RARE — Ultragenyx Pharmaceutical Inc.

Company context · prepared 2026-08-07 · company sweep 2026-08-07 · USD · framework v5.5.0

How to read this. Written for a reader who has not studied pharmacology or finance. General terms are defined once in framework/04-glossary.md. Tags: [VERIFIED — source] / [UNVERIFIED] / [WEB ESTIMATE — source, date].

Program analyses for this company are the sub-folders listed in C.2.

C.0 Company-tier coverage — CLEARED

No mandatory company-tier row reads NOT CALLED, so the company tier clears.

What changed at the gate since the v3.7.0 sweep. The one substantive company-tier gap in that sweep — that the second-quarter results were unreachable and had to be read from a third-party rendering — is closed. The press feed now carries the release, and the GlobeNewswire text itself was read through a newswire syndication [VERIFIED — GlobeNewswire, “Ultragenyx Reports Second Quarter 2026 Financial Results and Corporate Update”, 2026-08-04, read via manilatimes.net/tmt-newswire/globenewswire 2026-08-07]. Every second-quarter figure below now rests on the newswire text rather than on a summariser’s rendering of it. The primary ir.ultragenyx.com page still could not be opened: the request timed out after 60 seconds, which is the same failure the previous sweep recorded. That is the remaining limit of the check.

C.1 What the company is

Ultragenyx Pharmaceutical develops and sells drugs for rare and ultra-rare inherited diseases. It is not a single-asset company and it is not pre-revenue. Four products are already approved and selling — Crysvita (burosumab) for X-linked hypophosphataemia and tumour-induced osteomalacia, Dojolvi (triheptanoin) for long-chain fatty-acid oxidation disorders, Evkeeza (evinacumab) for homozygous familial hypercholesterolaemia, and Mepsevii (vestronidase alfa) for mucopolysaccharidosis VII — and together they produced $214 million of revenue in the second quarter of 2026, against a FactSet consensus of $183.1 million [VERIFIED — GlobeNewswire release 2026-08-04; consensus from MarketScreener earnings flash 2026-08-04, in the BPIQ press feed]. Guidance for the full year 2026 is $730–760 million, “which excludes revenue from potential new product launches” [VERIFIED — same release].

The development pipeline is nine BPIQ rows across eight distinct programs (C.2). Four of the nine rows carry a catalyst: two are United States regulatory decisions inside the next seven weeks, one is the Phase 3 Angelman readout analysed in this folder, and one is an early-stage Wilson disease readout.

The company is still loss-making. It lost $92 million, or $(0.90) a share, in the second quarter of 2026 and states that it remains “on path to profitability in 2027” [VERIFIED — GlobeNewswire release 2026-08-04].

C.2 Pipeline — every program

Every row returned by fetch_company_drugs on 2026-08-07 appears below, including the failed one. Catalyst dates are reported as catalyst_date_text; the “to a day?” column is yes only where that text names a day.

The four marketed products do not appear in this feed. fetch_company_drugs returned only development-stage rows plus the failed setrusumab row. Crysvita, Dojolvi, Evkeeza and Mepsevii — which are the entire revenue base — have no row. Anyone reading the pipeline table alone would conclude Ultragenyx has no product. It has four.

Why the whole table matters for timing, not only the analysed rows. The clustering pass (framework/02-connectors.md § Catalyst clustering) compares every has_catalyst: true row above against every other, analysed or not. On this ticker that is what finds the real collision: DTX401’s 2026-08-23 and UX111’s 2026-09-19 are both exact dates, and both sit inside the GTX-102 readout window. Neither has a program document of its own, so a detector that only opened program.json files would see none of it.

C.3 Financial position

On the runway-versus-catalyst tension. It re-verifies as real but not acute, and it points in opposite directions for the two analysed programs. At $32.3M a month the money lasts to roughly August 2027. The GTX-102 readout window opening 2026-09-01 sits comfortably inside that. The DTX301 readout guided to the first half of 2027 sits close to the edge of it. Three things soften the picture and one hardens it. It softens because the burn rate has halved quarter on quarter, from $197M to $97M, following the restructuring announced with the setrusumab failure; because two saleable priority review vouchers worth roughly $200M in aggregate become available if the August and September approvals land; and because revenue is growing and beat consensus by 17% in the second quarter. It hardens because the first-quarter figure is not a one-off distortion the company can repeat — management said explicitly that the first quarter is always the heaviest — so the true underlying rate sits between the two quarters rather than at the lower one.

