Phase Space AI

01b Unpublished Scoping and TAM

Intuitive Surgical [ISRG]

Intuitive Surgical [ISRG] — Unpublished Scoping & Bottom-Up TAM

Task 1 extension, produced under investment-memo v1.4.0

As of 27 July 2026 | Price $357.16 (Alpaca SIP last at pull; settled close $356.83 — 9bp, immaterial, and $357.16 is retained throughout for consistency with the model and charts)

This document exists because of one critique: what LLM-driven research systematically lacks is scoping what is not published. Everything here is built from corpora outside the filings — earnings-call transcripts counted over time, ClinicalTrials.gov, PubMed, and the FDA 510(k) database — and then sized from the bottom up.

---

Part 1 — The transcript corpus: what was retrieved, and what was not

Stated plainly, because the rule is not to imply more coverage than was retrieved: the target was 2022Q1 onward. 2022 was not obtained. Every "first-ever" or "versus all history" claim below is bounded by 2023Q1, and is written that way. The eight quarters added are the minimum the methodology requires, and they were enough to overturn two claims the previous six-quarter window had supported — which is the argument for extending history, made concrete.

A coverage artifact that materially distorts raw counts

Two transcripts in the cache are substantially shorter than the rest, with the Q&A largely missing:

This is not a footnote — it is the reason one of the prior memo's headline findings was wrong. Every count below is therefore reported per 10,000 words as well as raw, and the two short quarters are flagged on every chart (Exhibits 36 and 37).

---

Part 2 — Mention frequency run GENERATIVELY (the required order of operations)

Per references/mention-frequency.md, this was run before revising any view, in two passes:

Pass 1 — unsupervised. No term list. Every unigram and bigram in the last three quarters was scored against its own rate in 2023Q1–2024Q3, per 10,000 words, and the largest movers in both directions were listed with no interpretation attached. This is the pass that can find what neither the analyst nor the reviewer thought to look for.

Pass 2 — supervised. A deliberately broad 79-term set (every product line, every major geography, every named competitor, every disclosed procedure category, the channel vocabulary, and the standard business words) counted across all 14 quarters, split into prepared remarks versus Q&A.

Only after both passes were complete was any thesis revised.

The open-question list, exactly as generated (no interpretation attached)

This is the raw output of Pass 1, before any investigation:

Investigating each against an independent corpus

---

Part 3 — What the extended history actually changed

Correction 1 — endoluminal is NOT a first-ever mention

The prior memo reported endoluminal receiving its "first-ever mention" in 2026Q2. On the six-quarter window that was true. On fourteen quarters it is false.

In 2024Q3 Gary Guthart used it in prepared remarks about Ion: "Turning to our endoluminal system, we launched Ion in Q3 of 2019." The word is ISRG's standing descriptor for Ion.

What is genuinely new in 2026Q2 is endoluminal applied to the GI tract, not the word. That is a narrower and more defensible claim, and it is the one this memo now makes. This is exactly the failure mode references/mention-frequency.md warns about — "State the history window on every table. Six quarters is a useful sample; it is not 'all history.'"

Correction 2 — Ion is FLAT, not emerging. The prior series was an artifact.

The prior memo reported Ion running 4 → 10 → 25 → 18 → 17 → 18, read as "a sustained, management-driven priority" ramping hard. That ramp does not exist. The series started at 2025Q1 — the 2,400-word truncated transcript. Normalised across all 14 quarters:

* truncated transcript

Mean 24.2 per 10k, standard deviation 6.0, no trend. Ion is a stable, consistently-prioritised platform — which is a genuinely different thing from an emerging one, and it matters for the dissolved duration-variant test (old Gate 2B), where the second evidence leg requires a term that has moved.

Corroborating negative, disclosed and independent: Ion placements fell 28% in FY2025 (195 vs 271 in FY2024). Management's explanation is "a joint focus with our customers on increasing utilization." Both the language series and the unit series say the same thing: Ion is being harvested, not launched.

