Semiconductor Ultrasound
A medical device company repositioning as a chip platform — licensing its Ultrasound-on-Chip technology to partners building things Butterfly would never build itself.
Personal notes published for reference. Figures are sourced from company filings, releases, and press coverage as of the date above — the interpretation is mine and may be wrong.
CEO Joe DeVivo, CFO John Doherty, with David Horsley (SVP Innovation) on the technology side. The strategic call that matters is DeVivo's: recognizing the chip was worth more as a licensable platform than as a component in Butterfly's own handheld probes. That reframing is the entire current thesis.
This is the part that changed. Point-of-care ultrasound is a real but bounded market. Butterfly Embedded — the licensing and co-development arm — opens the chip to any third party building ultrasound-enabled hardware, in and beyond healthcare. The TAM stops being 'handheld probes sold' and becomes 'devices containing our silicon.'
Still the weakest criterion, but genuinely improving. Q3 2026 gross margin set another record at 73% (from 64% a year earlier) and adjusted EBITDA loss shrank to $1.2M from $8.1M — both driven by the Embedded mix shift. Cash burn was only $2.4M in the quarter against $122M of cash. Butterfly still loses money; it's no longer obviously bleeding out.
Q3 2026 revenue was $32.6M, up 52% YoY and 9% above the top of guidance, with Embedded revenue of $11.1M against $1.0M a year earlier. Full-year guidance has now been raised three times, to $123–126M. Two Embedded partners also went public with working hardware within a week of each other in June 2026 — external proof the platform strategy produces something other than press releases.
Butterfly proved its Ultrasound-on-Chip technology in point-of-care imaging with the iQ series, now on its third generation. Butterfly Embedded (formerly Octiv) licenses that same semiconductor platform to third parties.
Selling probes means competing on distribution, regulatory clearance, and sales headcount in a mature medical device market. Selling chips and licenses means collecting fees and per-unit revenue while someone else takes the commercialization risk. The margin profile and the capital intensity are not comparable — and the addressable market is set by other people's imagination rather than Butterfly's product roadmap.
Shares initially jumped sharply (reports put it around 23–24%) on the Q2 beat and second consecutive guidance raise, then gave a chunk of that back in subsequent sessions to around $7.11 — the same push-pull pattern as the June Midjourney reveal below.
Butterfly plans to launch its first home and community care program in October 2026, expected to contribute multiple millions of dollars in 2027 — small next to Embedded, but a second leg that doesn't depend on Midjourney or Aleph Neuro's execution.
The generative-AI company signed a five-year co-development and licensing agreement in November 2025, then revealed what it was building in June 2026: Midjourney Medical and the Midjourney Scanner, a full-body ultrasonic computational tomography device.
$15M one-time fee, $10M annual license fee paid quarterly across the five-year term, up to $9M in milestone payments, plus revenue sharing on hardware incorporating Butterfly chips and separate payments for the chip purchases themselves. Exclusive within a specified field of use.
Midjourney is targeting 5,000 “Midjourney Spas” at roughly 10 scanners each, with a first San Francisco location expected late 2027 and next-generation scanners in 2028. At ~40 chips per scanner, 50,000 units implies chip volumes on a completely different order than the probe business. Butterfly has said later versions may use more modules, not fewer.
Shares jumped sharply on the June reveal — reported moves of 31% to 56% depending on the window — to a 52-week high around $8. The stock then gave a chunk back, which is worth remembering: the reaction priced in a rollout that has barely started.
The scanner is an early-stage prototype, not a cleared medical device. FDA approval would be required for medical use; initial applications focus on body-composition mapping. William Blair flagged open questions on how this actually plays out. The $74M ceiling is a ceiling, not a forecast.
A research lab building non-invasive brain interfaces, and the second Embedded partner to go public with results — a week after Midjourney.
In June 2026 Aleph announced what it describes as the highest-resolution 3D images of the human brain ever captured from outside the skull, using Butterfly's Ultrasound-on-Chip platform — framed as “MRI-level detail” without drilling or an MRI machine's bulk.
