Healthcare Outcome Procurement + The Pioneering-Provider Wedge

The question

Given Diamandis/Wissner-Gross "Solve Everything" Phase 2 (2028-2031, Virtual Cell, body-as-software), what commercial wedges are actionable in 2026-2027 for a pioneering provider with patient-data-sovereignty as cost-side advantage? Origin: Michael Holzum's 2026-05-10 unprompted frame ("outcome procurement, compute escrow, data trusts") + today's harness-moat / Whoop-MCP work. See [[03-contacts/michael-holzum]].

What we already know (vault)

• [[06-reference/book-solve-everything-master-synthesis-2026-04-13]]: the four-stage revolution arc (legibility → harnessing → institutionalization → abundance) and the closing thesis that aiming becomes scarce once intelligence is cheap. The healthcare-data-trust + compute-escrow stack is exactly the institutionalization-stage rail this brief proposes building.
• [[06-reference/book-solve-everything-ch1-war-on-scarcity-2026-04-13]]: explicit Ch.1 imperatives - "Replace effort-based procurement with outcome-based contracts" + "Escrow compute tied to performance milestones." Holzum's frame IS Ch.1 applied to CMMI.
• [[06-reference/book-solve-everything-ch7-the-moonshots-2026-04-13]]: Organ Abundance + Double Human Healthspan moonshots. Outcome-procurement on cardiometabolic/CKD panels is the wedge that finances the data substrate the healthspan moonshot needs in Phase 2.
• [[06-reference/book-solve-everything-ch9-build-the-rails-2026-04-13]]: the ten-gear implementation engine. Three of the ten gears - outcome payment, compute escrow, data pooling via trusts - are precisely the three institutional rails Holzum named unprompted. His insight is the Ch.9 rails thesis instantiated into US healthcare.
• [[06-reference/2026-04-30-rdco-thesis-targeting-systems-feedback-loops]]: RDCO moat = "where to aim feedback loops." Outcome-bid pricing IS a targeting-system play.
• [[06-reference/concepts/2026-04-23-unhobbling]]: as model capability gets unhobbled, the verification layer is what survives. An outcome-scored healthcare panel IS a verification layer; the fiduciary data trust is the unhobbling-resistant infrastructure underneath.
• [[06-reference/concepts/2026-04-24-targeting-system]]: the canonical RDCO term. Sub-question (b)/(c) of this brief - data trusts + compute escrow - are the targeting-system mechanism instantiated in healthcare. Outcome-procurement contracts ARE the acceptance criteria; the panel-level scoring IS the test harness.
• [[06-reference/concepts/2026-04-24-three-decision-algorithms]]: in healthcare specifically, the implicit-targeting-system layer (carried clinical analogies, senior-provider taste) is exactly what value-based care has been trying to discipline for two decades. An outcome-scored panel forces the agentic-targeting-system shift the 3DA collapse is making cheap everywhere else.
• [[06-reference/concepts/2026-05-10-harness-moat-two-layers-portability]]: Layer-1 universal harness directly reusable as the back-office for an outcome-scoring pipeline.
• [[06-reference/2024-11-28-moonshots-ep133-ai-healthcare-fii-panel]]: Whoop-grade continuous monitoring as the leading-indicator data substrate.

The Solve Everything framing (Phase 2)

Phase 2 (2028-2031) collapses biology into a software problem once the Virtual Cell hits debuggable fidelity. Bottleneck: multi-scale biological complexity (compute-allocation, not insurmountable). Critical: the building blocks already in place in 2026 include "established clinical data pipelines." The pipelines ARE the wedge. Whoever owns consented, longitudinal, high-fidelity outcome data 2026-2028 rides Phase-2 cost collapse without rebuilding the data layer. Phase-2 timing is forward-looking; the wedge thesis stands even if Virtual Cell slips to 2032+.