What the decays say

• growth fell from 155 to 102 per 10k — the largest single decay in the corpus. Management is using the word "growth" a third less often than it did in 2023. That is not a metric; it is tone, and it is consistent with the deceleration the numbers show independently.
• china halved (48.7 → 20.5 per 10k) while the disclosed China situation deteriorated. De-emphasis of a deteriorating market is the less charitable reading and is worth naming.
• bariatric peaked in 2023Q3 (24.1 per 10k — the GLP-1 panic quarter) and has decayed to 3.7 while the bariatric business is still declining high single digits. The problem did not go away; the discussion did.
• backlog and capacity have gone to roughly zero since 2024Q3. The 2022–23 supply-chain and capacity-constraint era is over. Neutral-to-positive, and it removes a bull argument that no longer applies.
• prostatectomy, hysterectomy and colon resection have left the vocabulary entirely, while cholecystectomy, hernia and appendectomy persist. The mature US core is no longer discussed; the growth conversation has migrated to benign, high-volume general surgery — which is precisely the category the extended-use programme discounts and the ASC channel targets. Three independent findings converge on the same place.

The finding neither I nor the reviewer anticipated: the value/access cluster

Six terms rose together, and reading them separately misses the point:

cost in 2026Q2 is at its highest level in the entire 14-quarter window, and two thirds of it is in prepared remarks. Dave Rosa's 2026Q2 script names the strategy explicitly:

"These innovations are designed to help providers advance key health care objectives including clinical outcomes, patient experience, provider experience, access, and affordability. We are also innovating across manufacturing and supply chain operations to better serve value-sensitive markets. Programs such as XiR and EUP are intended to expand access to robotic-assisted surgery..."

Read as one cluster, this is a company publicly repositioning from a premium-only capital vendor to a segmented one with a deliberate value tier. XiR (cheaper hardware) and EUP (cheaper consumables) are the two instruments of the same strategy, announced in consecutive quarters. The prior memo treated EUP as an unexplained self-inflicted revenue headwind. It is not unexplained — it is one half of a coherent value-segmentation strategy whose other half is XiR, and that reframing is the single most useful thing this pass produced.

It cuts both ways, and both must be said: it is evidence of a real channel-expansion opportunity (Part 4, Vector D) and evidence that management sees enough price pressure to act pre-emptively on both hardware and consumables at once.

A genuinely undisclosed programme found in the registry: "JURA"

NCT07259759 — "JURA Study – Observational First-in-human Feasibility Study." Condition: Imaging. Intervention: Visualization / Imaging. Brief summary, in full: "Evaluate the functionality of the JURA System v. 1.0." n=20. COMPLETED 2025-10-31.

• "JURA" appears zero times across all 14 quarters of transcripts.
• It appears in no filing this workup examined.
• It is not da Vinci, not Ion, not SP.
• A completed first-in-human study means a working device exists and has been used in patients.

This is the purest example of the technique working: a company-sponsored, completed, first-in-human study of a named system that management has never once mentioned on a call. It is not sized here — an imaging or visualisation platform of unknown scope cannot be sized from a one-line registry summary, and pretending otherwise would be exactly the narrative-manufacturing this framework rejects. It is logged as a monitored item with a dated trigger (Task 6) and is offered to the reviewer as the highest-value open question this pass produced.

Correction 3 — the prior memo's GI corroboration was a misattribution

The prior memo cited "an IS-001 colorectal trial starting 2026-03 (GI)" as independent corroboration of the endoluminal GI vector. Checked against the registry record:

NCT07346066 — "IS-001 Injection in Patients Undergoing Robotic-Assisted Colorectal Surgery." Condition: Ureter Injury. n=146. NOT_YET_RECRUITING.

IS-001 is an injectable imaging agent for intraoperative ureter visualisation during transabdominal robotic colorectal surgery on da Vinci. It has nothing to do with endoluminal GI endoscopy. The word "colorectal" carried the inference; the condition field refutes it. Corrected. Its removal is what turns the dissolved duration-variant test (old Gate 2B) leg 1 for GI from "thin" to "empty."