Dilution history. No equity offering, at-the-market facility or withdrawn offering appears in the 406-item press feed for the 24 months to 2026-08-05. The feed carries an item dated 2026-06-16 referring to an “ESOP share shelf”, which is an employee-plan registration and not a capital raise, and repeated inducement-grant notices under Nasdaq listing rule 5635(c)(4) — most recently 44,409 restricted stock units to 15 new hires on 2026-06-19 and a further grant on 2026-07-25 — which are new-hire equity awards.

One plain line on whether a raise is likely before the next catalyst: unlikely before the two 2026 approval decisions and the Angelman readout, because the money reaches into the second half of 2027 without one and because selling equity into a binary readout is the worst available price. It becomes likely in 2027 if the Angelman readout misses and the voucher sales disappoint.

C.4 Valuation frame

Recomputing the market capitalisation. BPIQ reports $2,455,336,448, which at $25.82 implies 95,094,000 shares. The second-quarter net loss of $92 million at $(0.90) a share implies about 102,220,000 weighted-average shares, and a Schedule 13G filed 2026-05-12 reports State Street holding 5,021,441 shares as 5.1% of the company, which implies about 98,460,000. BPIQ is still carrying a share count roughly 7% too low. On 102.2 million shares the market capitalisation is about $2,639M.

Recomputing the enterprise value. BPIQ’s $1,809.4M is $645.9M below its own reported market capitalisation, yet its own cash field reads $413.0M. The figure remains internally inconsistent with the rest of the record and is not used. Recomputed: $2,639M of market capitalisation less $436M of cash at 2026-06-30 is $2,203M, and that is a floor, because the royalty-monetisation liabilities are real and their carrying value could not be confirmed.

The max_52_week_position bug is invisible today, and that is a coincidence, not a fix. BPIQ reports 0.30741. The documented defect is that the field uses the previous close; today the previous close is $24.93, which is also the committed price cache’s own most recent close, so both methods land on the same 30.7%. On the newer $25.82 close the correct figure is 34.9%. Since v5.1.0 this row is derived from data/prices/RARE.json with lib/prices.mjs rather than read from BPIQ at all, so the coincidence changes nothing about the method — it is recorded because a reader comparing the two numbers would otherwise conclude the bug had been fixed.

The shape of the year matters more than the level. The 52-week high of $39.89 belongs to the period before 2025-12-29, when setrusumab was still a live Phase 3 asset; the low of $18.29 was printed in the days after that failure. The stock then recovered to $32.38 by 2026-07-01 and has given back 23% of that in the five weeks since [VERIFIED — data/prices/RARE.json, 2026-07-01 close $32.38 against the 2026-08-05 close $24.93]. That drawdown is documented in the press feed as a nine-day losing streak, the longest since December 2024, and it coincides with the Russell index migration recorded in C.5 rather than with any company-specific news. This is the run-up baseline for every program under this ticker.

C.5 Ownership and flow

Insider flow. The 250-row file covers 2023-03-01 to 2026-07-01 and is byte-for-byte the same window as the previous sweep. Across that whole window there is not one open-market purchase by any officer or director. Every acquisition at a non-zero price is an employee stock purchase or an option exercise: five rows in total, three of them Theodore Huizenga at a flat $21.00 (2023-05-09, 2023-09-07, 2024-01-02) and two of them chief executive Emil Kakkis (47,853 shares) and chief business officer Thomas Kassberg (39,878 shares) at a flat $6.86 in October 2023, both patterns characteristic of plan purchases rather than conviction buying. Selling is continuous and mechanical: chief financial officer Howard Horn sold on the first trading day of January, February, April, May, June and July 2026, in sizes between 3,061 and 5,290 shares, which is the signature of a pre-arranged plan.