(a) ACA/CMS value-based contract landscape 2026

Bid window is the widest since CMMI's 2010 founding. Specifics:

• ACO REACH (sunset Dec 2026, replaced by LEAD). PY 2026: 74 ACOs, 125,909 providers, ~1.7M Traditional Medicare beneficiaries. Global Option = 100% Parts A+B risk, 3.5% mandatory discount, Quality Withhold raised 2% to 5% for PY 2026. Sunsetting - not the durable bet.
• LEAD Model (Jan 1 2027, runs 10 years through 2036). ACO REACH's successor. 10-year runway = longest CMMI commitment on record. Aimed at smaller, rural, high-needs clinicians. Application window opens 2026 for Jan 2027 start. Highest-leverage greenfield bid surface.
• ACCESS Model (launches Jul 1 2026; cohort-1 deadline Mar 20 2026 closed; cohort-2 to follow). Four tracks: early cardiometabolic/kidney (eCKM), comprehensive cardiometabolic/kidney (CKM), musculoskeletal (MSK), behavioral health (BH). Recurring Outcome-Aligned Payments (OAPs) tied to BP, HbA1c, lipids, eGFR, PHQ-9. Co-management ~$30/review + $10 onboarding ($100/beneficiary/year baseline floor). Health-tech companies must enroll in Medicare or contract through enrolled providers. eCKM/CKM = named cardiometabolic outcome-procurement market.
• Kidney Care Choices (extended through 2027). 74 KCEs, 7,534 providers, 237,000 CKD/ESRD beneficiaries. Global = 100% risk. PY 2023 produced ~$304M net loss to Medicare; 2026 re-pricing introduces participant discounts and cuts the kidney-transplant bonus. CMS is actively re-pricing because incumbents under-delivered.
• Enhancing Oncology Model (cohorts run to 2030). 28 practices, 2,000+ practitioners, six-month chemo episodes for 7 cancers. Mature, less greenfield.
• Sunset cluster. Making Care Primary, Primary Care First, Maryland TCOC, ETC all ending early per the 2025 CMMI restructuring.

Widest bid window 2026-2027: LEAD (2027 cycle, brand-new, 10-year) and ACCESS eCKM/CKM (2026 launch, OAP not yet calibrated).

(b) Data trusts in US healthcare 2026 - missing legal middleware

Status: still essentially undeployed at scale. UK NHS / DeepMind catalyzed the modern lit (Sean McDonald via Digital Public, CIGI 2024 "Reclaiming Data Trusts") but US healthcare did not adopt the fiduciary form. What deployed instead is structurally weaker:

• TEFCA = federated info-exchange protocol with QHIN intermediaries, NOT a fiduciary trust. Jan 1 2026: QHINs must adopt HL7 FAST for FHIR. Many large systems still hadn't signed by late 2025. Robust payer adoption pushed to 2026-2027.
• Datavant moves 60M+ records/year across 70K+ hospitals via PPRL. Commercial broker, not a fiduciary - the patient does not direct it.
• EU EHDS (in force Mar 26 2025) establishes Health Data Access Bodies. US has no federal equivalent.

The wedge: legal middleware Holzum named is genuinely missing. A pioneering provider that bundles a fiduciary vehicle (state-chartered trust or LLC-trust hybrid in Delaware/Wyoming/NH) with patient consent + a TEFCA QHIN connection holds a position no incumbent currently holds. First-mover risk real but defensible. Speculative confidence: medium.

(c) Compute escrow + PETs for clinical data

Tech maturity high; adoption bimodal (research wide, production narrow).

• Federated learning - production exists. Mayo Clinic Platform's "data under glass" (Sep 2019 Google partnership, 10-year): algorithms enter the enclave, data never leaves. Stanford AI Lab's 50-hospital federated network reported 89% accuracy gain on rare adverse-drug-event prediction. NVIDIA Clara FL is shipped product. Maturity 8/10. Outcome-contract integration ~1/10. Gap is commercial, not technical.
• Multi-party homomorphic encryption (MHE). FAMHE (Nature Communications 2021) is the reference architecture. HEPRS handles polygenic risk scores in clinician/modeler/evaluator topology. 2025 PubMed work handles HE-encrypted clinical-trial opt-out. Maturity 7/10, adoption ~1/10.
• Trusted Execution Environments. AWS Nitro Enclaves now ship zero-trust GenAI patterns for healthcare per AWS official blog. Caveat: Nitro doesn't encrypt memory like Intel SGX did; Intel deprecated SGX on consumer CPUs and is moving to TDX. Maturity 9/10, healthcare adoption ~2/10.

The wedge: "compute escrow" Holzum named is buildable today off-the-shelf (MHE for sensitive computations + Nitro Enclaves for inference + federated coordination). Nobody has packaged the three into a clinical-outcome-scoring product. Tech-risk low; integration + regulatory risk medium-high.

