Apex Capital Partners · CIO Intelligence

AI Medicine: The Discovery Spectrum

Tempus AI (TEM) Phase Classification & Comparative Alpha Analysis Against True Phase 3 Discovery Names
Report DateFebruary 18, 2026
ClassificationSector Deep Dive
Framework4-Phase Cyclical AI Value Capture
MethodologyAlpha Score v2.0 (Forward-Looking)

I. Executive Summary

The core question: Where does Tempus AI sit in the AI value chain — and does it belong in Phase 3?

Tempus AI ($57.39, +9.5% today on Pan-Cancer HRD-RNA algorithm launch) is the most financially successful AI-medicine company by revenue. At ~$1.27B FY2025 revenue growing 83% YoY, it dwarfs every company in our Phase 3 Discovery portfolio combined. But revenue scale ≠ Phase 3 classification within Apex's cyclical framework.

This report establishes why TEM is a Phase 2 Deployment name, provides full v2 Alpha Scores for TEM against our Phase 3 Discovery names (SDGR, RXRX, DNA, RLAY), and identifies whether the broader AI medicine narrative creates actionable opportunities for Apex Capital.

Key finding: RLAY has been fundamentally repriced since our original assessment — FDA Breakthrough Therapy designation for zovegalisib (Feb 3) and $596M cash runway into 2029 transforms it from a "cash runway risk / EXIT" call to the most de-risked AI-discovery clinical asset in our universe. This report includes a revised RLAY alpha score.

II. The AI Medicine Spectrum

Not all "AI + healthcare" companies create the same type of value — or the same type of compute demand.

The Spectrum: From Deployment to Discovery

Phase 2 · AI Applied Hybrid Phase 3 · AI Native Discovery

Phase 2: AI-Applied Medicine

What it does: Uses AI to analyze existing data (genomics, pathology, clinical records) to match patients with existing treatments more precisely.

Compute profile: Classification algorithms on structured data. Computationally modest — a 1,660-gene logistic regression model runs in minutes.

Revenue mechanism: Diagnostic testing fees + data licensing to pharma. Recurring, scalable, and already proven at $1.27B/yr (TEM).

Capex feedback: Minimal. Does not generate new demand for GPU clusters. Pharma pays for data access, not compute infrastructure.

TEMPUS AI sits here.

Phase 3: AI-Native Discovery

What it does: Uses AI to simulate molecular physics, fold proteins, and design molecules that don't yet exist. Creates new treatments, not just better use of existing ones.

Compute profile: Molecular dynamics across 1012 timesteps. Protein conformational landscapes with 10300 states. GPU-hours per candidate in the thousands. Exact or worthless — you cannot distill physics.

Revenue mechanism: Software licensing + milestone payments + clinical-stage therapeutics. Pre-revenue at scale, but monetization creates step-function compute demand.

Capex feedback: Massive. Clinical validation → every pharma needs dedicated GPU clusters → Phase 1 infrastructure demand restarts at higher intensity.

SDGR · RXRX · DNA sit here.

The RLAY Question: Relay Therapeutics occupies a hybrid position. Its Dynamo® platform uses molecular dynamics simulation (Phase 3 methodology) but applies it to a focused clinical pipeline (more traditional biotech execution). With zovegalisib now carrying FDA BTD and a Phase 3 trial underway, RLAY is the closest to proving that AI-guided drug design produces clinically superior outcomes. If ReDiscover-2 reads out positive in mid-2026, it becomes the single most important proof-of-concept for the entire AI-discovery sector.

III. Tempus AI — The Phase 2 Case

$9.92B market cap · $57.39 · Earnings Feb 24 · 83% revenue growth · Not yet profitable

Business Architecture

  • Diagnostics (~$955M, 75% of rev): NGS oncology + hereditary testing via acquired Ambry Genetics. Volume-driven, ~111% YoY growth (acquisition-boosted). Organic growth ~30%.
  • Data & Applications (~$316M, 25% of rev): Insights data licensing to 70+ pharma customers, growing 38% YoY. This is the crown jewel — recurring, high-margin, and moated by dataset scale.
  • Algorithmic Tests: HRD-RNA (launched today), PurIST, IPS, Paige Predict. Classification models that improve treatment selection — powerful but computationally light.
  • Tempus One: Generative AI clinical co-pilot integrated into EHR systems. Phase 2 deployment pure-play.

Why Phase 2, Not Phase 3

  • Data moat, not compute moat. TEM's competitive advantage is the world's largest linked clinical-genomic-imaging dataset (200K+ patients for Paige Predict). The defensibility is data accumulation, not computational methodology.
  • Pattern matching, not physics simulation. HRD-RNA is a 1,660-gene logistic regression. Impressive classification. But not molecular dynamics, protein folding, or de novo molecule design.
  • Optimizes existing treatments. TEM tells you which existing drug works best for which patient. Phase 3 companies design drugs that don't yet exist.
  • No capex feedback loop. TEM's success doesn't cause pharma to buy more GPUs for simulation. It causes them to buy data licenses — very different capital flow.

