Internet Labs

Investor presentation · October 2026

Frontier superintelligence for enterprises

The thesis

Frontier models have read the internet

The knowledge that runs an enterprise was never on it

That knowledge is private, tribal and fragmented, and no frontier lab will ever train on it. Internet Labs builds frontier superintelligence on it, inside each client’s own walls: compiled into governed memory, put to work in every department, and learning from every day of work.

The knowledge
Private, tribal and fragmented: in documents, mail, tickets and people’s heads.
The system
Internet SI OS compiles it into governed memory; approved work becomes golden data.
The model
Valor is post-trained on that golden data, which frontier labs never train on.

Nobody offers this. A few years from now, a company will not be able to run, let alone sustain itself, without Internet SI OS. That is why Internet Labs exists.

A long records-room aisle at blue hour, a lamp lit at its far end

Contents

Enterprise AI spending is up a third on a year

The returns lag, held back at the foundation

$173M
average planned AI spend over the next 12 months at $1B+ US companies, up 33% on a year
37%
of companies see any EBIT impact from AI
72%
of US leaders say they lack unified, accessible data

The main constraint is the enterprise’s own knowledge: fragmented, ungoverned and unsafe to hand to autonomous agents.

The opportunity

Spending rises; results don’t

Four leaks between the budget and the P&L

Planned AI spend

$173M

per $1B+ US company, over the next 12 months (KPMG, Sep 2026)

Knowledge
60%
of AI projects without AI-ready data abandoned through 2026 (Gartner)
Trust
65%
have seen agents act beyond their scope; 34% check authorisation when an action runs (EMA)
Ungoverned building
77%
say AI adoption is outpacing their governance (IBM)
After the vendor leaves
70%
of agentic AI built by vendors’ forward-deployed teams abandoned by 2028 (Gartner)

Reaching the P&L

37%

of companies see any EBIT impact from AI (McKinsey, Aug 2026)

42%

abandoned most AI initiatives in 2025, up from 17% (S&P Global)

What the platform must do

Compile the foundation
Keep it permissioned and cited
Let every team build, under IT’s control
Leave the client able to evolve it

Internet SI OS closes all four, inside the client’s own environment.

The opportunity

Why now

Five shifts in twelve months

Budgets doubled and moved to the business
AI spend is 3.3% of revenue, over 80% outside IT.
Compiling an enterprise got cheaper
A whole company wave compiles for an estimated $7,600 of model spend.
Sovereignty became a buying criterion
77% of companies weigh an AI vendor’s country of origin.
Forward-deployed delivery is in demand and capitalised
Palantir’s US commercial revenue grew 149%.
Gartner put a number on the failure mode
Gartner expects 70% of enterprises to abandon vendor-FDE-built agentic AI by 2028.

Internet Labs

Six businesses, one frontier stack

Business, in order of priority	Role	How it earns
01Internet SI OSThis round · built in 12+ months	Compiled memory, expert agents, continual learning	Setup $5M+, then 70% of setup a year
02FieldFrom the first paid Proof	Forward-deployed engineers who install the OS	Delivery in the OS fees; model builds $8M+
03ResearchAlongside the OS	Continual memory, private training, routing and evaluation	Compounds the OS and Valor
04ValorNext round · v1 on Gate 1	The frontier model line, post-trained on each client’s golden data	Per token: the cost to serve, plus a margin
05InfrastructureAfter Gate 1	The GPUs under Valor, owned and operated	About $4.50 a B200-hour; servers at cost plus 10–15%
06Protocol and platformOn the installed base	The Emissary as an open standard	Marketplace take-rate, certification, a per-seat edition

All six businesses are pre-revenue.

A man at a window at blue hour, a lamp lit on the desk beside him

Internet Labs

At a glance

What we are: A research lab building frontier superintelligence for enterprises.Internet SI OS is the lab’s first product.
Who buys: Every company with $100M+ of revenue.
How we earn: OS setup and annual fees; seats, Valor, Field and compute.
The market: TAM $10.3T by 2030. SAM $1.01T a year; year-5 revenue $7.1B.
Where we are: All six businesses are pre-revenue. Twelve months to build the OS.
The round: $20M pre-seed for 18 months.

