Executive Summary
- Hormuz structure confirmed: Iran's Foreign Minister and the IRGC are operating on independent command authority. Any diplomatic ceasefire must now be discounted by the probability the IRGC was not party to it. Force majeure clauses remain active on major shipping lanes. WTI structural floor at $90+ is more defensible than last week.
- Fed leadership vacuum: Senator Thom Tillis has blocked Warsh's confirmation until a DOJ investigation into Powell is dropped. Powell's term ends 15 May. A lame-duck handover with no confirmed successor removes the forward guidance mechanism that has anchored the yield curve since 2013. This is a tail-risk amplifier, not a directional rate call.
- Tesla Q1 misses on deliveries, raises on ambition: 358,000 vehicles delivered versus expectations — 40,000 short, 50,000 added to inventory. The more important number, the Terafab AI compute commitment, was not quantified. The market is no longer sure which metric to use.
- Q1 earnings season broadly solid: 86% of approximately 50 S&P 500 reporters beat EPS estimates. Earnings growth at 14% YoY for the fifth consecutive quarter. Hyperscaler capex guidance (MSFT, GOOGL, META) due week of 28 April — the decisive test for the AI infrastructure thesis.
- Citrini-Citadel score moves to 6–5: Apollo's Torsten Slok published a chart showing AI-adopting sectors pulling ahead in new firm creation — the first time the AI productivity dividend has shown up at the foundational economic layer, not just in equity valuations. Both theses can be simultaneously correct, operating on different populations.
- TLT now down 8.0% YTD: Ten-year yields rose from approximately 3.85% at the start of 2026 to 4.31%. Long-duration bond holders have quietly experienced a significant drawdown that shorter-duration benchmarks obscure. The Warsh Taylor Rule transition does not obviously resolve this.
Traffic Light Risk Dashboard
What This Means in Practice
Position size around binary outcomes, not directional conviction. The Warsh testimony and IRGC actions this week confirmed that uncertainty — genuine two-sided widening of outcomes — requires smaller positions and asymmetric structures. For equity investors: own the S&P into earnings season but hedge the tail. For bond investors: floating-rate structures are the only clean position when the Fed's reaction function itself is in transition. For energy names: the Hormuz re-escalation was not a surprise to the oil market. It was a surprise only to equity markets, which priced it as resolved. The insurance market, again, told the truth first.
Analytical Takeaway
Two cross-asset tensions defined the week, and neither resolved cleanly. Understanding them mechanically — not just noting that they exist — is what separates a position from a narrative.
The Oil-Equity Divergence: Four Consecutive Weeks
For the fourth week in a row, oil priced risk more accurately than equities. The mechanism is straightforward: oil futures price the physical cost of procurement — shipping delays, force majeure, insurance surcharges, routing diversions around the Strait. Equities price sentiment and the probability-weighted expectation of earnings. When a ceasefire is announced, equities rally because sentiment improves and the earnings-risk premium falls. Oil holds or rises because the physical market knows force majeure clauses have not been lifted and tankers are still being turned around.
The IRGC's seizure of three vessels on 23 April — hours after the "indefinite" ceasefire extension — was not a surprise to oil traders. It confirmed what the insurance market had been signalling for three weeks: the diplomatic track and the operational track in Tehran are not the same institution. Investors who relied on diplomatic headlines to price energy exposure took the wrong signal for the fourth consecutive week. This pattern is now sufficiently documented to warrant a formal update to the signal library.
The Warsh Hearing: What Actually Happened
The market's framing of the Warsh testimony was "how hawkish is he?" The correct framing turned out to be "what happens if there is no chair?" Warsh confirmed his full regime change plan: abandoning forward guidance, dropping core PCE as the inflation measure, anchoring decisions to the Taylor Rule, and reducing the frequency of post-FOMC press conferences. Markets had modelled this as a binary between flexibility (base case, 55%) and hawkishness (stress case, 45%). The Tillis blocking manoeuvre introduced a third scenario that was not in the probability distribution.