TEM Financial Snapshot

MetricFY2024FY2025 (prelim)Growth
Total Revenue~$693M~$1,270M+83%
Diagnostics Revenue~$452M~$955M+111%
Data & Applications~$241M~$316M+31%
Insights (Data Licensing)~$316M+38%
Adj. EBITDA~($105M)~$5M (est.)+$110M
Net Revenue Retention~126%
Cash & Securities~$764M
Market Cap$9.92B
P/S (TTM)~7.8x

Today's Catalyst: Pan-Cancer HRD-RNA Algorithm

The HRD-RNA algorithm uses RNA gene expression (1,660 genes) to dynamically assess Homologous Recombination Deficiency — a key predictor of response to platinum-based chemotherapy and PARP inhibitors. Unlike static DNA-based "genomic scar" tests, this RNA approach captures real-time tumor functional status, detecting HRD even in cancer types where genomic scarring is difficult to identify.

Significance: Incrementally positive — adds to TEM's algorithmic test suite (PurIST, IPS, Paige Predict), demonstrates continued product innovation ahead of Feb 24 earnings. Currently research-use only; clinical availability expected later in 2026. The +9.5% move is pre-earnings positioning on an oversold name (down 45% from 52-week high), not a thesis inflection.

IV. Alpha Score v2 — Comparative Analysis

Forward-looking 6-month scores using Apex v2 methodology. Scoring what conditions will be, not what they are today.
Schrödinger
SDGR · $11.87 · NASDAQ · Phase 3 Discovery
63
v2 Composite
Moat Durability
72
Catalyst Runway
68
S/D Trajectory
58
Edge Decay
55
Regime Alignment
60

Moat (72): Physics-based FEP+ platform is structurally differentiated — competitors (Isomorphic Labs, etc.) are ML-only. All top 20 pharma as software customers. Novartis $2.5B deal. Lilly TuneLab integration. Software licensing is recurring floor under the stock. Neptune ammonia catalyst validates platform beyond pharma.
Catalyst Runway (68): Feb 25 earnings, SGR-1505 Phase 1 data completion 2026, SGR-3515 data H1 2026, Copernic catalyst pilot demos, potential new pharma collaborations. Mix of confirmation and new information events.
S/D (58): Software demand stable but drug discovery revenue lumpy. Competition from Isomorphic Labs and large pharma in-house teams could compress licensing pricing over time.
Edge (55): "Physics + AI" platform thesis is increasingly known. Stock down 58% from 52-week high — bad sentiment, but the thesis hasn't been disproven. Remaining edge is in clinical data readouts that haven't yet arrived.
Regime (60): FDA end-of-animal-testing policy (Apr 2025) is structural tailwind. But biotech funding environment remains tight. Rate sensitivity moderate.

Recursion Pharmaceuticals
RXRX · $3.41 · NASDAQ · Phase 3 Discovery
57
v2 Composite
Moat Durability
58
Catalyst Runway
65
S/D Trajectory
52
Edge Decay
55
Regime Alignment
55

Moat (58): 65PB multimodal imaging dataset + OS 2.0 + Exscientia merger is differentiated. NVIDIA partnership. But platform validation remains clinical — until pipeline drugs work in humans, the "AI discovers better" claim is unproven at scale. REC-1245 DAHLIA H1 2026 readout is make-or-break.
Catalyst Runway (65): DAHLIA Phase 1 data H1 2026 (new information), REC-617 ovarian cancer data ongoing, Sanofi/Roche/Bayer collaboration milestones. Front-loaded catalyst calendar.
S/D (52): Cash runway to mid-2027 ($509M). But at $3.41 (down 72% from 52wk high), the market is deeply skeptical. Revenue (~$100M est.) heavily collaboration-dependent. No product revenue in sight for 3+ years.
Edge (55): NVIDIA backing is known. AI-drug thesis is increasingly covered. But clinical data readouts will be genuinely new information — either validating or destroying the thesis.
Regime (55): FDA end-animal-testing policy helps. But biotech funding environment challenging for cash-burning pre-revenue companies. Dilution risk real.

