GET IN TOUCH

Investor Deck — 2026

Fidelity The world’s first confidential AI development environment See the coverage

The Confidential Agentic Stack

Nutmeg is the confidential stack for AI agents in defense, finance, and
regulated industry from factory floors to the front lines.

Built by a world-class team previously at

nutmeg-logo-white
nutmeg-logo

02 — Problem

The most sensitive teams are locked
out of AI.

Every keystroke in a mainstream AI tool sends a company’s most valuable IP to infrastructure it cannot control — third-party servers, LLM APIs, and systems far beyond its security perimeter.

Compliance certificates check boxes; they don’t stop source code and proprietary data from leaving the building. Today every AI tool asks customers to trust its infrastructure. For defense, finance, and regulated industries, vendor opacity, unaudited inference, and training-data leakage make that disqualifying — so they ban the tools or quietly absorb the risk.

The status quo

Vendor opacity
Unaudited inference
Training-data leakage
No proof of where code ran

03 — Why now

AI adoption is colliding with regulation,
security, and sovereignty.

01

AI is mandatory, trust is not optional

Every regulated enterprise is being pushed to adopt AI while its own security teams refuse to send sensitive code or data to opaque cloud APIs.

02

AI is mandatory, trust is not optional

Every regulated enterprise is being pushed to adopt AI while its own security teams refuse to send sensitive code or data to opaque cloud APIs.

03

AI is mandatory, trust is not optional

Every regulated enterprise is being pushed to adopt AI while its own security teams refuse to send sensitive code or data to opaque cloud APIs.

04 — Solution

Confidential by design. Verifiable
by default.

Orgn was founded on the conviction that confidential code deserves confidential AI. The platform secures the development environment end-to-end and lets customers choose the right assurance level per workflow.

For confidential workflows, Orgn generates cryptographic attestation that proves the workload ran inside a verified enclave — making confidentiality something customers can verify, rather than trust.

Our charter

To make AI development safe for the most sensitive code on Earth.

05 — Platform

End-to-end infrastructure for AI
workloads that cannot be exposed.

NUTMEG

Develop with AI assistance while repositories, prompts, terminals, and execution context stay inside your security boundary.

NUTMEG

Develop with AI assistance while repositories, prompts, terminals, and execution context stay inside your security boundary.

NUTMEG

Develop with AI assistance while repositories, prompts, terminals, and execution context stay inside your security boundary.

NUTMEG

Develop with AI assistance while repositories, prompts, terminals, and execution context stay inside your security boundary.

06 — Target market

Built for teams that cannot compromise
on security.

01

AI is mandatory, trust is not optional

Every regulated enterprise is being pushed to adopt AI while its own security teams refuse to send sensitive code or data to opaque cloud APIs.

02

AI is mandatory, trust is not optional

Every regulated enterprise is being pushed to adopt AI while its own security teams refuse to send sensitive code or data to opaque cloud APIs.

03

AI is mandatory, trust is not optional

Every regulated enterprise is being pushed to adopt AI while its own security teams refuse to send sensitive code or data to opaque cloud APIs.

04

AI is mandatory, trust is not optional

Every regulated enterprise is being pushed to adopt AI while its own security teams refuse to send sensitive code or data to opaque cloud APIs.

07 — Market

A large, expanding market for
confidential AI infrastructure.

TAM $1.3T

Total AI spending by 2029, with agents as the primary driver (IDC).

TAM $1.3T

Total AI spending by 2029, with agents as the primary driver (IDC).

TAM $1.3T

Total AI spending by 2029, with agents as the primary driver (IDC).

TAM $1.3T

Total AI spending by 2029, with agents as the primary driver (IDC).

33% of enterprise apps will embed agentic AI by 2028, up from under 1% in 2024 (Gartner).

08 — Traction

Shipping product, broad model coverage,
and independent validation.

250+

TEE and Zero Data Retention (ZDR) models routable through one Gateway API.

4 layers

TEE and Zero Data Retention (ZDR) models routable through one Gateway API.

World's first

TEE and Zero Data Retention (ZDR) models routable through one Gateway API.

09 — Business model

One ladder: self-serve credits up to enterprise ACV.

Pay as you go

$20to start

Prepaid credits at provider cost plus a transparent 10% markup. No subscription — the on-ramp for individuals and pilots.

Pay as you go

$20to start

Prepaid credits at provider cost plus a transparent 10% markup. No subscription — the on-ramp for individuals and pilots.

Pay as you go

$20to start

Prepaid credits at provider cost plus a transparent 10% markup. No subscription — the on-ramp for individuals and pilots.

Pay as you go

$20to start

Prepaid credits at provider cost plus a transparent 10% markup. No subscription — the on-ramp for individuals and pilots.

10 — Competitive landscape

The pieces exist. The full confidential dev stack does not — except Orgn.

Capable AI dev tools are cloud-dependent and unverified. Confidential inference is real but stops at the API. Air-gapped and ops platforms are narrow or built for operations, not software engineering.

Adjacent players lead one axis — confidential inference (Tinfoil, Edgeless, Opaque) or agentic dev (Cursor, Tabnine, Palantir AIP). None deliver both: a verifiable, attested stack across IDE, Gateway, Agents, and Attestation.

11 — The plan

The roadmap

Trusted Execution Environment (TEE) inference ships in the first release.
Then it scales with owned GPU infrastructure.

Pay as you go

$20to start

Prepaid credits at provider cost plus a transparent 10% markup. No subscription — the on-ramp for individuals and pilots.

Pay as you go

$20to start

Prepaid credits at provider cost plus a transparent 10% markup. No subscription — the on-ramp for individuals and pilots.

Pay as you go

$20to start

Prepaid credits at provider cost plus a transparent 10% markup. No subscription — the on-ramp for individuals and pilots.

Pay as you go

$20to start

Prepaid credits at provider cost plus a transparent 10% markup. No subscription — the on-ramp for individuals and pilots.