---

Part 4 — Bottom-up TAM by vector

Built per references/tam-sizing.md: units × penetration × price/unit × attach rate; every input labelled DISCLOSED, EXTERNAL or ASSUMPTION; no top-down industry-report headline used anywhere. All arithmetic lives on the model's TAM & Vectors tab and every figure below was read back from Excel after calculation.

Vector D — XiR / ASC (the vector neither the memo nor the reviewer had named)

Why it leads: it is the only vector with a rising prepared-remarks mention series, a cleared 510(k), a disclosed unit count, and a competitor that raised capital specifically to attack the same channel.

What is disclosed:

The units build:

Combined Vector D TAM: $1,443m per year at full penetration — ~11% of FY2027E consensus revenue, and the largest of the four vectors.

Penetration-path sensitivity (US ASC leg):

Time to revenue: FY2026E ≈ $170m (1.4% of revenue); FY2027E ≈ $250–300m (~2%); >5% of revenue not before ~2030–2032.

What would falsify it: ASC placements failing to exceed 27/quarter in any quarter through 2027; XiR placements falling below the 64 printed in Q2 2026; management disclosing an XiR ASP below ~$500k (which would make the channel volume without economics); or CMS declining to narrow the ASC-versus-HOPD reimbursement gap that management named as the binding constraint.

Base rate, named as required: new site-of-care migrations in US surgery have historically taken 10–20 years (ophthalmology, orthopaedics, ENT). Nothing here assumes faster than that; the Base path fills ~68% of the addressable pool in nine years, which is at the fast end of the historical range and is labelled as such.

Vector B — da Vinci cardiac

Years to reach the cleared-pool SAM at the disclosed +39% growth rate: 6.8 → roughly 2033. Management has said cardiac-specific instruments "will take several years to bring to market." Not >1% of revenue before ~2030.

Independent corpus, honestly graded: PubMed "robotic cardiac surgery" grew 6 → 46 papers/yr (2016→2025) — real, but 7.7x over a decade against "robotic bronchoscopy" at 57x. "Robotic mitral valve repair" is essentially flat (8 → 23). One registry trial, n=30, not yet recruiting. FDA dV5 cardiac clearance is real and dated (December 2025). This is a genuine programme at an early stage, not an inflection.

Vector A — Ion / lung

Ion I&A revenue today: $429m/yr. US Ion procedures today: 179,400.

Implied US penetration: 81.5% of the current-practice pool.

This is the disciplined finding, and it runs against the bull narrative. On a mid-range estimate, Ion has already captured most of the US pool that exists at current lung-cancer screening rates. The range is wide (51% at a 350k pool, ~100%+ at 150k), but the conclusion is robust across it: US Ion growth from here must come from pool expansion, not share. The three expansion levers are lung-cancer screening uptake (~16% of an eligible ~14.5m today — a public-health variable ISRG does not control), international (12 countries, early), and moving Ion from diagnosis toward therapy (the ROSE and EBUS programmes).

Ceiling: $1,155m/yr of I&A (US at aggressive screening $770m + international at 50% of US $385m).

Independent corpus — the strongest in the set. PubMed "robotic bronchoscopy": 1 (2018) → 52 (2025) → 57 in a partial 2026 already above the full prior year. "Shape-sensing robotic-assisted bronchoscopy" — the Ion-specific term — 0 (2020) → 27 in partial 2026. Eight registry lung studies including a 1,301-patient US Ion registry and a 1,200-patient European Ion registry (started 2025-09-09). Two Ion 510(k)s cleared in September 2025. Lung Cancer is the #1 indication across all 37 Intuitive-sponsored trials started 2022+.

But the transcript series is flat and placements fell 28% (Part 3, Correction 2). The clinical evidence is inflecting; the commercial series is not.

Vector C — endoluminal GI: deliberately not sized

What is real and dated: a 510(k) submitted in Q2 2026 for a "foundational noncommercial next-generation flexible robotic endoscope system for use in the gastrointestinal tract." That is a specific, verifiable regulatory act.

What the independent corpora return:

Transcript signal: GI = 0 across thirteen consecutive quarters, then 5 in 2026Q2 — all five in Q&A, zero in prepared remarks. Management, verbatim:

"I want to reinforce this remains early. The 510(k) submission is for a noncommercial endoluminal system... We are not ready today to describe timing or size of the opportunity."

