Capital is starting to follow physical constraints. One FDA approval is not a platform.

Capital is following physical plant. SLB bought cooling today. nVent bought switchgear last week. Two governments put money behind minerals. FDA approved a RAS drug; one approval is not a platform. Pentagon is scaling expensive interceptors. 13F for Q1 was already converging in those ecosystems — lagged, not today's tape. Gold is consensus. No consumer or distress story.

What is forming. What is accelerating. What is still unproven.

This morning’s strongest Momentum signal is not a single stock. It is a broader move from narrative into physical commitment.

Three themes we were already watching now show the same behavior: AI infrastructure is moving from compute into power and cooling; strategic-mineral policy is moving toward government capital and offtake; defense urgency is moving into longer production commitments. A fourth development is different: a first-in-class RAS inhibitor crossed from scientific possibility into FDA-approved use in pancreatic cancer. The drug is confirmed. Whether that becomes a broader RAS platform is not.

The common thread is capital, procurement, or a regulator following a real constraint. That is stronger than attention. Theme-level LongShort 13F is now tested on Q1 2026 vs Q4 2025: ownership was already converging on the AI-power, minerals, and defense-prime ecosystems. Q2 2026 filings in the warehouse are still incomplete, so we did not use them. That lag does not confirm this morning’s deals.

Momentum Map

AI / Technology — Strengthening
The AI constraint continues to move from GPUs into electricity, cooling, and electrical equipment. A $3.4 billion cooling acquisition announced this morning and a $1.75 billion switchgear acquisition on August 24 are two capital prints. They do not yet confirm the transformer queue or interconnection.

Critical Minerals / Industrial — Strengthening
Government concern is becoming ownership, financing, and guaranteed demand — not policy rhetoric alone. Two countries. Two minerals. Same mechanism.

Healthcare / Biotech — Emerging
FDA approval of daraxonrasib for metastatic pancreatic adenocarcinoma confirms a long-pursued idea: directly targeting RAS-driven cancer in this setting. The drug is confirmed. The broader platform thesis is not.

Defense / Geopolitics — Strengthening
Seven-year U.S. missile-component agreements, Taiwan’s unmanned-systems law, and faster jet-powered drones in Ukraine point to sustained capacity pressure. These are related. They are not one fact.

Gold / Monetary — Consensus
Long-end yields and U.S. fiscal concerns still support the monetary-hedge case. Treasury’s larger long-bond buybacks (announced August 19) show market-function concerns were large enough to elicit intervention. This is no longer undiscovered. Crowding needs attention. This is not this morning’s news.

Consumer / Societal — No Butterfly this morning
Nothing cleared the evidence bar. That is a discovery gap, not proof of stability.

Corporate Distress — Watch / not enough for a new thesis
Refinancing pressure remains relevant. This morning’s scan did not produce an independent new deterioration signal.

1. AI infrastructure: power and cooling — State: Strengthening

The AI-power thesis gained a new type of evidence this morning.

SLB agreed to acquire cooling-equipment maker Kelvion for $3.4 billion in cash and about $700 million of assumed debt. SLB says data centers are Kelvion’s largest and fastest-growing market. It guides Kelvion data-center revenue of about $1.2–$1.3 billion in 2026 and targets $4.5–$5 billion of data-center revenue by 2028. Close is expected in the first half of 2027, subject to regulators. The deal is announced. It is not closed.

This matters because it is not another forecast that AI needs more electricity. It is a large industrial company committing billions of dollars to the physical plant required to run AI data centers.

It does not prove interconnection queues or transformer lead times.

Last week a second print hit the electrical layer. On August 24 nVent agreed to buy Maverick Power, a data-center switchgear and power-distribution manufacturer, for $1.75 billion, plus up to $550 million of earnout, close targeted in the fourth quarter of 2026. That is company primary (nVent release, August 24).

Cooling was bought this morning. Switchgear was bought last week. A utility 8-K that names data-center load or interconnection is still missing.

Causal chain

AI compute growth
→ higher rack density and electricity use
→ power-delivery and thermal constraints
→ scarcity in transformers, switchgear, turbines, and cooling
→ strategic procurement and M&A
→ capacity expansion
→ earnings transmission later, if orders follow

What should happen next
If the thesis is strengthening rather than becoming fashionable, we should see more capacity announcements, more long-term equipment commitments, and backlog growth outside the obvious electrical-equipment names. Timing: weeks to two quarters.

What would weaken it
Lead times normalize without a demand explanation. Data-center projects are cancelled. Equipment capacity expands faster than demand. Acquisition headlines continue while orders and backlog stop improving. The Kelvion deal fails to close.

Capital / attention
Strategic corporate capital: confirmed on cooling (SLB, August 31) and on electrical distribution (nVent, August 24). Two independence groups.
Theme-level LongShort 13F: tested on Q1 2026 vs Q4 2025 (Q2 2026 filings in the warehouse are still incomplete). Ownership was already converging across cooling, electrical-equipment, and independent-power names into Q1. That is lagged. It does not confirm this morning’s deal.
Attention: already high. Do not infer overnight 13F from one M&A headline.

2. Critical minerals: policy is becoming balance-sheet commitment — State: Strengthening

The condition is not that Western governments dislike dependence on China. That has been known for years.

The change is that governments are acting like capital providers and customers.

August 24: USA Rare Earth reported a $1.55 billion U.S. government-backed SPV around Serra Verde offtake. That includes a $750 million Department of War investment with take-or-pay terms and price floors, a $500 million bank commitment letter, and a U.S. government forward purchase of at least $300 million of product over five years.

