What the money is doing

ABBY

This is Agent Mode AI. I'm Abby. Today we're checking three claims at once: AM-118, AM-119, and the vendor-economics section of AM-121. They describe three institutional capital cohorts reading the same AI risk surface from three different angles. The asymmetry between how they're behaving is what we want to walk through.

AVERY

I'm Avery. Frame the asymmetry.

ABBY

Three cohorts of institutional capital. Pension funds and sovereign wealth funds, who manage the longest-dated capital in the system. Reinsurance, who underwrite the catastrophic tail of the insurance market. SaaS investors, who fund the AI vendors at primary market and secondary. All three read the same AI risk surface. Pension capital is silent. Reinsurance capital is repricing. SaaS investor capital is celebrating. Three institutional cohorts, one risk surface, three different reactions.

AVERY

Start with AM-118. The pension fund piece.

ABBY

AM-118 is the claim that the major institutional pension funds and sovereign wealth funds — NBIM in Norway, CalPERS in California, similar entities globally — have published essentially nothing on AI risk as of mid-2026. NBIM has approximately 1.7 trillion dollars in assets and a published governance position on virtually every other category of systemic risk: climate, cyber, anti-corruption, board independence, voting policies. On AI specifically, the published corpus is thin to absent.

AVERY

What does the absence mean.

ABBY

The absence is the editorial signal. NBIM publishes detailed governance positions on emerging risks specifically because of its scale and time horizon. The fact that AI is named in their governance documentation as an area of interest but not anchored to a published policy position by mid-2026 is a structural data point about how the longest-dated capital in the system is reading the AI risk surface.

AVERY

Two readings of the silence.

ABBY

One reading: pension capital is waiting for the empirical evidence before committing to a public position. The argument from this side is that publishing too early on AI risk locks the fund into a stance the data doesn't yet support. The other reading: pension capital sees AI risk as more diffuse than other categories — not concentrated in any specific holding, hard to express as a portfolio adjustment. The argument from this side is that the absence reflects an analytical gap, not a strategic choice.

AVERY

Either reading is consistent with the observed data.

ABBY

Both are. The point isn't to choose between them. The point is that the longest-dated capital in the system has not committed to a position. That's a data point worth tracking against AM-119 and AM-121.

AVERY

Now AM-119. The reinsurance side.

ABBY

AM-119 is the upstream piece. Lloyd's of London, Munich Re, and Swiss Re have actively repriced AI tail risk into 2026 cyber treaties. The published research from all three is consistent. Lloyd's Futureset systemic-risk content explicitly treats AI as an emerging cyber tail-risk category. Munich Re's annual Cyber Insurance Risk Report names AI as a catastrophic-scenario category and provides loss-modelling frameworks that primary carriers reference. Swiss Re Institute's sigma research published 2025 work on AI-related liability and cyber-physical convergence.

AVERY

The downstream effect.

ABBY

Treaty terms have tightened. Catastrophe-bond issuance for cyber risk has shifted in the direction the reinsurance signal predicts. Per-event retentions at the reinsurer level have moved. The signal travels downstream to primary cyber renewals on a six-to-twelve month delay.

AVERY

So the reinsurance market is not silent. It's actively committing capital to its position.

ABBY

That's the difference. Reinsurance capital is shorter-dated than pension capital and has direct exposure to catastrophic loss. The published research is the artefact, the treaty terms are the commitment. Both are observable.

AVERY

Now the third cohort. SaaS investors via AM-121.

ABBY

AM-121's vendor-economics section. ServiceNow's RPO at the end of Q1 FY26 was 27.7 billion dollars, up 25 percent year over year. Customers at 5 million dollars or more in annual contract value: 630, up from 516 a year prior. Datadog's Q3 FY25 10-Q discloses that its AI-native cohort contributed approximately eight percentage points of year-over-year revenue growth for the quarter, with the company's filing language flagging the concentration as a risk factor.

AVERY

That's vendor-side disclosure. What does it tell us about investor behaviour.

ABBY

Investor behaviour follows the vendor-side disclosure. ServiceNow's market capitalisation reflects the RPO trajectory. Datadog's market capitalisation reflects the AI-native cohort's contribution. The capital-market response to AI-vendor revenue exposure has been to reward growth, not to discount for the concentration risk that the company itself flagged.

AVERY

The asymmetry.

ABBY

Pension capital silent. Reinsurance capital actively repricing AI tail risk in 2026 cyber treaties. SaaS investor capital celebrating AI-attributable revenue growth at the vendor level. Three institutional cohorts, all reading the same underlying AI risk surface, all behaving in different directions.

AVERY

Why is this the editorial point.

ABBY

When three independent capital cohorts read the same risk surface and behave in three different directions, the divergence itself is the editorial signal. One of three readings is going to look most accurate in retrospect. We don't know which one yet. What we know is that the pension capital position — silence — is the position that has the most option value if the AI risk surface materialises in either of the directions the other two cohorts are pricing.

AVERY

What the CIO listening should take from this.

ABBY

Two specific things. First, the reinsurance pricing signal is concrete and observable in your 2026 cyber renewal. The terms tightened upstream. Read your renewal accordingly. Second, the AI-vendor concentration flagged in Datadog's 10-Q is a leading indicator for how multi-year SaaS commitments price AI exposure. If your vendor's AI-native customer cohort is concentrated, your renewal trajectory carries that concentration risk.

AVERY

Verdicts.

ABBY

AM-118 is Holding. The absence-as-signal framing is supported by NBIM's published governance corpus on every other risk category. Cadence is sixty days because the next scheduled NBIM governance update will be the first observable signal change.

AVERY

AM-119.

ABBY

AM-119 is Holding. Same reading as Episode 1 — Lloyd's, Munich Re, Swiss Re published research is consistent, the downstream effect on primary cyber renewals is observable. Cadence is sixty days.

AVERY

AM-121.

ABBY

AM-121 is Holding. The SEC-filed RPO and AI-native cohort numbers are audited. The investor reaction is observable in market capitalisation. Cadence is sixty days.

AVERY

What would change any of them.

ABBY

For AM-118: a published NBIM or CalPERS AI governance position would close the absence-as-signal frame. For AM-119: a coordinated reinsurance-market softening on AI tail risk. For AM-121: a material change in ServiceNow's RPO trajectory or Datadog's AI-native cohort growth contribution.

AVERY

Final word.

ABBY

The three claims, the SEC filings, and the published reinsurance research are linked at agentmodeai dot com slash holding. The Sunday brief ships every week with what moved on the ledger.

AVERY

Holding-up. See you next Sunday.