Ginkgo Bioworks
DNA · $9.12 · NYSE · Phase 3 Discovery
59
v2 Composite
Moat Durability
60
Catalyst Runway
65
S/D Trajectory
50
Edge Decay
65
Regime Alignment
58

Moat (60): OpenAI + Ginkgo autonomous lab experiments (Feb 5: 36,000 experiments, GPT-5 proposed novel reagents) positions DNA as the physical execution layer for AI biology. The "NVDA of Phase 3" thesis — but financials remain deeply distressed (Z-Score -7.01). Platform is unique; solvency is the question.
Catalyst Runway (65): Autonomous lab expansion with OpenAI, synthetic biology commercial deals, potential biosecurity contracts. Mostly new-information events — market hasn't modeled autonomous lab scale.
S/D (50): Revenue declining, margin negative, cash burn significant. Customer demand for cell programming exists but hasn't scaled to profitability. Supply of competing bio-foundry capacity is limited (moat), but demand hasn't arrived at price.
Edge (65): Autonomous AI biology is genuinely non-consensus. Most investors view DNA as a failed SPAC, not the physical execution layer for AI-driven biological discovery. If OpenAI partnership produces validated results, massive re-rating potential.
Regime (58): Biosecurity policy tailwinds. FDA animal testing reform helps. But financial distress and dilution risk weigh heavily.

Relay Therapeutics REVISED ↑
RLAY · $9.18 · NASDAQ · Phase 3 Discovery (Hybrid)
72
v2 Composite
Moat Durability
75
Catalyst Runway
85
S/D Trajectory
62
Edge Decay
60
Regime Alignment
72

Moat (75): Dynamo® platform uses molecular dynamics simulation (D.E. Shaw Research heritage) to capture protein motion — a structural differentiation no competitor has replicated. Zovegalisib is the first pan-mutant selective PI3Kα inhibitor ever developed — discovered through computational protein dynamics. FDA BTD validates the approach. Phase 3 trial (ReDiscover-2) underway. This is no longer speculative — it's clinically validated and regulatory-endorsed.
Catalyst Runway (85): This is the highest catalyst density in our AI-medicine universe. March 16: ESMO data (400mg BID fed Phase 3 dose — first public disclosure). Mid-2026: ReDiscover-2 topline Phase 3 data. NRAS program advancing. Fabry disease clinical initiation. Vascular malformations expansion. 5+ distinct catalysts, majority are new-information events that the market has not modeled.
S/D (62): PIK3CA-mutant HR+/HER2- breast cancer = ~140K patients/yr in US alone. Plus 170K vascular malformation patients. Addressable market is massive if zovegalisib delivers. Competitive threat from Celcuity's gedatolisib (broader PI3K inhibitor) — but RLAY's mutant-selectivity may offer superior tolerability.
Edge (60): BTD announcement moved the stock but it's still 91% below 52-week high of $9.54. Wait — actually RLAY is at $9.18 vs 52-week high of $9.54, meaning it's near highs now after massive recovery from $1.77 lows. The BTD and Oppenheimer upgrade are becoming known. But Phase 3 topline data is genuinely unmodeled — new information.
Regime (72): FDA extremely supportive — BTD granted. Precision oncology regulatory pathway is well-established. $596M cash runway into 2029 eliminates the funding risk that killed RLAY's thesis previously. This is the most de-risked AI-discovery asset in the universe.

Prior Assessment (Feb 4): Alpha 50 · EXIT · "Down 47%. Low alpha. Cash runway risk."

Revised Assessment (Feb 18): Alpha 72 · HOLD / ADD · FDA BTD transforms risk profile. $596M cash runway eliminates funding concern. Stock has re-rated from $1.77 to $9.18 (+419%). Phase 3 data mid-2026 = highest-conviction near-term catalyst in AI discovery.

What changed: Everything. BTD is the FDA explicitly endorsing Dynamo-discovered molecules as clinically meaningful. This is the proof-of-concept we said Phase 3 names needed to validate the capex feedback thesis. RLAY is delivering it.

V. Master Comparison Matrix

All five AI-medicine names ranked across key dimensions. Prices as of Feb 18, 2026 intraday.
Attribute TEM RLAY SDGR DNA RXRX
Phase PH.2 PH.3 PH.3 PH.3 PH.3
Price $57.39 $9.18 $11.87 $9.12 $3.41
Market Cap $9.92B $1.59B $874M $518M $1.37B
52w High / Low $104 / $36 $9.54 / $1.78 $28.47 / $11.15 $17.58 / $5.00 $12.36 / $2.98
% from 52w High -45% -4% -58% -48% -72%
vs 50D MA -12% below +13% above -28% below ~flat -22% below
vs 200D MA -18% below +71% above -40% below -12% below -31% below
Revenue (TTM) ~$1.27B ~$8.4M ~$225M ~$75M* ~$100M*
Revenue Growth +83% N/M +21% declining +22%
Profitable? Near breakeven No No No No
Cash Runway $764M $596M → 2029 Funded by SW rev At risk $509M → mid-27
v2 Alpha 68 72 63 59 57
Signal WATCH HOLD / ADD HOLD SPEC HOLD SPEC HOLD
AI Methodology Classification / pattern matching on clinical data Molecular dynamics simulation (protein motion) Physics-based FEP+ molecular simulation Autonomous AI-driven biological experiments High-dimensional imaging + ML drug discovery
Compute Intensity LOW HIGH VERY HIGH HIGH HIGH
Capex Feedback Minimal — data licensing, not compute demand Moderate — validates AI drug design Strong — pharma buys compute for FEP+ Strong — autonomous labs need GPU clusters Moderate — imaging pipeline is GPU-intensive
Next Catalyst Feb 24 earnings Mar 16 ESMO data Feb 25 earnings OpenAI lab updates DAHLIA H1 2026

