Apple Hands the Hardware Engineer the Keys, Vertiv Takes a 400bp Bow, Intel Drafts Its Own Eulogy

The Lead

Tim Cook is stepping down as Apple’s CEO effective 1 September 2026, and the Board chose John Ternus, SVP Hardware Engineering, ahead of the obvious services and operations candidates. That choice is the signal — not the transition itself, and not the standard bromides about continuity. A board that wanted to defend the services-margin story picks Eddy Cue. A board that wanted to professionalise an aging operations footprint picks Jeff Williams. A board that picks Ternus is telling you the next decade is about silicon, Vision Pro, on-device AI, and whatever displaces the iPhone — and that the person running the P&L will come from the engineering side of the building.

Falsifiable form: Ternus’s first formal capital allocation decision — disclosed by Q1 FY27 — will tilt towards hardware R&D and away from services-led capital return. If FY27 buybacks expand and R&D stays flat, the thesis was wrong and the choice was about boring continuity.

Threads

Cook to Ternus — the hardware-first reading

The 8-K (17 April, furnished 20 April) gives the bare facts: Cook becomes Executive Chair, Art Levinson moves to Lead Independent Director, Ternus joins the Board. AAPL 8-K, 17 April 2026. What the filing does not give you, but the choice does, is the strategic posture.

Ternus, 50, joined Apple in 2001 and has run hardware engineering since 2021. The previous internal CEO conversation was always assumed to be between Cue (services), Williams (COO), and a handful of AI-native external hires. Picking the engineer is a vote against optimising the existing revenue mix. Cook stays in the room until probably late FY27 as Executive Chair, which buys six to twelve months of overlap and softens the discontinuity for institutional holders.

Read-throughs: TSM gets paid more (Apple silicon roadmap accelerates under hardware-first leadership). NVDA is the question — does on-device Apple Intelligence become a serious enough competitor to NVDA’s edge inference story to matter inside three years? Probably not at the training layer; possibly at the small-model inference layer. AAPL itself: the multiple gets re-rated if Ternus delivers a credible new hardware category, de-rated if FY27 is “iPhone 19 with marginal improvements.”

Vertiv prints the floor — +30% / +400bps / +153% OCF

The Q1 numbers from the 22 April 10-Q answered the only question worth asking after Issue 01: was the $15B backlog disclosed last quarter a ceiling that needed a reality-check, or a floor under accelerating demand? It was the floor.

• Net sales $2,649.5M vs $2,036.0M prior-year, +30.1%
• Gross profit $999.7M (37.7%) vs $686.5M (33.7%) — +400bps
• Operating cash flow +153% YoY
• Americas segment alone +53%

VRT 10-Q, 22 April 2026.

The 400bps gross-margin expansion is the more important number than the revenue line. You don’t take 400bps on a backlog that customers can substitute away from — pricing power of that magnitude only happens when the order book is locked, the lead times are extending, and the customer universe is structurally hyperscaler-and-neoclouds. Vertiv’s own response is to raise FY26 capex to $425–525M to convert backlog faster, which is the company telling you that the demand curve isn’t levelling off.

Read-through: DLR (announced Q1 earnings same week, 23 April 8-K) and EQIX both benefit from the same demand vector — colocation operators are the order pipeline. ANET reports next; if their data-centre switching revenue mirrors VRT’s pace, the read-through becomes a swept-floor for the whole AI-infrastructure stack.

Lab-calibrated confidence: future_bet × bullish setups in our historical sample show **+2.5% mean direction-adjusted alpha at 20 days, 57.5% hit rate (n=365)*. This thread sits on the strongest historical calibration of any in this issue.

Intel drafts the eulogy for 14A (lab-cautioned)

This is the buried lead of the week. Page-deep in Intel’s 24 April 10-Q risk factors:

“If we are unable to secure sufficient committed demand for Intel 14A through product design wins with potential significant external customers and our Intel products roadmap, we face the prospect that it will not be economical to develop and manufacture Intel 14A and successor leading-edge nodes on a go-forward basis. In such event, we may pause or discontinue our pursuit of Intel 14A.”