Founded by CEO Marley Xiong with co-founders Raffi Hotter and Lev Chizhov. Early-stage, and explicitly a research lab rather than a commercial product company — so there is no near-term revenue attached.
Two independent partners, two completely different applications — full-body tomography and brain-computer interfaces — from the same silicon, within eight days. That's the strongest available evidence that Embedded is a genuine platform rather than one bespoke deal. Brain interfaces are also a category where the ceiling is nowhere in sight.
The highest-profile Embedded partner yet: on Sept 1, 2026, Sam Altman's brain-computer-interface startup picked Butterfly as its exclusive chip-based ultrasound supplier.
Merge Labs launched in January 2026 with $252M in seed funding from OpenAI, Bain Capital, Interface Fund, Fifty Years, and Gabe Newell, co-founded by Sam Altman alongside researcher Tyson Aflalo. It's one of the two best-funded non-invasive BCI efforts in the US — the other, Forest Neurotech, works from the same underlying ultrasound-BCI science via Mikhail Shapiro's lab.
Butterfly becomes Merge's exclusive CMOS-MEMS chip-based ultrasound partner, licensing the Poseidon family of Ultrasound-on-Chip technology under a multi-year Butterfly Embedded agreement. Terms include upfront payments, milestone-based license payments, hardware purchase commitments, access fees, and ongoing royalties — no dollar figures were disclosed. Butterfly retains the right to license the same technology to other companies in the field.
This is now two independent, well-funded groups building non-invasive brain interfaces on Butterfly's silicon — and one carries Sam Altman's name and OpenAI's capital, a different order of credibility than a research lab. It's the strongest evidence yet that Embedded is a real multi-vertical platform rather than one lucky partnership (Midjourney) plus a research curiosity.
No dollar figures were disclosed, and Merge is a brain-computer-interface startup at the earliest stage — no product, no regulatory pathway started, no revenue timeline. Read this as a credibility and optionality signal, the same caveat that already applies to Aleph Neuro.
On Sept 22, 2026, Needham's David Saxon initiated coverage at Buy with a $10 price target, built on a sum-of-the-parts model (roughly 16x 2027 EV/sales) that treats the chip — not the probe — as the business, and names Merge Labs as evidence the silicon travels outside traditional ultrasound. The stock hit a new 52-week high the next day, and the consensus target has since climbed from $7.06 to roughly $9.30–10.50 depending on the tracker. That's the first time the sell side has formally priced the platform thesis this page is built around.
A week later Evercore ISI initiated at Outperform with a $15 target, the new street high. It models $550M+ of revenue by ~2030 and positive EBITDA around 2028, and makes the point I've been making here: this is a multi-partner platform, not a Midjourney trade. Two buy-side initiations in a week took the consensus to roughly $11.40.
Mendaera's Focalist — a handheld robot that guides needles for vascular access, nerve blocks, and organ access, already FDA 510(k)-cleared — is now commercially available with a customized Butterfly iQ3 probe. Midjourney, Aleph, and Merge are all years from a product. Mendaera is the first Embedded partner shipping something a hospital can buy today, which makes it the first test of whether Embedded becomes recurring hardware revenue rather than license fees.
Mike Noonen joins the board on Oct 9. He's a 30-year semiconductor executive who ran MixComm (sold to Sivers) and Swave Photonics. It's a small item, but it fits the pivot: a probe company wouldn't add him; a chip licensor would.
On Oct 6, 2026, Bridge Neurotech came out of stealth with a multi-year R&D agreement under Butterfly Embedded. Bridge is building a fully non-invasive brain-computer interface that images brain activity through the skull with ultrasound and decodes it with AI. It's backed by General Catalyst, has raised about $13.5M, and is enrolling its first human study in Redwood City. Terms follow the Merge Labs pattern: technology-access and milestone payments plus hardware purchase commitments, with no dollar figures. That makes three independent groups (Aleph Neuro, Merge Labs, Bridge) betting that ultrasound is the route to non-surgical BCIs, all on the same chip. Bridge is much smaller than Merge, so read it as more evidence for the platform, not as revenue.