The Macro Compass (Alf Peccatiello) flagged a useful secondary risk: if Warsh's confirmation triggers an accelerated Fed balance sheet run-down, bank reserves as a percentage of US GDP approach the threshold that caused the September 2019 repo market blowup. That was a technical liquidity event that briefly forced Fed intervention, and it happened when reserves were otherwise comfortable. In a Hormuz-driven energy shock environment, the margin for error is smaller.
| Scenario | Probability | 2Y Target | 5Y Target | 10Y Target | Equity Impact |
|---|---|---|---|---|---|
| Warsh confirmed, Taylor Rule path | 40% | 4.15–4.35% | 4.05–4.25% | 4.20–4.40% | Neutral to mildly negative; curve flattens |
| Leadership vacuum (no chair, May FOMC) | 35% | 4.40–4.65% | 4.25–4.50% | 4.30–4.55% | Vol expansion; tail risk premium repriced |
| Tillis deal struck; Warsh confirmed fast | 25% | 4.05–4.20% | 3.95–4.10% | 4.15–4.30% | Relief rally; 2Y anchors, long end eases |
The most underpriced scenario is the middle one. Options markets are pricing directional rate risk — hawkish vs dovish — not the structurally different risk of a period with no forward guidance mechanism at all. The VIX implied-to-realised spread is the instrument to watch: if institutional hedge books are being rebuilt, the spread widens even as realised volatility remains contained.
What We Know
- IRGC/FM dual-track is confirmed by direct operational evidence, not inference
- Warsh's Taylor Rule plan is confirmed; 2Y at 4.15% reflects partial pricing of new regime
- TLT down 8.0% YTD — long-duration bond holders have already absorbed a significant drawdown
- Five consecutive quarters of 14%+ S&P 500 EPS growth; AI cycle structurally intact
What We Infer
- The insurance market is a more reliable Hormuz signal than diplomatic announcements — it has been correct four consecutive weeks
- The Tillis blocking manoeuvre was not priced into the rates market before the hearing; it may still be underpriced
- Global liquidity peaked Q3 2025 (Howell); assets that depend on secondary market liquidity for their exit assumptions are the most exposed
- Hyperscaler capex guidance week of 28 April is the single highest-signal data point for the AI infrastructure thesis
What Could Change
- If Tillis drops his block, the leadership vacuum risk collapses and yields ease toward the base case range
- If any hyperscaler signals capex reduction or AI revenue miss, the behind-the-meter power trade loses its fundamental anchor
- If IRGC escalates beyond ship seizures to a kinetic exchange with US Navy, the oil-equity divergence becomes an oil-equity correlation (both down sharply)
- Q1 GDP advance (Fri 25 Apr) — a number below 1.5% would add recession risk to the current stagflation narrative
The Grid Problem Drew Baglino Decided to Solve
Nineteen years at Tesla. Then a DOE report. Then Heron Power.
In spring 2022, the United States Department of Energy published a report that, outside a small community of power engineers and utility planners, received almost no attention. It documented what the DOE called the interconnection queue crisis: 2,600 gigawatts of new power generation projects waiting to connect to the US grid, with average wait times of four years and rising. The bottleneck was not land, not permits, not capital. The bottleneck was transformers.
The conventional design, an oil-cooled device built in Germany or South Korea with a lead time of twelve to twenty-four months, is a programme stopper for any hyperscaler trying to connect a 200-megawatt data centre to the grid in six months. Drew Baglino read the same DOE report. He had been at Tesla for nineteen years, most recently as Senior Vice President of Powertrain and Energy, the division that built the Powerwall and Megapack. He spent two decades making high-voltage DC conversion hardware work at manufacturing speed. He understood the problem. In 2023, he left Tesla and founded Heron Power.