INTC 10-Q, 24 April 2026, risk factors section.

This is the first time Intel has put in writing the option of exiting leading-edge node manufacturing entirely. The condition is “significant external foundry customers” — i.e., the AAPL/QCOM/NVDA/AMD type wins that have not materialised in the post-18A pipeline. If 14A pauses, the IDM premium that the market still attaches to Intel compresses materially, and the post-18A roadmap moves to TSMC like everyone else’s.

Layered on top: Scott Gawel resigned as Corporate Vice President / Chief Accounting Officer on 24 April (8-K filed 25 April), with the CFO assuming the role. CAO departures one day after earnings are not benign. INTC 8-K, 25 April 2026.

Layered on top of that: Intel disclosed it was named near the top of an Iranian retaliation target list published in late March 2026 — relevant because a “significant portion” of Intel 7 production runs through the Israel fab, and Intel is self-insured against war damage but not insured for war-driven business interruption. INTC 10-Q, 24 April 2026.

Lab-calibrated confidence — read with care. This thread is built on three bearish-direction signals (two strategic_shift × bearish, one asymmetric_info × bearish). In our 4,669-row historical sample, strategic_shift × bearish setups deliver -4.25% mean direction-adjusted alpha at 20 days, 35.6% hit rate (n=118) — the single worst setup in the dataset. Asymmetric_info × bearish delivers **-2.86% mean, 46.8% hit (n=156). The system is structurally bad at calling bears in this universe.

So treat this as a hypothesis with a falsifiability date rather than a directional call. The 14A language stands or falls on whether Intel discloses an external customer commit before its FY26 10-K (filed late January 2027). If they do, it was a routine disclosure boilerplate. If they don’t, the pause-or-discontinue language gets escalated to a definitive statement, and the multi-trillion-dollar question — who makes the chips below 18A? — becomes the only question.

Tesla quietly joined the hyperscaler capex club

The 23 April 10-Q has the line that resets the modelling exercise:

“We currently expect our capital expenditures to be in excess of $25 billion in 2026, driven by our AI initiatives, including investments in compute infrastructure and data centers…”

TSLA 10-Q, 23 April 2026.

The $25B+ figure with explicit AI attribution puts Tesla in the same sentence as Microsoft, Google, Meta, Amazon, Oracle and CoreWeave for the first time. Q1 capex was $2.49B, +66% YoY — the FY guide isn’t a plan, it’s already a run-rate. Cortex (the GFTX training cluster) gets the bulk. The 10-Q also discloses an April 2026 AI hardware acquisition with the dollar value XBRL-masked.

The auto-only multiple is becoming insufficient to model TSLA. The re-rating question (auto multiple → AI infrastructure / platform multiple) is a 24-month thesis, not a quarter. Q1 also delivered auto gross margin 21.1% (vs 16.2% prior year) and energy gross margin 39.5%, with operating cash flow $3.94B — i.e., the AI capex is being self-funded.

Read-throughs: NVDA gains another tier-1 customer for the GPU portion of Cortex. VRT and other power/cooling specialists benefit from the data-centre buildout. TSM benefits because Tesla’s in-house silicon ambitions still need a foundry partner.

CoreWeave re-taps at 102 — $7.75B in eight days

On 21 April 2026, CoreWeave issued an additional $1B of 9.75% Senior Notes due 2031 at an issue price of 102.000% of par. CRWV 8-K, 21 April 2026. The original tranche on 14 April priced at par. Re-tapping at a premium one week later means the secondary market has firmed up — buyers are paying more for the same coupon than the primary holders did.

Stacked with the 14 April convertible ($4B), the 14 April senior ($1.75B), and the 15 April Jane Street equity placement ($1B), the three-week capital raise is now $7.75B. Use of proceeds includes “repayment of outstanding indebtedness” — they’re refinancing the old high-yield while adding new, an indication that the tighter spreads are the point of the exercise rather than incremental capacity.