August 25: the UK National Wealth Fund agreed up to £71 million of equity and debt for Tungsten West to restart the Hemerdon mine, plus an exclusive period to negotiate offtake on up to 50% of feasibility-study tungsten output. The offtake is not signed. The equity and debt are.

Two geographies. Two minerals. Same mechanism:

strategic dependence
→ geopolitical vulnerability
→ government financing
→ guaranteed or negotiated demand
→ project bankability
→ new non-Chinese capacity, if the plants run

A speech does not build a mine. Financing plus committed demand can.

What should happen next
More government equity or debt packages. Signed offtakes, not only negotiation windows. Processing FID. The strongest confirmation would be a third government, independently, using a similar structure on a different mineral. Timing: weeks to quarters for more packages; quarters to years for tonnes.

What would invalidate it
Projects fail despite public money. Processing stays uneconomic outside China. Commitments are delayed or withdrawn. New supply arrives without sustainable economics.

Capital / attention
Government / strategic capital: strengthening.
Theme-level LongShort 13F: tested on Q1 2026 vs Q4 2025. Ownership was converging across the named minerals ecosystem into Q1. Lagged. Not confirmation of the August SPV or the UK package.

3. Healthcare: RAS crossed into pancreatic-cancer reality — State: Emerging

On August 26 the FDA approved Revolution Medicines’ Rasonque (daraxonrasib) for adults with metastatic pancreatic adenocarcinoma after prior systemic therapy, or when multiagent therapy is unsuitable.

In the 500-patient randomized trial, median overall survival was 13.2 months versus 6.7 months on standard chemotherapy. Median progression-free survival was 7.2 months versus 3.6 months.

That is a hard clinical and regulatory fact (FDA, August 26).

Momentum should not turn one approval into an entire oncology-platform thesis.

The larger Butterfly — Early / Unconfirmed — is whether broad RAS inhibition becomes a reusable approach across other RAS-driven tumors.

What should happen next
If that broader thesis is real, we should see trial success in other tumor types, competing mechanisms attracting capital, combination regimens, and more research intensity around pan-RAS. Timing: quarters, not days.

What would invalidate the broader thesis
No meaningful efficacy outside this indication in the next two registrational cycles. Toxicity that blocks combinations or dosing. Resistance that cuts durability. A competing approach that is clearly better.

Individual drug approval: confirmed.
Broader RAS-platform implication: inference. It still needs proof.

That distinction matters. Being early is not the same as promoting a platform.

Hunt note: this morning the FDA also approved a Besremi label expansion into essential thrombocythemia. Real primary. Not a new mechanism. Not the RAS condition. Not seeded.

RVMD 13F (named issuer only, Q1 2026 vs Q4 2025): holder count rose. That is lagged and pre-label. It does not prove a RAS platform.

4. Defense: industrial capacity and a cheaper-intercept question — State: Strengthening

The defense signal is broader than one weapon. It is also not one story.

This morning the Pentagon announced seven-year agreements with General Dynamics Ordnance and Tactical Systems and Lockheed Martin to expand missile-component production for THAAD and PAC-3 MSE. The stated aim is to triple PAC-3 MSE output and quadruple THAAD. Minimum annual quantities are guaranteed, subject to annual appropriations. This is industrial capacity for expensive interceptors.

That is not the first print in this chain. On August 3 the Department of War signed separate seven-year frameworks with Northrop Grumman (about $3 billion) to add a second source of PAC-3 solid rocket motors and to expand THAAD structural components. Same condition. Different supplier layer. Two independence groups.

On August 27 Taiwan’s legislature passed a six-year unmanned-systems law of about NT$240 billion (about $7.3–$7.6 billion). Funding runs through the annual budget, with a NT$40 billion yearly cap in principle. Taiwan’s defense minister has said that cap can slow planned procurement. This is a large industrial-policy commitment with budget risk. It is not a locked production run.

Also this morning: Russia fired about 1,500 drones at Ukraine from Thursday to Sunday, about 800 of them jet-powered. Ukraine is using combat aircraft while interceptor-drone makers iterate. That is a battlefield cost problem, getting worse.

Causal chain

high-volume autonomous attack
→ interceptor use and cost asymmetry
→ need for cheaper, scalable counter-drone systems
→ longer procurement commitments
→ manufacturing expansion
→ new suppliers and architectures, if buyers actually purchase the cheap layer

Butterfly to watch
The interesting edge is not drones. That narrative is mature.

It is the economics of counter-autonomy: inexpensive detection, autonomous interception, and manufacturing built to defeat large numbers of cheap attackers. The U.S. is scaling expensive interceptors. Cheap interceptors are not yet a U.S. purchase.

What should happen next
More multiyear component deals. A U.S. or allied purchase of low-cost interceptors, with cost-per-intercept in the requirement. Taiwan actually funding through — or raising — the annual cap. Timing: weeks to two quarters for contracts; longer for output.

What would invalidate it
Inexpensive interception cannot scale against jet-powered drones. Procurement stays in expensive legacy interceptors. Taiwan’s annual cap and politics prevent a domestic build.

Butterfly Ledger — this morning

Strengthened: AI power and cooling — strategic M&A adds a new capital channel. Grid equipment and interconnection still not printed.

Strengthened: critical minerals — policy moving toward financing and guaranteed demand in two countries.

New Butterfly — Early / Unconfirmed: broader RAS…