KROS — Keros Therapeutics, Inc. · 2026-09-07 · Verdict: WATCH · Conviction 2
Price $11.32 (screen row in universe_under2b.csv, rebuilt 2026-09-07; no live quotes in this sandbox) · Mkt cap $224.4M (19,827,188 shares) · EV -$7.9M after all liabilities (screen says -$33.2M because it nets only cash) · EV/EBIT n/m, no product revenue · FCF yield n/m (screen's 47% is the one-time CY2025 Takeda upfront) · Net cash $232.3M, $11.72/share · ADV $2.0M
Sources read: 10-K filed 2026-03-04 (Items 1, 1A, 7), 10-Q filed 2026-08-03 (MD&A), DEF 14A 2026-04-22, 8-Ks 2026-06-03, 2026-06-25, 2026-07-22, 2026-07-30 and 2026-08-03 (Items 2.02 and 7.01, including the corporate presentation), Form 4s trailing 12 months. No earnings call transcript or prepared remarks were filed; the "transcript" in the bundle is the Q2 press release.
1. What the business actually does
Keros is a clinical-stage biopharmaceutical company that engineers ligand traps against the TGF-beta family of proteins (10-K 2025, Item 1). Two molecules matter. Elritercept (KER-050) treats cytopenias in myelodysplastic syndromes and myelofibrosis and is licensed worldwide outside greater China to Takeda, which since January 2025 has been responsible for all development, manufacturing and commercialization in its territory (10-Q Q2 2026, MD&A). Rinvatercept (KER-065) is an activin and myostatin trap now in an open-label, multi-cohort Phase 2 basket trial of about 24 late-ambulatory and early non-ambulatory Duchenne muscular dystrophy patients, dosed every four weeks for 48 weeks, with initial data expected in the first half of 2027 (corporate presentation, 8-K 2026-08-03, EX-99.1). An ALS Phase 2 is at the regulator-engagement stage (same source). A third asset, cibotercept, had its Phase 2 terminated (10-K 2025, Item 7). There is no product revenue and none is expected for years.
2. Why it is mispriced, and why the discount may not close
The screen edge is real and arithmetic. At $11.32 the equity is $224.4M against $257.6M of cash and $25.3M of total liabilities, of which $15.7M is operating leases (10-Q Q2 2026 balance sheet). Net cash after every liability is $232.3M, or $11.72 a share, and the $20M Takeda milestone triggered by first patient dosing in the ELRiSE MDS Phase 3 was announced after the quarter closed (8-K 2026-07-30, Item 7.01), lifting pro-forma net cash to about $12.72 a share. So the buyer pays roughly nothing, or less than nothing, for rinvatercept, for up to $1.08B of remaining Takeda milestones, for low-double-digit to high-teens royalties on an asset Takeda is funding through two Phase 3 trials, for the Hansoh China deal worth up to $167.5M in milestones plus royalties, and for the preclinical library (10-Q Q2 2026, MD&A).
The seller is not irrational, and this is where the case weakens. Through 2025 two activists, the ADAR1 and Pontifax parties, pushed Keros toward returning capital. On October 15, 2025 the company bought their entire 10,176,595 shares at $17.75 for $180.6M, two Pontifax directors resigned, and both parties signed standstill and voting commitments running until certification of the 2028 annual meeting vote; a tender at the same $17.75 took another 10,950,165 shares for $194.4M (10-K 2025, Item 7). Total 2025 buybacks were $375.0M, and treasury stock stands at 21,126,760 shares costing $384.6M, an average of $18.20. The party that would force the cash out has been paid off and silenced for exactly the period over which the cash will be spent. That is the honest reason a negative enterprise value persists here.
3. Unit economics
There are none in the usual sense. Q2 2026 revenue was zero against $22.3M of R&D and $8.6M of G&A, for a $28.7M net loss (8-K 2026-08-03, EX-99.1). H1 operating cash use was $24.0M, but that included a $14.8M working capital release, mostly a $15.2M drawdown of prepaid expenses that cannot repeat, so underlying burn was about $38.8M for the half, near $19M a quarter (10-Q Q2 2026, MD&A). Q2 opex of $30.9M less roughly $5.8M of stock compensation and $2.3M of dividend income puts the current run rate closer to $23M. Management's own runway language implies more: $257.6M funding operations "into the first half of 2028" is six to eight quarters, or about $35M a quarter. Note the endpoint did not move when cash fell $29.8M over six months, since the 10-K gave the same first-half-2028 date from $287.4M (10-K 2025, Item 7; 10-Q Q2 2026, MD&A). Either the guidance carries slack or the DMD and ALS ramp is about to double the burn. Against a $224M market cap, $19M to $35M a quarter is 8% to 16% of the equity per quarter.