Two rows are worth naming. Horn’s 2026-07-01 sale of 4,698 shares at $33.13 is the highest price in the entire file, and the shares are 22% lower five weeks later. That is a pre-arranged sale that happened to land on the high, not evidence of foreknowledge, but it does establish that the stock traded at $33 as recently as five weeks before this analysis. And the connector file is now demonstrably behind the press feed: two Form 4 items dated 2026-08-04 — the chief financial officer selling under a 10b5-1 plan and the chief legal officer selling 1,900 shares at $24.92 — appear in the press feed and in no row of the transactions file. Both are consistent with the same mechanical pattern; neither is a new signal.

Concentration and fund positioning. Six hedge funds are on file for 2026Q2, holding 7.95 million shares in total, or about 7.8% of the company: RTW Investments 3,648,795 shares, Sofinnova Investments 3,145,888, Palo Alto Investors 596,495, Ikarian Capital 482,571, DAFNA Capital 61,100 and ADAR1 Capital 13,400. The quarter is internally consistent — every one of the six implies $20.95 a share, which is the 2026-06-30 close — so the documented 1000× value bug does not appear on this ticker. Two changes stand out. Baker Bros. Advisors held 2,772,692 shares for six consecutive quarters through 2026Q1 and is absent from the 2026Q2 file, which reads as a full exit ahead of the Angelman readout. Sofinnova went from 83,300 shares to 3,145,888 in the same quarter, a 38-fold increase, and RTW added 649,000. Deep Track Capital, Frazier Life Sciences, Avoro, BioImpact and Perceptive were all on file in 2025 and are all gone by 2026Q2. The specialist money did not leave; it changed hands.

Index flow, which is not conviction. Ultragenyx was dropped from the Russell 1000 and the Russell 1000 Growth and Dynamic indices and added to the Russell 2000, Russell 2000 Value and Russell Small Cap Comp Value indices in the last week of June 2026 [VERIFIED — BPIQ fetch_company_press_releases, marketscreener items 2026-06-28 to 2026-06-30]. A large-cap-to-small-cap migration forces mechanical selling by index funds and is a plausible non-fundamental contributor to the July drawdown.

Litigation flow. The press feed carries law-firm investigation notices (Johnson Fistel 2026-07-15, Kuehn Law 2026-06-05) referring to alleged false statements on drug trial results, and a 2026-05-16 item recording that the Delaware Chancery Court tossed a stockholder complaint against the company. Investigation notices are solicitations, not filed claims, and are recorded here as flow rather than as fact.

C.6 Options chain and its readability

Plain-English takeaway. The chain is listed, deep and now carries real volume as well as real open interest — thousands of contracts changed hands at single strikes around the second-quarter release, where the previous sweep found tens. But the quotes have got wider, not tighter, so a single number for the expected move is less defensible than it was a week ago, not more. A bracket is what the chain supports. For the expiry that spans the Angelman readout the market is pricing something between a 41% and a 70% move in either direction by mid-October. Even the narrow end of that is enormous, and it is the clearest single statement anyone makes about how binary the next ten weeks are.

Artifact check. The documented implied-volatility floor of 0.01488 appears on the 2026-08-21 calls at $2.50, $5.00, $7.50, $10.00, $12.50, $15.00 and $17.50, on the 2026-11-20 $5.00 and $7.50 calls, on the 2027-01-15 $2.50 and $5.00 calls, and on the 2028-01-21 $2.50 call — all of which also carry a delta of exactly 1 and a gamma and vega of exactly 0. Those contracts are unusable. The documented ceiling of 9.99512 does not appear anywhere in the chain. Every figure quoted above avoids both.

C.7 Reaction to past catalysts

Intraday moves are intra_day_price_change_percent from fetch_company_historical_catalysts; gaps are open_price_gap_percent. Every date was cross-checked against the press feed for the same day. No new row has been added since the previous sweep; the table below is re-verified rather than extended.

What this table says as a whole. Ultragenyx has an asymmetric reaction function. Approvals and positive Phase 3 data on secondary assets move it 0–9%. Phase 3 misses on the lead asset move it −13% to −45%. There is no row in the file where good clinical news produced a move of more than about 9%. Any scenario that assumes a positive Angelman readout is worth 50% has to explain why that would be the first time.