Lab-calibrated confidence: asymmetric_info × bullish setups deliver *+0.68% mean direction-adjusted alpha at 20 days, 52.1% hit rate (n=117) — modest edge. The base rate says: this is the right direction but not a high-conviction trade. The trade is in how it confirms the broader AI-capex financing thesis, not in CRWV itself.

SMCI’s BDO ratification kills the audit-tail

Annual meeting on 15 April; the 20 April 8-K reported the votes. BDO USA P.C. ratified as auditor with 98.8% support of voted shares (426M for, 4.6M against, 1.1M abstain). SMCI 8-K, 20 April 2026.

The audit-credibility tail-risk that has overhung the stock since the 2024 EY resignation is now removed. Doesn’t make the company cheap; does make it institutionally re-investable. Class I directors re-elected with mixed results (Charles Liang strongest at 234.6M / 45.7M, Sherman Tuan weakest at 172.3M / 108.0M against). The 2020 equity plan was amended (15M new shares) with 81% support — read as routine.

Watchlist Updates

Ticker	Direction	Driver	Live score (salience)	Lab-calibrated reading
INTC	bearish	14A discontinuation language; CAO departure; Iran target list	96.59	Caution — bearish strategic_shift has -4.25% mean alpha (n=118). Treat as hypothesis.
INTC	bearish	Iran retaliation target list disclosure	96.59	Caution — bearish asymmetric_info has -2.86% mean (n=156).
TSLA	bullish	$25B+ AI-attributed capex	95.48	Strong — bullish future_bet has +2.5% mean (n=365).
VRT	bullish	Q1 +30% sales, +400bps GM, +153% OCF	94.39	Strong — same calibration as TSLA.
CRWV	bullish	$1B re-tap at 102, $7.75B 8-day total	82.03	Modest — bullish asymmetric_info +0.68% mean (n=117).
AAPL	neutral	Cook → Ternus transition	79.37	Best calibration in issue — neutral strategic_shift +3.03% mean, 58.6% hit (n=111).
INTC	bullish	DCAI revenue +$926M YoY, server ASPs +27%	mid-tier	Modest — partial offset to the bearish thread above.
SMCI	bullish	BDO ratification, audit-tail removed	mid-tier	Modest.
DLR	neutral	Q1 earnings furnishing	mid-tier	Read in conjunction with VRT.

The Week Ahead

The big-tech triple-header lands on Wednesday 30 April after-close: MSFT (Q3 FY26 results), META (Q1 FY26), GOOG (Q1 FY26). AMZN and AAPL report Thursday 1 May after-close.

The capex commentary is the ball to watch. With Tesla now in the hyperscaler club at $25B+, and Vertiv’s order book confirming the demand side, the market is positioned for combined FY26 capex from the four hyperscalers (MSFT/META/GOOG/AMZN) of approximately $320–360B. Anything below $80B annualised per name reads as deceleration relative to the VRT/CRWV/TSLA inputs — and would be the first credible deceleration signal since the AI capex cycle began. Anything above $90B per name extends the runway by another two quarters at minimum and strengthens the read-through into VRT, ANET, EQIX, DLR, and the second-derivative power names.

What would change the Intel thesis: an external-customer commit announcement before earnings on 25 April (already past) — or, more likely, an explicit walk-back of the 14A language on the conference call.

Methodology and disclosures

Filings Intel Digest is a bona fide financial publication. Nothing here is personalised investment advice. The editor may hold positions in companies discussed; current positions and policy are at /about. All claims are sourced to publicly filed documents linked inline. Backtested or historical figures are direction-adjusted and calibrated against the signal lab; calibration is partial — treat any forward statement as a hypothesis, not a forecast.

Salience scores in the watchlist table are not edge — they are a prior on attention. Direction-adjusted alpha calibration against historical outcomes shows bearish-direction signals in this universe have negative mean alpha at 20 days (-2.5% mean, 45.9% hit), so any bearish thread in this issue is presented as a falsifiable hypothesis rather than a directional call. Bullish future_bet setups (the VRT and TSLA threads) carry the strongest historical calibration in our sample (+2.5% mean alpha at 20 days, n=365).

Frozen as of close-of-trading 24 April 2026.