*Revenue estimates based on available data. Confirm via earnings releases. All prices intraday Feb 18, 2026.

VI. The Capex Feedback Loop — Why Phase 3 Matters More Than Phase 2

Phase 3 discovery is the single largest accelerant of Phase 1 infrastructure demand in existence.

The Mechanism

Phase 3: AI discovers novel molecule
Drug generates $15B/yr peak revenue
Every pharma needs GPU clusters for simulation
Phase 1: NVDA sells more GPUs
Hyperscalers build more DCs
Power, cooling, memory demand spikes
Phase 1 capex restarts at 2-3x intensity

TEM's Capital Flow

TEM's success → pharma buys data licenses → $ flows to TEM, not to GPU manufacturers. The capital stays within the Phase 2 deployment layer. More patients sequenced, more data collected, more algorithms trained — but on existing clinical infrastructure, not new GPU clusters.

Infrastructure demand impact: Incremental.

Phase 3's Capital Flow

SDGR/RXRX/DNA/RLAY success → pharma buys compute infrastructure → $ flows to NVDA, hyperscalers, power providers, memory manufacturers. The capital propagates from discovery through the entire Phase 1 supply chain. One successful AI-discovered drug validates the entire computational approach for an industry.

Infrastructure demand impact: Exponential.

The RLAY inflection: If zovegalisib's Phase 3 data reads out positive in mid-2026, it will be the first drug designed through AI-powered molecular dynamics simulation to demonstrate Phase 3 clinical efficacy. That single data point re-rates the entire AI-drug discovery sector — SDGR, RXRX, DNA, and potentially TEM's pharma partnerships — while simultaneously triggering the compute demand cycle that validates Phase 1 infrastructure investments. This is why RLAY's alpha score just jumped from 50 to 72.

VII. Investment Implications for Apex Capital

Actionable conclusions within the 4-Phase Cyclical Framework.

RLAY — Upgrade to Active Position Candidate

Prior: EXIT at $4.50, Alpha 50. Revised: Alpha 72, highest in AI-medicine universe.

The FDA BTD, $596M cash runway into 2029, and Phase 3 trial in a massive indication (PIK3CA-mutant breast cancer, ~140K US patients/yr) fundamentally changed this name. At $9.18, the stock has already re-rated +419% from $1.77 lows — but the Phase 3 topline data hasn't arrived yet. That's the highest-conviction catalyst in AI discovery for 2026.

Risk: Phase 3 failure is binary. Celcuity's gedatolisib is a competitive threat. At $1.59B market cap, RLAY is pricing in some success. The stock is near 52-week highs — chasing is suboptimal.

Sizing: Treat as an option — 0.5-1% portfolio weight. If adding, wait for pullback toward $8 support (near 50-day MA of $8.12). Stop below $6.50.

SDGR / RXRX / DNA — Hold Existing Thesis Positions

No changes to Phase 3 Discovery option book. SDGR (Feb 25 earnings) is the most important near-term catalyst for the computational engine thesis. RXRX's DAHLIA data in H1 2026 is the highest-risk/highest-reward readout. DNA remains the speculative long-shot on autonomous AI biology.

All three names are significantly below their 52-week highs (58%, 72%, 48% respectively). The market is deeply skeptical of AI-drug discovery. If any of the three catalysts validates the thesis, the sector re-rates violently. This is why we hold them as options, not as core positions.

Summary Action Table

TickerPhasev2 AlphaSignalPositionRationale
TEM PH.2 68 WATCH 0% Evaluate post-Feb 24 earnings. Best AI-applied medicine. Not Phase 3.
RLAY PH.3 72 ADD 0.5-1% Highest alpha in AI-medicine. BTD validates. Phase 3 data mid-2026.
SDGR PH.3 63 HOLD 1% Computational engine. Feb 25 earnings. Software floor supports.
DNA PH.3 59 SPEC HOLD 0.5% Autonomous AI lab thesis. Financial distress risk. High edge.
RXRX PH.3 57 SPEC HOLD 1% DAHLIA H1 2026. Binary. NVIDIA-backed. 65PB dataset.

VIII. The Bottom Line