(d) Sizing - 1-3 outcome-category bids

Pick: (1) ACCESS eCKM/CKM, (2) LEAD ACO 2027, (3) KCC Global as fallback.

Bid 1 - ACCESS eCKM/CKM cardiometabolic. Addressable Original Medicare population with hypertension + 2 cardiometabolic comorbidities ~ 12-15M (CDC: 60% of 65+ have HTN; ~30% diabetic/pre-diabetic; heavy overlap). CMS spends $95.7B/year on CKD alone (2024 USRDS = ~1 in 4 Medicare dollars). Avg diabetic Medicare beneficiary: complications add ~$10-15K/year. Baseline OAP ~$100/beneficiary/year is the floor. A 5% reduction in cardiometabolic event rate on a 100K-beneficiary panel saves Medicare ~$50-75M/year; Global-style 50% shared savings = $25-37M/year to provider. Pioneering-provider gross margin 40-60% if measurement infrastructure is amortized across tracks (the data-sovereignty cost advantage). Plausible bid: 100K-beneficiary panel, $25-50M annual revenue at maturity.

Bid 2 - LEAD ACO 2027. Inherits ACO REACH structure (Global = 100% Parts A+B risk, 3.5% discount, 5% Quality Withhold). Per-beneficiary Traditional Medicare spending ~$13K/year. A 50K-beneficiary high-needs ACO at $13K = $650M annual benchmark; 3% net savings = ~$20M to provider after discount. 10-year runway means LTV stacking favors data-infrastructure-heavy entrants - cost-side moat compounds. Highest-LTV bid in the brief.

Bid 3 - KCC Global (fallback / 2027 entry). Per-beneficiary CKD/ESRD ~$70-90K/year (USRDS dialysis-stage). 5K-beneficiary KCE at $80K = $400M benchmark, 5% savings = $20M. PY 2023 $304M Medicare loss tells us bid pricing currently mispriced AGAINST CMS, hence 2026 re-cut. Wait for re-priced terms before bidding.

Recommendation: lead with ACCESS eCKM/CKM (smaller bites, faster proof, cohort-2 cycle), build the data-trust + compute-escrow infrastructure on its back, then bid LEAD 2027 with infrastructure already operating.

Synthesis for RDCO

Michael Holzum's "outcome procurement / compute escrow / data trusts" insight is not a novel external thesis. It is the targeting-system framework [[06-reference/concepts/2026-04-24-targeting-system]] applied to value-based healthcare. The "pioneering provider with patient-data-sovereignty as the cost-side advantage" is the targeting-system instantiated as a healthcare bet. The Diamandis/Wissner-Gross "Build the Rails" frame [[06-reference/book-solve-everything-ch9-build-the-rails-2026-04-13]] is what Michael recognized when he named "outcome procurement, compute escrow, data trusts" as the three institutional rails healthcare needs. RDCO already operates the targeting-system as a discipline; the question is whether to instantiate it as a healthcare bet, partner with a healthcare-native operator (Michael being the obvious candidate), or stay in the harness layer and let others build the bets.

The pioneering-provider wedge is structural and time-boxed. Structural: outcome procurement requires measurement; measurement requires consented longitudinal data; consented data requires a fiduciary vehicle; the fiduciary vehicle does not exist in US healthcare today. Whoever builds it captures the measurement layer for the next decade of value-based care + the Phase-2 Virtual Cell era. Time-boxed: ACCESS launches Jul 2026, LEAD applications open 2026 for Jan 2027 start.

What RDCO would need:

• Build: agent harness running outcome-scoring, consent UX, dashboards. Whoop MCP is a tiny proof of the data-sovereignty UX. COO-agent harness work directly reusable.
• Partner: a Medicare-enrolled provider entity (or rapid enrollment path), MHE/TEE specialist (Lifebit, OpenMined-adjacent), healthcare-compliance lawyer fluent in TEFCA + state trust law.
• License: Datavant for record linkage; AWS Nitro for compute escrow; an EHR-FHIR integrator.

Co-founder fit - Michael Holzum. Strong on three dimensions: finance = bid pricing/contract structuring/capital; healthcare = payer relationships, regulatory navigation, provider credibility; structural-thinking pattern (the unprompted Solve-Everything-on-CMMI insight) = strategy seat, not operational. Most likely role: co-founder + healthcare lead, Ben on product/agent, Ray as harness. Single biggest unknown: salary tolerance. Worth a structured conversation, not a casual ask.