C.8 Company data-quality flags

• max_52_week_position reproduces the documented previous-close bug, but invisibly today. 0.30741 is what the formula returns on the previous close of $24.93; on the last price of $25.82 the correct figure is 34.9%. It happens to match the committed price cache’s own last close, so a reader comparing the two would wrongly conclude the bug had been fixed. Since v5.1.0 this row is derived from the cache and the BPIQ field is not read.
• BPIQ’s enterprise value is internally inconsistent with its own cash field. The reported $1,809.4M is $645.9M below the reported market capitalisation, but the reported cash is $413.0M. The two cannot both be right and no debt figure reconciles them. The reported figure is not used.
• BPIQ carries a stale share count. Its market capitalisation implies 95.1 million shares against roughly 102.2 million implied by the second-quarter loss per share and roughly 98.5 million implied by State Street’s 2026-05-12 Schedule 13G.
• BPIQ’s cash field is a narrower line than the company’s. $413M against the company’s $534M at the same date; the difference is marketable securities. Comparing the BPIQ figure to a company-reported figure from another quarter would produce a fictitious cash movement.
• BPIQ’s burn fields are one quarter stale. qtr_burn −$197M and monthly_burn −$65.7M are the first-quarter figures. The second quarter used $97M, half as much. Quoting the BPIQ field would overstate burn by 2×.
• tute_own reads 1.04713 and is rejected. As a fraction that is 104.7% of shares outstanding. The field has drifted further from plausibility since the previous sweep (1.02165) rather than being corrected. It is not passed through in any form.
• short_float units are NOT resolved for this ticker. The field reads 0.14201 and no second source was found.
• The drug feed omits every marketed product. fetch_company_drugs returns nine rows and not one of them is Crysvita, Dojolvi, Evkeeza or Mepsevii, which are the entire revenue base.
• Rows 20225 and 20222 are a duplicate. Both are “UX016 (Substrate Replacement)” / “UX016 (Substrate Replacement Therapy)” in GNE myopathy, Phase 1/2 Initiation, H2 2026, with note fields that paraphrase each other. Counting rows overstates the pipeline by one program. Same class of defect as the drug-name dirtiness documented in 02-connectors.md.
• open_price_gap_percent and intra_day_price_change_percent disagree in sign on the two most important rows in C.7 (2025-12-29: −43.49% against +2.07%; 2018-10-26: −13.85% against +10.66%). The connectors file prefers the intraday figure. On these two rows that preference would invert the finding, so both are reported alongside an independent close-to-close check.
• fetch_company_earnings_transcript is one call behind. It still returns the 2026-05-05 first-quarter call three days after the second-quarter call took place, and the second-quarter transcript is reachable only through a third-party rendering in the press feed (MarketBeat, 2026-08-05). No claim in this document is sourced to that rendering.
• fetch_company_insider_transactions is behind the press feed. Two Form 4 filings dated 2026-08-04 appear in the press feed and in no row of the 250-row transactions file, which still ends 2026-07-01.
• The primary investor-relations site remains unreachable. ir.ultragenyx.com timed out after 60 seconds on this sweep as it did on the last. The second-quarter release was read from a GlobeNewswire syndication instead, which is the newswire text rather than a summary — an improvement on the previous sweep’s third-party rendering, but still not the primary page or the Form 10-Q.
• One press-feed item is mis-dated. An item titled “Ultragenyx Announces Partnership with GeneTx to Advance Treatment for Angelman Syndrome” carries the date 2026-06-20 and points at an Ultragenyx investor-relations URL. Ultragenyx acquired GeneTx outright in July 2022, so a new “partnership” in June 2026 is not credible; this is almost certainly the 2019 partnership release carrying a wrong date in the feed. It is not used as evidence of anything.
• The committed price cache lags BPIQ by one trading day. data/prices/RARE.json ends 2026-08-05 ($24.93); BPIQ reports the 2026-08-06 close of $25.82. Both are recorded above with their own dates. No price fetch was run on this analysis, per its own instructions.