Reported Oct 29. Another clean beat and a third guidance raise, but the stock fell anyway (about 5% premarket from a $7.81 close). The quarter answered "is the pivot real?" and left open "what comes after Midjourney?" — which is the question every analyst asked.
Revenue beat the top of guidance by 9%; U.S. revenue rose 71% to $27.5M. Software and services (where Embedded fees sit) is now more than half of revenue at $17.0M. Adjusted EBITDA loss was just $1.2M, and full-year guidance moved up to $123–126M with the EBITDA-loss guide narrowed to $12–15M. FedRAMP authorization came through, which opens federal buyers in 2027. Management still targets EBITDA and cash-flow breakeven in 2028 and repeated the $550M+ revenue goal for 2030, which it says assumes only about a quarter of Midjourney's potential.
The Q4 guide of $31–34M is flat with Q3 and only ~3% YoY at the midpoint, because Q4 2025 included the Midjourney upfront. International fell 5% to $5.1M; management says it leaned too hard on distributors selling to individual clinicians and is switching to the enterprise playbook that works in the U.S. The core probe business grew just 5% ($21.5M vs. $20.5M). And remaining performance obligations, the contracted backlog, slipped from $99.6M at year-end to $84.2M: Butterfly is recognizing signed deals faster than it is signing new ones of the same size.
This was the theme of the call. DeVivo kept coming back to it: "a very large pipeline" in Embedded, outbound targeting of large incumbents with mature products (who could produce revenue "very quickly"), at least one non-healthcare conversation, and "one deal can completely change the complexion of the company." His answer on timing was simply "It'll happen." Nothing was named, sized, or dated, and management guides only to signed contracts, so a Midjourney-scale deal would arrive as an 8-K and a guidance raise rather than something already in the numbers. Read it as real pipeline confidence from a CEO who has delivered on Embedded so far, but it is still speculation until signed — and it's what the stock is pricing.
The 10-Q says two customers each accounted for more than 10% of revenue in Q3 (none did in Q3 2025), and two customers are also over 10% of receivables. They aren't named, but given the timing and the CFO calling Merge Labs "a big contributor," Midjourney and Merge are the likely pair. That means roughly a fifth or more of quarterly revenue, at ~100% gross margin, rests on two private, pre-product partners. Midjourney's floor is contractual ($10M a year in license fees through 2030), but milestones, chip purchases, and Merge's payments depend on those companies continuing to fund their programs. If either pulled back, the hit would land on the highest-margin line, and the EBITDA progress would reverse faster than revenue falls. This is the biggest reason the next Embedded deal matters: it's diversification, not just growth.
Home and community care starts scanning in November with a congestive-heart-failure program in nursing homes (DeVivo cites $10–12K saved per avoided readmission); meaningful revenue is a 2028 story. Mendaera's Focalist is launched but meaningful revenue is "a year or two" out. A next-generation P5.1 chip comes in early 2027, and Butterfly plans to enter the ~$8B cart-based ultrasound market in 2028. Butterfly filed a $150M shelf and a $150M ATM, called "housekeeping" with "no immediate need for cash" — but an ATM on a story stock is a dilution tool worth watching. 2027 guidance comes at JP Morgan in January.
Butterfly's cash-flow weakness makes this one worth watching closely — institutional buying into a story stock is a different signal than institutional buying into a name that can already fund itself.
FMR LLC (Fidelity) raised its stake to roughly 11.8%, BlackRock holds about 7.8%, and Larry Robbins/Glenview Capital about 6.2%. That's a mix of large passive managers and at least one concentrated, activist-style holder — not a purely mechanical index flow.
The 30–56% move on the Midjourney reveal happened before this data was fully captured in filings; some of the increase is likely funds buying the news rather than the underlying platform thesis. Worth re-checking next quarter's filings to see if the position holds through the pullback.