The company will not size it and neither will this memo. The underlying non-robotic pools are genuinely enormous — endoscopic submucosal dissection and peroral endoscopic myotomy have 8,337 and 1,689 publications respectively, so the clinical field is mature and real. But there is no robotic literature, no trial, no clearance, no disclosed unit, and no management number. Any TAM built here would be manufactured.

The honest answer, and it is a strong one: the signal is real, and the opportunity is unsizeable today.

The dated triggers that would change that, in ascending order of significance — all free and checkable:

1. A ClinicalTrials.gov registration by Intuitive with a GI or endoluminal condition. Currently zero. This is the single best leading indicator and would precede everything else. Check monthly.
2. FDA 510(k) decision on the pending submission. Submitted ~Q2 2026; typical real-world decision 4–8 months → expect between Q4 2026 and Q2 2027. Checkable at openFDA against applicant "Intuitive Surgical".
3. Management giving the programme a name, a phase, or any number on the Q3 2026 or Q4 2026 call — or GI appearing in prepared remarks rather than only in answers.
4. First peer-reviewed publication on robotic endoluminal GI therapy. Base rate from Ion: FDA clearance in 2019 → 8 publications in 2019 → 52 by 2025. Literature follows clearance by 1–2 years and precedes revenue by 3–5.

---

Part 5 — The required implied-penetration statement

Per references/tam-sizing.md, the rigorous form of EV/TAM is to reverse the question into physical units.

At $357.16, ISRG's enterprise value requires roughly 7.7 million annual procedures at a 5.0x terminal EV/revenue — 2.4x the 3.24 million performed in FY2025, and 85% of the entire 9-million-procedure pool management itself described as being "in direct line of sight" at JPMorgan in January 2026.

Two things make this conservative rather than aggressive: revenue per procedure is held flat, when the disclosed trend is down (da Vinci I&A per procedure $1,860 → $1,830, with EUP and XiR both pushing it lower); and 5.0x is the middle of the peer range (SYK ~5x, BSX ~3.5x, EW ~8x) for a business with 70% gross and 40% operating margins.

Even after a 41% drawdown, the price requires near-complete capture of the company's own stated opportunity.

The cross-check that answers the reviewer directly

The three pipeline vectors the reviewer named, plus the fourth this pass found, are real, are being pursued, and are worth roughly 1.56 million incremental annual procedures sized generously. The price requires ~4.4 million. The pipeline therefore carries about ONE THIRD of the growth embedded in today's valuation. The other two thirds must still come from deeper penetration of core da Vinci indications in existing and new geographies.

The pipeline is the option. It is not the thesis.

That is the most useful sentence this exercise produced, and it is offered as a direct, quantified response to the reviewer's framing rather than an agreement with it.

EV/TAM — secondary sanity check only, with its denominator's elasticity stated

Elasticity, stated as the reference requires. Management moved this denominator from 6 million procedures (2023 investor presentation, quoted back to them by Michael Polark on the 2023Q3 call) to 7 million to 9 million (JPMorgan, January 2026, quoted back by Frederick Wise on the 2025Q4 call) — a +50% redefinition in three years, at its own discretion, with no external audit. In 2023 it simultaneously cited a 20 million total soft-tissue procedure pool; at that denominator EV/TAM would be 1.9x.

The ratio moves by more than 2x on a management slide. That is precisely why it is reported here as a secondary check and never as the headline, and it is the concrete illustration of the definitional-elasticity problem the reference file describes.

---

Part 6 — Provenance of hypotheses (required)

Score: of the seven substantive findings, three came from the generative process, three came from priors (one confirmed, two materially downgraded), and one prior finding was falsified outright. The generative pass produced the largest TAM in the memo and the only vector with a rising prepared-remarks series. That is the argument for running it first, before a view exists.

---

Part 7 — Disclosed limitations, each stating which corpora were attempted

Per the rule that a "not disclosed" finding is a research task, not a limitation to report: