MAIA Biotechnology makes a drug that tripled survival times in terminal lung cancer patients. Phase 2 data: 17.8-month median overall survival vs. 5.8 months on standard of care. In the hardest NSCLC population — third-line, failed both checkpoint inhibitors and chemotherapy. These patients have nothing left. Ateganosine gave 38% of them an objective response where chemo gives 6-10%. Eighty-five percent achieved disease control. A 30-month survivor is still going.

At $1.46 per share, $89M market cap. If Phase 3 confirms, the company is worth $500M to $2B — 5-20x. The EV is positive under every scenario model we ran.

We're not buying. Here's why that's the interesting part.

The Drug

Ateganosine exploits a vulnerability in 85% of cancers and less than 1% of normal cells. Telomerase — the enzyme cancer cells use to maintain their telomeres — incorporates ateganosine into telomeric DNA, causing structural collapse and rapid cell death within 24-72 hours. Normal cells, which barely express telomerase, are unaffected. The dose is 14x below the NCI-tested maximum tolerated dose.

The second mechanism is what makes this more than another cytotoxic: ateganosine converts immunologically "cold" tumors into "hot" ones. Give it before a checkpoint inhibitor to a patient who already failed checkpoint therapy. The tumor that was invisible to the immune system becomes visible. Published in Cancer Cell (2020), co-invented by MAIA's CSO Sergei Gryaznov and Jerry Shay at UT Southwestern. IP-protected to 2041 via exclusive UTSW licenses, with milestones only on commercial sales.

Three tier-1 pharma companies supply checkpoint inhibitors for free. Regeneron gave cemiplimab and demanded exclusive NSCLC PD-1 combination rights during the study period — they're positioning for ownership. BeOne (formerly BeiGene) gave tislelizumab for HCC/CRC/SCLC expansion. Roche gave atezolizumab for a master supply agreement. These companies run their own due diligence on the science. Their commitment is not nothing.

Geron's imetelstat — a different telomere-targeting agent that inhibits telomerase rather than weaponizing it — was approved in June 2024 for MDS and did $184M revenue in its first commercial year. Guiding $220-240M for 2026. But GERN trades at $1.57, down 10% YTD despite growing revenue. The market discounts telomere biology even when it works.

Twenty-Two Patients and the Ghost of Eftilagimod

The Phase 2 data is from 22 patients in the intent-to-treat population. Single-arm. No concurrent control. The 5.8-month SoC comparator is a historical benchmark, not a randomized control arm.

To be fair about what makes this data more reliable than typical Phase 2: overall survival is a hard endpoint — not a surrogate like ORR or PFS that can mislead. A 3x improvement is enormous and hard to produce through selection bias alone. The 95% CI lower bound (12.5 months) is still more than double SoC. Low toxicity means Phase 3 doesn't face the usual safety surprise risk.

To be honest about what makes it less reliable: n=22 is extremely small even for Phase 2. Single-arm trials tend to select healthier patients. The external SoC comparator may not match the Phase 3 control arm. And the base rates are brutal.

The BIO/QLS/Informa definitive study — 12,728 clinical transitions, 2011-2020 — puts oncology Phase 2 to Phase 3 success at 24.6%. Phase 3 to NDA: 47.5%, the lowest of all 14 disease areas. Combined likelihood of approval from Phase 2: roughly 10-12%. Citeline's more recent data (2014-2023) shows the trend is worsening, not improving. For drugs that received Accelerated Approval in oncology, 15% (30 of 198) have been withdrawn — and 87% of the withdrawals since 2016 came from drugs initially approved on response rate surrogates.

The ghost at this particular feast is eftilagimod alfa. Phase 2 in first-line NSCLC: 48.3% ORR, 11.2-month median PFS, 35.4-month median OS. By every measure, superior to MAIA's numbers. The Phase 3 trial, TACTI-004/KEYNOTE-F91, was halted at futility. The IDMC judged it unlikely to meet co-primary PFS and OS endpoints. Better data, same disease, dead.

It's not alone. Nivolumab received Accelerated Approval for SCLC in 2018 based on CheckMate-032. Two confirmatory Phase 3 trials — CheckMate-451 and CheckMate-331 — both failed primary OS endpoints. The indication was withdrawn. PACIFIC-2 (durvalumab in Stage III NSCLC) didn't meet its primary PFS endpoint despite clear theoretical rationale. Lung cancer chews up impressive Phase 2 data.

We put MAIA's conditional Phase 3 success probability at 15-25%. Call it 20%.

The Vehicle

The science is compelling. The company is concerning.

In October 2025, MAIA's board authorized holding up to 90% of liquid assets in Bitcoin, Ethereum, and USDC. Announced it publicly the next day. As of the 10-K filing: $0 in crypto. Strategy "on hold."

This is a 13-employee company that was five months from running out of cash, running an international Phase 3 oncology trial. The authorization came October 6, between a private placement on October 1 and another on October 13. We searched for comparable micro-cap biotech crypto treasury adoptions: Hoth ($1M BTC), Acurx ($1M BTC), Propanc ($100M via convertible preferreds). MAIA's 90%-of-liquid-assets authorization is the most extreme version we found in any clinical-stage biotech with an active Phase 3. Jefferies' David Windley called the pattern a "last-ditch effort to reignite investor interest." Kaiko's analyst called it a "death rattle." Post-announcement stock performance for these companies: uniformly negative after the initial pop.