Realistic timeline:

• 2026-Q3: Whoop MCP ships; Michael installs as second user; founder-Michael strategic conversation; legal entity formed.
• 2026-Q4: ACCESS cohort-2 application; MHE/TEE prototype on Mayo-style federated test bed.
• 2027-Q1: LEAD application cycle; first paying outcome contract live (sub-1K beneficiary panel).
• 2027-Q4: LEAD acceptance; first ACCESS performance year complete.
• 2028: scale on running infrastructure when Phase-2 Virtual Cell discourse goes mainstream.

Risks + objections

• Regulatory uncertainty. State trust law for healthcare data is uncodified; aggressive state AG could stall. Mitigation: permissive jurisdiction (DE/WY/NH), fiduciary-LLC with explicit beneficiary class.
• Incumbent capture / political cycle. 2025 CMMI restructuring sunset four models; LEAD or ACCESS could get cut. Mitigation: ACCESS launches under existing authority, hard to undo mid-cycle; LEAD has 10-year statutory commitment.
• Consumer-side adoption. Patients may not value data-sovereignty enough to switch providers. Mitigation: B2B2C through ACO panel - panel-attribution does the choosing.
• Tech-stack assembly. MHE+TEE+federated coord not packaged as a clinical product. Mitigation: lead with TEE (lowest risk), add MHE only for cross-institution computations.
• Founder bandwidth. RDCO already has Squarely, Sanity Check, MAC, COO unhobbling. This is a fifth bet. Honest read: doesn't fit the portfolio without dropping something OR Michael takes operating-lead so Ben stays on existing four.
• Phase-2 timing. Virtual Cell may slip to 2032+. Wedge thesis still works on VBC economics alone, but longevity-bet upside compresses.

Open follow-ups

• ACCESS Model OAP dollar levels - CMS hasn't published bonus thresholds. Federal Register watch.
• LEAD application timing + risk-adjustment methodology - not yet final.
• State trust statutes that would support a healthcare-data fiduciary - legal research, not internet research.
• Is any 2026 startup already targeting this wedge (Truveta, Verana, Komodo are adjacent but not the wedge as framed)?
• Patient-side acquisition cost for a data-sovereignty-first care brand - no analog; closest is Iora/One Medical at ~$300-500 CAC for a Medicare panel.
• Whoop or Oura partner on consumer data-pipe vs build in-house?
• Find out if Michael actually wants to BUILD this vs treat it as an interesting frame.

Sources

Vault: ~/rdco-vault/03-contacts/michael-holzum.md, book-solve-everything-ch1-war-on-scarcity-2026-04-13.md, concepts/2026-05-10-harness-moat-two-layers-portability.md, 2026-04-30-rdco-thesis-targeting-systems-feedback-loops.md, 2024-11-28-moonshots-ep133-ai-healthcare-fii-panel.md.

Vault (Solve Everything + RDCO canonical concepts, added on patch): [[06-reference/book-solve-everything-master-synthesis-2026-04-13]], [[06-reference/book-solve-everything-ch7-the-moonshots-2026-04-13]], [[06-reference/book-solve-everything-ch9-build-the-rails-2026-04-13]], [[06-reference/concepts/2026-04-23-unhobbling]], [[06-reference/concepts/2026-04-24-targeting-system]], [[06-reference/concepts/2026-04-24-three-decision-algorithms]].

Web (accessed 2026-05-10): solveeverything.org (Phase 2); cms.gov ACO REACH / LEAD / KCC / EOM PY2026 fact sheets; manatt.com ACCESS Model unlocked; milliman.com ACO REACH PY2026; usrds-adr.niddk.nih.gov 2024 ESRD expenditures; healthit.gov TEFCA; datavant.com; cigionline.org Reclaiming Data Trusts (McDonald); nature.com FAMHE (s41467-021-25972-y); lifebit.ai 2025 federated-learning; aws.amazon.com Nitro Enclaves zero-trust healthcare; ncbi.nlm.nih.gov Mayo-Google "data under glass" (NBK594445); pubmed.ncbi.nlm.nih.gov 40588977 (HE-encrypted trial opt-out 2025).