Ownership percentages for this name vary meaningfully by source and date — a byproduct of a small, actively-traded float and a share count that moves with warrant exercises and financing activity. Treat the headline number as approximate.
Q3 was another clean beat, but the stock fell on it. The market has stopped paying for execution on signed deals; it now wants the next large Embedded contract and proof that revenue isn't resting on two private partners.
If this plays out: ~$5
The stock has already run to ~$9.40 — within a few percent of Needham's new $10 target and the raised consensus — so the upgrade cycle may be priced in rather than ahead of it. The Merge Labs deal — announced Sept 1 — disclosed no dollar figures at all, same as Aleph Neuro. Two unnamed customers were each over 10% of Q3 revenue, almost certainly Midjourney and Merge. Contracted backlog shrank from $99.6M to $84.2M this year, and the Q4 guide is flat. The single biggest relationship, Midjourney, is pre-revenue at the product level: a scanner prototype, not an FDA-cleared device, with the first physical location not expected until late 2027. A famous logo isn't the same as a booked contract, and the stock has already proven it can give back 20+ points of a guidance-driven pop in the same news cycle.
If this plays out: ~$10
Q2's narrowing losses and record 71% gross margin show the pivot in the income statement, not just partner announcements, and Merge Labs is now the third Embedded partner in four months — each in a different application (imaging, brain interfaces, and now a second, better-funded brain-interface effort). None of that is disclosed revenue yet. Expect continued outsized moves around every Embedded update and FDA milestone, in either direction, for a stock this small and headline-driven.
If this plays out: ~$15
If Embedded scales the way three independent, well-capitalized partners now suggest it can, Butterfly stops being a struggling hardware seller and becomes a licensor sitting underneath multiple companies' hardware roadmaps — at the 71% gross margin Q2 just proved is achievable. Sam Altman and OpenAI's capital behind Merge Labs is a credibility signal the Aleph Neuro deal alone didn't carry. Evercore's $15 target — the new street high, built on $550M+ of revenue by ~2030 — is what that looks like priced in.
About $7.40 after Q3 (it closed at $7.81 the day before and fell ~5% premarket on the print), well off the $10.21 52-week high, with roughly 267M shares outstanding. Before earnings it was $8.25 at the Oct 9 close, down about 9% that week despite the Bridge deal. Needham (Buy/$10, Sept 22) and Evercore ISI (Outperform/$15) lifted the consensus target from $7.06 to roughly $11.40, so the gap to consensus is now ~38%. Insiders have been sellers into the run: CEO Joe DeVivo sold 35,790 shares (~$359K) on Oct 6 under a pre-arranged 10b5-1 plan, after an officer's ~10K-share Form 144. The same source flags the stock trading at 17.5x sales against a "fair" 5.9x — a reminder that undisclosed-terms partnerships can move the narrative faster than the multiple. The scenario targets above are my own rough, directional estimates — not formal price targets — anchored to the sell-side range rather than a full model.
Butterfly's upside now runs through other companies' execution. Midjourney's spa rollout, FDA pathways, and Aleph's and Merge's research all sit outside Butterfly's control.
Two customers were each over 10% of Q3 revenue and two are over 10% of receivables, up from none a year ago. Both are almost certainly early-stage private partners. One pulling back would hit the highest-margin revenue first. Until another large Embedded deal lands, this is the risk that matters most.
A new $150M ATM and $150M shelf went live with Q3. Management calls it housekeeping, but selling stock into rallies is the obvious use.
First Midjourney location late 2027, next-gen scanners 2028, 50,000 units by 2031. The meaningful chip volume is years out; the stock has already moved on it.
The handheld probe business still loses money. Compass AI and Butterfly Garden are attempts to build higher-margin software revenue around it, but the base business needs to stop bleeding regardless of what Embedded does.
This is a small cap that moved 30–56% on a single partner announcement and 19% down in another session. Headline-driven in both directions.