The Product
The Heron Link is a solid-state transformer. It converts medium-voltage grid AC directly to 800-volt DC, the bus voltage specified in Nvidia’s AI rack reference design. A data centre built to that specification does not need a separate conversion stage at each rack. The Heron Link does it at the grid interconnection point. Connection time: weeks, not twelve to twenty-four months. Before the February 2026 Series B, customers had expressed interest in 40 gigawatts of capacity. The $140 million raised, co-led by Andreessen Horowitz American Dynamism and Breakthrough Energy Ventures, was production capital, not exploration capital.
The Strategic Uncertainty
Every deep-tech hardware company faces the certification question. The IEEE 1547 standard governs grid interconnection of distributed energy resources. A first-of-kind solid-state transformer has no certified device pathway, and building one takes twenty-four to thirty-six months. Competitors who enter the queue behind Heron Power are locked out for at least eighteen months. The decision in early 2026 was whether to prioritise certification speed or manufacturing scale. Scaling before certification creates inventory risk; waiting creates competitive exposure. The company chose a hybrid: limited manufacturing for the most straightforward application category while pursuing certification for the utility-scale version. It is the kind of decision that reflects engineering discipline — and that investors from the Anduril and Commonwealth Fusion Systems portfolios would immediately recognise as correct.
What This Teaches
The transferable lesson is more specific than “experienced founders win.” When the moat is physics, when the product is constrained by materials science rather than software architecture, the founder who has spent two decades at the frontier of those constraints has an advantage that a talented newcomer cannot replicate in a standard Series A timeline. Baglino spent nineteen years building the foundations that made the Heron Link possible. The DOE crisis spent four years becoming acute enough that a manufacturing-scale solution would find immediate demand. The $1.4 trillion US utility capital expenditure cycle is the force that will pull Heron Link through procurement, past certification, and into the installed base. The market problem and the specific person qualified to solve it arrived at the same moment. That is rarer than it sounds.
⚠ The Technology at Scale — Risk and Distribution
Solid-state transformers and programmable grid infrastructure are not neutral technologies. A grid that routes power with millisecond precision based on willingness to pay creates new categories of energy access inequality. Industrial and hyperscale users who contract directly with next-generation infrastructure experience the AI power economy as abundance. Residential users in areas where utilities are slow to upgrade experience it as congestion, rising bills, and degraded reliability caused by demand they cannot see. The regulatory framework for utility interconnection does not currently distinguish between a household and a hyperscaler in its grid access queue. That will change as solid-state transformer capacity becomes scarce. When it does, the question of who controls the prioritisation logic will carry consequences for energy equity that are difficult to reverse once the infrastructure is built. Heron Power is building the hardware layer. The governance layer has not yet been designed.
The Week That Was
Nothing resolved. A few things got actively worse.
Saturday 25 April — Pakistan Talks Collapse
The Witkoff and Kushner mission to Islamabad did not depart. Iran’s Foreign Minister Baghaei confirmed that no meeting is planned. The ceasefire extended “indefinitely” on 22 April now has no active diplomatic infrastructure behind it. This is not a delay. It is a confirmation that there is no negotiating track currently in operation on either side. The next dialogue pathway, if one exists, must be built from scratch — a process that required six weeks of groundwork, three countries, and Omani mediation the first time around.
Kevin Warsh
The Senate Banking Committee confirmation hearing on Tuesday 21 April produced the most analytically significant Fed testimony since the post-pandemic inflation debates. Warsh confirmed the Taylor Rule transition, the elimination of forward guidance, and the reduction of press conference frequency. He handled the questions with the precision of someone who had been preparing for this specific set of questions for six months. What he could not handle — because no one can — was Senator Thom Tillis's announcement, delivered with the timing of a man who had been saving it, that he would block the confirmation until the DOJ investigation into his predecessor was dropped. The Senate Banking Committee had apparently not scheduled a follow-up vote for the reason that they had not anticipated needing one. The Fed chair race, as of Friday, remains unresolved.