4. Balance sheet and capital allocation
No debt. The ATM is dormant: no shares were sold in 2025 and, having lost well-known-seasoned-issuer status, Keros stated it "may not offer and sell any ATM shares" as of December 31, 2025 (10-K 2025, Item 7). In October 2025 the board committed to distribute 25% of any net Takeda cash proceeds received on or before December 31, 2028 (10-K 2025, Item 7), and reaffirmed it for the $20M milestone (10-Q Q2 2026, MD&A). That leaves 75% to be spent. Insider alignment is thin: CEO Jasbir Seehra holds 312,498 actual shares against 1,559,184 near-term options, and all officers and directors are 12.6% mostly through options (DEF 14A 2026-04-22). BlackRock owns 10.1%. Trailing-12-month Form 4s show net selling of $131,920; the only buyer is director Jean-Jacques Bienaimé, three purchases of 1,000 shares each at $10.19, $10.61 and $10.27 (Form 4s 2026-06-17, 2026-07-17, 2026-08-19), which is a gesture, not a signal.
5. Management: said versus did
They said they would return capital and did, at $17.75, to a stock now at $11.32. They said the DMD Phase 2 would start in Q2 2026 (10-K 2025, Item 1) and it did (8-K 2026-08-03, EX-99.1). Governance is unsettled: CFO Keith Regnante resigned effective August 3, 2026 with the corporate controller as interim principal financial officer and a search underway (8-K 2026-07-22, Item 5.02), a new director was added July 1 (8-K 2026-06-25), and at the June 3 annual meeting Bienaimé drew 5,156,153 withheld votes against 10,093,559 for, roughly a third of votes cast (8-K 2026-06-03, Item 5.07).
6. Valuation
Bear (40%, $5.50): burn runs at management's implied $35M a quarter, the open-label 24-patient DMD readout is uninterpretable, ALS starts anyway, and by late 2028 cash is near $100M with the standstill only then lapsing. Base (40%, $9.50): burn holds near $20M a quarter, the stock keeps tracking a shrinking net cash per share, no rerating. Bull (20%, $24.00): the H1 2027 DMD data show functional benefit consistent with the Phase 1 muscle, bone and fat pharmacology (corporate presentation, 8-K 2026-08-03, EX-99.1), and elritercept advances toward the $280M of commercial milestones. Probability-weighted value is about $10.80, within noise of the $11.32 price. Reverse DCF: the price implies the market values the entire pipeline, the Takeda royalty and milestone stream, and the Hansoh rights at roughly minus $8M, which is a precise statement that every dollar Keros spends from here is expected to destroy slightly more than a dollar. That is a defensible view, not an obvious error.
7. Catalysts and timeline
Receipt of the $20M milestone and payment of the 25% distribution, H2 2026. Regulator engagement on the ALS Phase 2 design, H2 2026. Q3 and Q4 2026 cash balances, which will settle whether burn is $19M or $35M a quarter. Initial DMD Phase 2 data, H1 2027. Standstill expiry after the 2028 annual meeting.
8. Kill criteria (pre-registered)
- Cash and equivalents below $200M at any quarter end on or before June 30, 2027, which would confirm burn at the higher implied rate.
- Any equity issuance, restored ATM, or PIPE priced at or below the then-current net cash per share.
- Initial DMD Phase 2 data in H1 2027 showing no improvement in the functional secondary endpoints, or any dose-limiting toxicity or serious adverse event.
- Public withdrawal of the 25% Takeda distribution policy, or failure to pay the distribution on the $20M milestone by the FY2026 10-K.
9. Verdict and summary
WATCH, conviction 2. Keros trades at $11.32 against $11.72 a share of net cash after every liability, $12.72 once the $20M Takeda milestone lands, so the market is paying less than nothing for a DMD trap in Phase 2, a Takeda-funded Phase 3 asset carrying up to $1.08B of milestones and low-double-digit to high-teens royalties, and a China deal on top. The reason that discount is not free money is on the record: in October 2025 the company spent $375M buying out the two activists who wanted the cash returned, at $17.75 a share, and bound them to a standstill until the 2028 annual meeting, which is precisely the window over which the remaining $257.6M gets spent at $19M to $35M a quarter. Nothing now forces the cash back to shareholders except a voluntary 25% pass-through of Takeda proceeds. The stock is therefore worth watching for two specific things and buying for neither yet: the Q3 and Q4 2026 cash balances, which decide whether the burn is the $19M a quarter the cash flow statement shows or the $35M management's runway implies, and the H1 2027 DMD readout, which is the only event that can make the pipeline worth more than the cash it consumes.
Research for discussion, not investment advice. Positions and sizing are the reader's decision.