2026-05-13

- Eos Energy Enterprises (EOSE) reported Q1 FY2026 EPS of $0.12 this morning — the biggest surprise of the two BMO prints, because this is a company that has been burning cash for years and just posted its first positive quarterly earnings. EOSE sits in L22 (Secure the Power) with a 4/10 moat score — the zinc-bromine chemistry is genuinely differentiated versus lithium-ion for long-duration energy storage, but the company has been operating at negative gross margins and the technology remains unproven at commercial scale. The $305M DOE loan guarantee provides a financial backstop that most battery startups would envy, and the behind-the-meter data center backup use case ties EOSE directly to the AI infrastructure power thesis. A $0.12 print against what has been a consistently negative earnings trajectory suggests the manufacturing ramp at the Turtle Creek facility may be reaching a cost inflection — management had targeted gross margin positive in H2 2026, and this result moves the timeline forward. The chain watch runs to Fluence Energy (FLNC, 5/10 moat) and Bloom Energy (BE, 6/10 moat) in the same L22 layer — both are further along the commercialization curve but lack Eos's chemistry differentiation. The broader read is about whether long-duration storage economics are improving faster than expected, which would loosen the grid-constraint bottleneck that has limited data center site approvals in Virginia, Texas, and Phoenix.

- Vishay Intertechnology (VSH) printed Q1 FY2026 EPS of $0.05 this morning, and the number underscores exactly why this name sits at a 3/10 moat score — the broadest passive component portfolio in the world, but not an AI story. VSH is in L18 (Run the Cables) alongside Corning (GLW, 8/10 moat), Amphenol (APH, 7/10), and TE Connectivity (TEL, 8/10), all of which carry meaningfully higher moat scores because their AI data center exposure is concentrated and growing. Vishay's data center revenue is less than 10% of total sales, and the $0.05 print reflects a cyclical recovery in passive components — resistors, capacitors, inductors, diodes — that are inside every rack but face intense Asian competition with short qualification cycles. The read-through to the broader L18 layer is limited: Vishay tells you about the general electronics cycle, not about AI infrastructure demand specifically. Corning's optical fiber and Amphenol's high-speed connectors are the AI-relevant reads in this layer, and Vishay's commodity positioning means its results travel in a different orbit. The $0.05 is an improvement from Q4 2025's $0.01, and book-to-bill at 1.20 signals improving orders, but the Newport SiC wafer fab integration continues to drag margins by 130–150 bps. This is a passive component cycle story, not an AI infrastructure inflection.

- Cisco Systems (CSCO) reports after the close today — the marquee print yesterday's brief flagged, and the most consequential single result of the week for the AI infrastructure stack. Cisco sits in L15 (Connect Servers to Other Servers) with a 7/10 moat score anchored by 100M+ deployed devices, 60%+ enterprise switching share, and a certified workforce moat that competitors cannot replicate quickly. The question heading into Q3 FY2026 is not whether Cisco's enterprise business is stable — it is — but whether Silicon One is converting hyperscaler design wins into revenue at a pace that justifies the roughly 19x forward P/E the stock carries. AI infrastructure networking orders have been running north of 50% growth, and tonight's print will show whether that trajectory is accelerating into the second half or plateauing as hyperscalers split wallet between Cisco and Arista Networks (ANET, 8/10 moat). The chain read is unusually clean: Arista is the direct competitive benchmark in hyperscaler switching, but Cisco's optical networking exposure through L11 also touches Coherent (COHR) and Lumentum (LITE), both of which reported earlier this cycle. If Cisco's AI data center orders are growing, that confirms the physical networking buildout is running at or above the pace the chip and power layers have already validated.

- The divergence worth tracking this week sits between the silicon layers and the physical infrastructure layers — and it continues to widen. Over the past two weeks, the earnings cycle has confirmed that chip design (CEVA, L01), semiconductor equipment (ONTO, ACLS, L02), materials (Q, L03), foundry (GFS, L04), packaging (CAMT, L05), and finished chips (AMD, ARM, ALAB, L06) are executing cleanly. Power generation has been strong — Constellation Energy's $2.74 beat on Monday anchored the L21 layer, Eaton (ETN) at $2.81 confirmed the L23 power delivery chain, and Cameco's Q1 print kept the nuclear fuel story intact. But the operational layers — the companies that actually run data centers and generate revenue from compute — have been messier. CoreWeave (CRWV) and Iris Energy (IREN) both flagged cost pressures, TeraWulf (WULF) posted a -$0.44 loss, and the gap between "building the infrastructure" and "operating it profitably" is becoming the central tension in the AI supply chain. Today's Cisco print sits at an interesting inflection point in this divergence: networking equipment is a physical infrastructure layer, but Cisco's recurring software and services revenue gives it an operational margin profile closer to the chip companies than to the data center operators. How the market treats tonight's number will say something about whether investors are still paying for buildout momentum or starting to demand operational proof.

- The networking bifurcation read: Cisco tonight, then silence. After Arista's print earlier in the cycle, tonight's Cisco result completes the L15 layer picture for Q1. The market has already priced Arista as the hyperscaler winner — ANET carries an 8/10 moat versus Cisco's 7/10 — and the question is whether Cisco's Silicon One push is narrowing that gap or whether the enterprise-to-hyperscaler migration thesis is stalling. If Cisco's AI networking orders reaccelerate, the entire L15 layer re-rates. If they plateau, Arista's premium over Cisco widens further, and the networking layer splits into an enterprise incumbent story and a hyperscaler growth story with very different multiple trajectories. The optical interconnect names — Coherent (COHR), Lumentum (LITE), MACOM (MTSI, which printed $1.09 earlier this cycle) — are the secondary chain to watch, since faster networking drives faster optics.

- The power-versus-operations divergence watch: the earnings cycle's defining tension is not resolved. Constellation Energy, Eaton, and Cameco all delivered on the power side. TeraWulf, CoreWeave, and Iris Energy all disappointed on the operations side. The question for the back half of the cycle is whether this divergence is a timing lag — operators will eventually scale into profitability as utilization ramps — or a structural feature of the AI infrastructure stack where value accrues to the picks-and-shovels layers and gets competed away at the compute delivery layer. No reporters from our universe are scheduled for the rest of this week to test this thesis directly, but the Cisco print tonight offers a partial read: Cisco sells the physical networking equipment that both hyperscalers and independent operators buy, and the split between enterprise and hyperscaler orders in its guide will hint at where networking capex is concentrating.

- The energy storage proof-of-concept watch: EOSE's positive print opens a question for the L22 layer. Fluence Energy (FLNC) and Bloom Energy (BE) are further along commercially, but neither has the chemistry differentiation that makes Eos interesting as a long-duration storage play for data center backup power. A $0.12 EPS from a company that has been consistently negative is a data point, not a trend — but if the Turtle Creek manufacturing ramp is real, EOSE's behind-the-meter positioning becomes more interesting as hyperscalers increasingly cite power availability as their primary capacity constraint. The chain watch extends to the broader L21 power generation layer: if data center operators cannot secure enough grid power, behind-the-meter storage and generation become the bypass, and EOSE, BE, and FCEL are the names positioned for that trade.