General Hossein Salami and FM Araghchi
Iran contributed two statements and three ship seizures. FM Araghchi's statement last Friday — "the Strait of Hormuz is completely open" — was either the most optimistic diplomatic announcement in the region's recent history or the most misleading, depending on whether you asked the Foreign Ministry or the IRGC. The IRGC, commanded by General Salami, answered the question definitively on Wednesday by seizing two vessels and attacking a third in the hours following Trump's "indefinite" ceasefire extension. The structural fact this confirmed — that Tehran's diplomatic track and its military operational track operate on genuinely independent command authority — was noted in last week's edition as a risk. It is now a confirmed analytical input, not a hypothesis.
Bubble & Risk Scan
Private Credit — The Slow Leak
The private credit market is not in crisis. It is doing something more dangerous: showing early-stage deterioration in a segment of the market that was specifically sold as "uncorrelated" to public markets. BDC non-accruals are approaching 5% at smaller players. Capital formation is down 40% year-on-year. Large commercial banks are re-entering mid-market lending at more competitive terms, which compresses the yield premium that justified BDC allocations in the first place. Private credit default rates climbed to 9.2% for privately monitored ratings by end of 2025. None of these numbers constitute a systemic event. Together, they describe a market that is repricing risk in ways that do not yet show up in the top-line yield figures institutional investors are using to justify allocations. The ERDR section this week covers BDCs in detail.
AI Cybersecurity — From Academic to Operational
Jack Clark (Import AI) published a benchmark result this week that moved from the research literature into the operational threat category: current AI models achieve a 50% success rate on offensive cybersecurity tasks that take human experts approximately 3.2 hours of professional work. The number matters because it crossed a threshold. Below some success rate, AI-assisted offensive security is a research curiosity. At 50%, it is a force multiplier. The implication is not that AI is replacing human hackers — it is that the cost of a sophisticated cyberattack just fell by a factor that is difficult to model with prior threat frameworks. Regulatory bodies have not caught up. Boardrooms have not priced it. Corporate cyber insurance has not repriced it yet.
The Electricity Maths Problem
Morgan Stanley estimates AI power demand will surge by 126 gigawatts annually through 2028. The US grid queue currently holds 2,600 gigawatts of waiting projects with an average connection wait time of five years. Terafab — if it proceeds at the scale announced — would add approximately 1,000 gigawatts of demand to a queue that is already five years long. The arithmetic does not resolve without either a step-change in grid permitting reform or a large-scale adoption of behind-the-meter generation (SOFC, small modular reactors, on-site gas turbines). This is not a constraint that exists in theory. It is showing up as capital allocation decisions right now, in the form of colocation deals and site selection criteria that prioritise proximity to existing generation over proximity to population centres.
The Speed of Now
The Citrini-Citadel Score: 6–5 (Citrini leads)
The running debate between the "AI destroys value" thesis (Citrini) and the "AI creates value" thesis (Citadel) added one point to each side this week, leaving the score at 6–5 with Citrini holding the edge.
Citrini point — Ghost GDP gets foundational: Michigan Consumer Sentiment fell to 47.6 — an all-time low — against a backdrop of S&P 500 record highs. This is the distributional paradox at the heart of the Citrini thesis: aggregate output statistics rise (the S&P reflects corporate earnings growth) while median household experience deteriorates (sentiment surveys reflect lived experience). AI's economic gain is concentrating in the hands of those who own the infrastructure and the equity. The consumption multiplier — the mechanism by which corporate profits historically become household income — is not firing. Citrini calls this "Ghost GDP": output that never circulates.
Citadel point — Firm creation at the base: Apollo's Torsten Slok published data on 22 April showing sectors with high AI adoption rates are now pulling ahead of low-adoption sectors in new firm creation. This is categorically different from the productivity and equity return evidence that has powered the Citadel case so far. New firm formation is the seed layer of future employment and GDP. If AI is enabling new businesses to form at disproportionately higher rates in high-adoption sectors, the value creation is beginning to show up where it matters most for long-run economic health. The Citrini and Citadel theses are, it turns out, simultaneously correct — operating on different populations, different time horizons, and different economic layers. The right framing is not "which side wins" but "which population are you in?"