Then there's FGMK. FGMK, LLC provides accounting, tax, and valuation services to MAIA — paid in stock. Its affiliate, FGMK Business Holdings, is a 5%+ stockholder that invested $4M+ in private placements and received a warrant repricing from $1.87 to $1.30 in September 2025. The firm valuing the company has financial interest in the company's survival and stock price. Grant Thornton audits independently, which mitigates but doesn't resolve the conflict.

Dilution: 26.2M shares (December 2024) to 38.6M (December 2025) to 60.7M (March 2026). That's 132% in 15 months. Fully diluted: ≈86.7M shares, another 43% overhang. Burn accelerating from $15.7M to $18.8M to an estimated $25-30M as Phase 3 and multi-indication expansion ramp. Of the G&A increase, $1.5M went specifically to investor relations. Another raise before end of 2027 is 85% probable.

The Competitive Landscape MAIA Doesn't Mention

MAIA's 10-K positions the 3L CPI-resistant NSCLC space as essentially empty. We searched 5,737 earnings call transcripts for "third-line NSCLC" and "CPI-resistant." Zero mentions. No large-cap oncology company is focused here. That part is true — this is a micro-cap frontier. It cuts both ways: real unmet need, but also a population so small that big pharma hasn't bothered.

What the 10-K doesn't say is that MAIA isn't alone on that frontier.

Candel's CAN-2409 — a viral immunogene therapy — showed 24.5-month median OS in CPI-resistant NSCLC. Numerically superior to MAIA. Different mechanism, similarly small n, similarly unreliable cross-trial comparisons. But it's initiating Phase 3 on essentially the same timeline. ImmunityBio's ANKTIVA is already approved for bladder cancer, giving it commercial infrastructure MAIA doesn't have. ADCs — particularly Dato-DXd, now NCCN-preferred in 3L — are encroaching from the EGFR-mutated side.

MAIA's differentiation is real: sequential ateganosine-then-CPI in wildtype (non-EGFR) CPI-resistant NSCLC. Nobody else converts cold tumors to hot. But "first mover in 3L NSCLC" is not the same as "only mover."

Where the Edge Isn't

Here is the finding that changed our conclusion.

MAIA trades at beta -0.02, idiosyncratic volatility 95.8%. From a portfolio construction perspective, this looks ideal — pure stock-specific risk, way above the 75% idio variance target. But the 95.8% is misleading. The stock is entirely idiosyncratic not because we've isolated our edge, but because a single binary event dominates everything.

We decomposed the thesis into six independent factors:

The positive EV comes entirely from the clinical factor. A 20% probability of 585% upside overwhelms a 35% probability of 76% loss. The math works.

We have zero informational advantage in the clinical factor. Our 20% is a base rate, not an informed view. We don't have patient-level data, enrollment signals, biomarker insights, or FDA feedback beyond the 10-K.

Where we have edge — governance, competition, dilution — everything is bearish. The crypto timing between placements. The FGMK conflict. Candel's superior Phase 2 numbers on the same timeline. The dilution trajectory to 100M+ shares before Phase 3 reads out.

The marginal retail buyer of MAIA probably doesn't see any of this. Our edge is real. It argues against the trade.

Per Paleologo: size proportional to alpha from factors where you have informational advantage. Our advantage is concentrated in 15% of outcome variance. All of it is bearish. Edge-adjusted alpha is negative.

What the Market Prices

At $1.46, the market implies 6-11% success probability:

Our base rate estimate: 20%. That gap — 2-3x — looks like mispricing. But micro-cap governance discounts are often overdone. Institutions refuse to own stocks with crypto treasury policies and FGMK-type conflicts, creating selling pressure with no matching bid. The market's 6-11% may already be the governance penalty, not clinical pessimism.

We can't distinguish "market is too pessimistic on the drug" from "market correctly discounts for governance." Without clinical edge, this is unresolvable.

All three positive. The most conservative still +13%. But the entire upside rests on a factor where we're no better than random.

The Trade

At $1.46: pass. Thin margin of safety for zero clinical edge. Stock is within pennies of the March offering price. Twenty-two million new shares still being absorbed.

At $1.00: tracking position (0.3-0.5%). Insider buying zone — CEO Vitoc and five directors bought here in November-December 2025, $300K+ of personal capital. Even market-implied EV gives 65% upside. Governance discount extreme enough to plausibly overshoot.

At $0.87 (52-week low): small position (0.5-1%). Clinical option is essentially free. Market-implied EV gives 90% upside.

The timing: the stock may drift lower before binary catalysts arrive (AA filing 55% in 2026, Phase 3 interim 65% in 2026). The next capital raise (85% by end 2027) is the best entry window. Buy the dilution-driven weakness, position before the AA filing catalyst.

What Changes This

Bullish: Enrollment velocity showing THIO-104 >50% enrolled by Q3 2026. AA filing accepted. Governance cleanup — rescind crypto policy, replace FGMK. Institutional buyer appears in 13F.

Bearish: Phase 3 interim DCR below 60%. Capital raise at discount before any milestone. Crypto policy reactivated. Candel reports positive Phase 3 data first.

MAIA has exceptional science in a problematic vehicle. The Phase 2 data is the kind of result that makes oncologists stop and stare — 3x survival in patients who have nothing left. But twenty-two patients is twenty-two patients. The governance tells you something about who's running the company. And everywhere we actually have informational advantage, the signal is bearish.

We'll watch for the price to come to us. If it does, we're buying the clinical option at a discount deep enough that even our bearish edge factors can't eat all the upside. Until then, this is a name we know well and a position we don't have.

Scenarios (18 months)

Evidence