai-unwind-is-forcing-a-historic-market, a Forward Guidance co-host: "a lot of this I would attribute to the rotational mechanical flows of a positioning unwind and squeeze" — a crowded-trade unwind wearing a handoff's clothes. Worse, the productivity the bull case requires is the exact thing another concept says isn't showing up: ai-roi-reckoning cites realized enterprise ROI of "0 to 2%." If capex moderates without the handoff, it's 2022 again — and the book is positioned mostly at the bottom of the ai-cake-layer-framework, which is the losing side of that coin. Implied price: RSP −10–15% in a broad de-rate; the relative trade still works even as the absolute one doesn't. Cited: 2026-07-17-podcast-forward-guidance-the-ai-unwind-is-forcing-a-historic-market, ai-roi-reckoning.

Currently undervalued vs base case? No — RSP sits 0.6% from its 52-week high, and nothing on this page argues it is cheap. The temptation after a session like 2026-07-16 is to read the tape as confirmation and buy the confirmation. Three reasons on this page's own evidence not to:

1. The day's green was defensive, not adopter-driven. DG +4.8%, ROST +3.1%, TJX +2.5%, XRT +1.5% is a consumer trade-down signature. That is a different mechanism from "healthcare and financials capture AI productivity," and reading it as this concept's thesis would be the exact error the Forward Guidance co-host names.
2. The falsifiable test has not been passed. Brown's check is specific and clean: "You could have 500 companies in the financial sector and the healthcare sector say something on an earnings call. Like, we credit the beat to these investments we made a year ago in updating our workflows and processes to the AI era." UNH's Q2 is the closest datapoint in the wiki and it does not clear the bar — 70% provider deployment and a 13–16% long-term growth target are a deployment statistic and an aspiration, not an attribution.
3. Everyone making this argument is paid to make it. jonathan-thomas runs an asset manager whose Avantis value/small-cap products benefit directly from a broadening market, and the episode is substantially an interview about those products; josh-brown runs an RIA. And Brown himself doesn't believe it yet — "I don't know if it's gonna work" — and volunteers that he has "been burned too many times of thinking, here comes the value rotation."

The honest position: this is the most consequential open question in the book, and it is not yet a trade. It determines the sign of capex normalization for every AI-infra chain the project owns — which is why it earns a page — but the one thing that would convert it is a quarter of earnings-call attributions, and that has not happened. Thomas's own dot-com analogue is the caution: the internet's value did accrue above the infrastructure, but he dates full realization to 2020, twenty years after the capex bubble. A handoff that takes two decades is not a trade.

Catalyst path:

• Q3 2026 healthcare/financial earnings season (October) — the Brown test. Count the calls that credit a beat to AI-era workflow investment. This is checkable, dated, and the cleanest confirmation available.
• UNH Q3 2026 — does the 70%→90% ambient-AI deployment convert from a deployment stat into a margin attribution? The nearest single-name read.
• Mag 7 Q3 capex guidance — the moderation this whole question is conditioned on. If capex guides down before any adopter attribution appears, the bear case runs first.

Related

• ai-cake-layer-framework — the structural version: value accrues at the app layer
• ai-roi-reckoning — the direct counter: realized enterprise ROI "0 to 2%"
• steroid-era-earnings-inflation — what happens to earnings if the handoff doesn't come
• mega-issuance-peak-to-ai-capex-derate — the hedge that pays only if the handoff fails
• open-source-share-shift-bullish-for-compute — the cost-side route to adopter margin
• jonathan-thomas · josh-brown · sean-russo · tyler-neville
• unh — a live adopter deployment datapoint