This week's Moment That Mattered: The Jack Clark benchmark result — AI achieving 50% success on 3.2-hour professional offensive security tasks — moved the technology from "theoretical risk" to "operational reality" in a single data point. Not a gradual progression. A threshold crossing. The speed at which AI capabilities are moving from academic benchmarks to operational deployment is the defining characteristic of this technological transition, and it is moving faster than any regulatory framework was designed to handle.
When Anthony ran this, Claude correctly identified financial services and healthcare administration as sectors where AI adoption is highest but where new firm creation rates are lowest — suggesting efficiency gains are consolidating into incumbents rather than spawning new entrants. That is a different risk profile from software development, where firm creation is accelerating alongside adoption and challenging incumbents from below. The implication: AI may be simultaneously entrenching existing financial institutions and creating new competitors in software. Most equity research teams are not yet asking the question at portfolio level.
One observation: none of the large consultancies Anthony spoke with this week had run this specific analysis on their own portfolios. At an AI adoption rate that just cleared the 3.2-hour offensive security threshold, the competitive intelligence gap between those who run these prompts routinely and those who do not is widening faster than most boards realise.
Geopolitical Watch
The Two-Track Problem
The most important structural insight from this week is not that the ceasefire broke. It is why it keeps breaking. Iran's Foreign Ministry and its Islamic Revolutionary Guard Corps are operating on independent command authority. This is not a communication failure. It is a deliberate feature of the Iranian state architecture: the FM's diplomatic track operates under presidential and foreign policy constraints; the IRGC's operational track operates under Supreme Leader authority with different objectives, different timelines, and no obligation to coordinate with the diplomatic track's commitments.
The practical implication for investors is specific: when Iran's Foreign Ministry makes an announcement — ceasefire, reopening, negotiation framework — the correct analytical response is to discount it by the probability that the IRGC is not a party to it. The force majeure clauses that shipping insurers refused to lift after FM Araghchi's "completely open" announcement on 18 April were not excessive caution. They were correct structural analysis. The IRGC confirmed their validity within 48 hours of the "indefinite" ceasefire extension on 22 April. The insurance market has beaten the diplomatic market as an analytical signal four times in four weeks. That is now a signal library entry, not a coincidence.
Sovereign Bond Spreads — The Geography of Energy Exposure
The UK Gilt now yields 4.52%, versus the US 10-year at 4.31% — a spread of 21 basis points. The German Bund yields 2.78%. These three numbers tell a geographic story about energy exposure scoring: the UK is a net energy importer with significant gas exposure and a consumer price inflation problem that is directly linked to European wholesale gas prices. The Gilt market is pricing that exposure. Germany — insulated somewhat by its early investments in renewable capacity and its industrial demand destruction — shows a materially lower yield. The US, with its domestic energy production advantage, sits in the middle. Sovereign bond spreads in a Hormuz disruption scenario are not just a credit and monetary policy story. They are an energy geography story.
The Tillis Variable — Domestic Policy as Geopolitical Risk
Senator Thom Tillis's decision to link his Warsh confirmation vote to the DOJ's investigation into former Fed Chair Powell is, on one reading, a standard piece of Washington procedural leverage. On another reading — the one relevant to market practitioners — it is an example of domestic political volatility creating a tail risk that is structurally different from the Fed risks the market was already pricing. The Fed leadership vacuum created by Powell's 15 May departure with no confirmed successor is not a hawkish risk or a dovish risk. It is an institutional credibility risk that removes the forward guidance mechanism — the 13-year-old architecture through which the Fed has managed market expectations. Options markets price directional rate risk. They do not have an established framework for pricing the absence of the communication mechanism itself. That is the gap to watch.