Setup

Butterfly Network (BFLY) makes handheld ultrasound devices built on a proprietary semiconductor chip (Ultrasound-on-Chip) that replaced piezoelectric transducers. The Q1 2026 10-Q filed April 30 is the first quarter where a structurally distinct second business — Butterfly Embedded, which licenses the chip and co-develops applications with non-competitive partners — has reached enough scale to matter. Embedded revenue jumped from $2.3M Q1 2025 to $5.7M Q1 2026 (+148% YoY). The stock fell ≈23% the week before the print. We're looking now because Embedded's scale and contract structure have changed without the multiple framework changing.

What the filing says

Q1 2026 revenue $26.5M (+25% YoY). Gross margin 68.9% (+590bps); software/services GM 84.1% vs 71.4%. Operating loss improved $4.6M on revenue +25% / OpEx +1%. Cash $138M, ≈10-11 quarters runway. One customer exceeded 10% of revenue — concentration that didn't exist Q1 2025.

Two facts the filing does not name explicitly but that map to disclosures already on the public record:

• The >10% customer is Midjourney. A November 18, 2025 8-K (Item 1.01) discloses a five-year Co-Development & Licensing Agreement: $74M total — $15M upfront, $10M/yr ratable license ($2.5M/qtr), up to $9M milestones, future chip royalties. Q4 2025 Embedded ($6.8M) and Q1 2026 ($5.7M) reconcile to this contract. Recurring multi-year structured revenue, not a one-time milestone.
• Forest Neurotech (Oct 2023, $20M, BCI implant) is the structural template. Same five-year, upfront-plus-license-plus-milestone-plus-royalty scaffold. Midjourney is 3.7× Forest. CEO commentary on the Q1 call: 9 partners signed, ≈40 in pipeline (pharma, BCI, neuromod, robotics).

Counterweight: RPO declined $8.2M sequentially ($99.6M → $91.4M); new bookings replaced ≈29% of recognized revenue. Hardware units +4.9% YoY against handheld POCUS sectoral CAGR 5.75% / portable 9.0% — modest share loss to GEHC Vscan Air SL and PHG Lumify. CEO, CFO, GC, COO sold during the quarter; founder Rothberg adopted a 10b5-1 plan March 13 authorizing up to 7.8M shares (≈$32M) starting July 14.

What the market thinks

At $4.16 / ≈11× trailing revenue with -50% operating margin, BFLY is at or above unprofitable medtech multiples. Not cheap on conventional screens.

Sum-of-parts:

• Hardware $80M run-rate × 1-2x = $80-160M
• Embedded $23M run-rate × 8-15x (chip-IP comp) = $184-345M
• Cash $138M
• Per share: $1.51-2.42 LOW / $2.42-4.15 HIGH

Current $4.16 is at the HIGH end — but only if Embedded gets full chip-IP multiple, which it isn't getting today.

Back-solving market-implied probabilities of the 4-state factor scenario (durable_platform / midjourney_anchor / lumpy_single_deal / midjourney_walks): market ≈20/45/27/8. Our view: 30/40/22/8. Edge is +10pp on the durable_platform tail — ≈$0.42/share, ≈10% of current price.

Options confirm: July IV 108% > May IV 94% (term structure inverted), calls 3.1× puts, max pain $4.50.

Why the gap exists

1. The Midjourney 8-K was buried. Item 1.01 binding agreements get less attention than 7.01 press releases. Linking the November 2025 contract terms to the Q1 2026 customer-concentration footnote requires reading both filings together. Sell-side hasn't synthesized this in published notes.
2. No public chip-IP-into-vertical-applications comp set exists. ARM is software-IP. CDNS/SNPS are EDA. NVIDIA CUDA is different scale. BFLY would be creating the medtech chip-as-platform category — without a comp set, analysts default to hardware multiples for both businesses.
3. Bear signals are loud and on-screen; bull signals are inferred. Insider selling, RPO decline, hardware deceleration are visible. Embedded's structural shift requires reading 8-Ks plus call commentary plus platform hires (David Horsley ex-TDK InvenSense / Chirp Microsystems for chip-as-platform commercialization; Caroll Neubauer ex-B. Braun USA for medical-device commercial governance).

Risks (ranked)

1. Midjourney walks or pauses. ≈8% probability × -58% E[r] = -4.7% expected return drag. Private gen-AI lab at ≈$500M ARR per The Information; burn rate unobservable. Largest single-counterparty risk.
2. Q2 Embedded prints <$4M. Reframes Embedded as 1-2 lumpy milestone deals. Multiple compresses to hardware comp; fair value drops to $1.51-2.42.
3. Hardware deceleration accelerates. Q2 <+3% YoY = continued share loss; Embedded must do more lifting alone.
4. Founder + C-suite supply collision. Rothberg 10b5-1 opens July 14 — 30 days before Q2 print. Forced supply could dominate the immediate window.
5. Embedded comp set never establishes. Without a partner commercial product launch (timeline unknown), the chip-IP multiple stays hand-anchored.

Catalysts

• mid-May: Form 4 cadence read.
• June – early August: Possible 8-K Item 1.01 (10th partner; CEO guided 11+ by mid-2026 vs 9 signed at Q1 call).
• July 14: Rothberg 10b5-1 plan opens (negative supply).
• mid-August — Q2 2026 10-Q: THE event. Resolves Embedded revenue, RPO direction, hardware units, customer concentration.
• mid-November — Q3 2026 10-Q: Durability test #2; second 10%+ customer disclosure window.

What would change our mind

Bearish: Midjourney 8-K Item 1.02 → durable_platform probability collapses. Auditor change or restatement on Embedded over-time recognition. Q2 Embedded <$4M with RPO continued decline. Capital raise announcement at depressed price.

Bullish: 10th+ Embedded partner 8-K before July. Apollo chip tape-out with named customer. Insider Form 4 code P (purchase). Midjourney announces commercial AI imaging product requiring BFLY chip. Second customer >10% disclosed.

Frame

Probability-weighted E[r] +7.7% over 98 days (≈30% annualized), Sharpe ≈0.43. Concentrated single-driver bet (Embedded ≈70% of expected return). Doorway state with a sharp dated catalyst. Modest edge, not exceptional. Q2 print is the size-up event. Pre-Q2, this is a starter, not a pound-the-table.

Evidence

LR signal: 1.3 (bullish — market underweighting; modest, not exceptional).