The first product$5.5M to build the OS and reach Gate 1

Internet SI OS

The operating system for enterprise superintelligence

01
Compiled
memory
Documents, mail and archives, compiled into cited, permissioned pages, kept current.
02
Expert
agents
One for every department; a person approves anything sent, spent or published.
03
The Emissary and
the Orchestrator
Departments exchange verified answers, not files; requests reach the right model.
04
Continual
learning
Finds its own gaps, tests fixes; what passes becomes golden data.

One install: the whole company
Unlimited departments: added as data, no code change
Any environment: cloud, on-prem or air-gapped
Any model: on no single vendor

Internet SI OS

One system for the whole company

Four technologies at its core

Internet SI OS compiles what a company knows into governed memory, puts expert agents to work in every department, and learns from every day of work, in the company’s own environment and under human approval.

01
Memory
vault
Every document, mail, ticket, recording and scanned archive, compiled into cited, permissioned pages.
02
Pragmatic
model routing
Sensitivity decides first, then difficulty: every request reaches the model that serves it best.
03
Reversible semantic
pseudonymisation
Identifiers become realistic twins, so models learn without seeing a real identifier.
04
Continual
learning
Finds its own gaps, tests fixes on real work and keeps what passes.

One install: the whole company
Unlimited departments: added as data
Any environment: cloud, on-prem or air-gapped
Any model: on no single vendor

Internet SI OS

How it works

Knowledge in, intelligence out, every change approved

Daily work read bottom to top

- Company material
- Scanned paper
Knowledge Compiler a source on every page
Memory vault
Orchestrator pragmatic routing
- Expert agents
- The Emissary
- Approved models
People, in one workspace

Improvement read top to bottom

Capture
Agents propose
The owner approves
Build or learn
Tested on the living benchmark
A person releases it

Internet SI OS

One install, every department

Nine presets ship ready; any other is added as data, not code

HR
Leave requests, onboarding checklist, policy questions
Finance
Invoice approvals, expense checks, month-end checklist
Sales
Proposal drafts, discount approval routing, account research
Marketing
Brand review requests, campaign briefs, launch checklists
Design
Asset requests, design review intake, research summaries
Engineering
Incident review drafts, runbook lookups, on-call handover
Legal
Contract intake against the playbook, NDA requests
IT
New-hire provisioning, access requests, leaver offboarding
Security
Alert intake and triage, phishing triage, endpoint isolation with analyst sign-off
Any other department
Added as data: no code change, no redeploy

The technology

The Knowledge Compiler

From raw material to cited, permissioned pages

Raw material

A desk covered in loose papers under a lamp

Documents
Decks
Wikis
Mail
Tickets
Recordings
Paper

The compiler

01Map owners and permissions, before reading anything
02Label sensitive files before they are taken in
03Screen every source for planted instructions
04Compile versioned pages, a source on every claim
05Recompile only what changed, with one-step rollback

A compiled page Example · Finance

finance / wiki / expense-approvals.mdFinance · all staff

Expense approvals

Claims above a manager’s limit go to Finance for approval.¹

Travel booked outside the travel tool needs a written exception.²

Every claim carries its receipts.^{1 3}

1 Expenses policy.pdf · page 4
2 Travel guide · Finance wiki
3 Month-end checklist.xlsx

v14 · approved by the Finance ownerunder Git

Every claim opens to its source; access is fixed before the page exists.

90%
of each department’s in-scope material reached (target)
~10 days
to compile an enterprise wave (target)
$7,600
estimated model spend for a whole company wave, on an assumed corpus

The technology

The memory vault

The company’s long-term memory, compiled once and kept current

Raw sources

Documents
Mail
Tickets
Recordings
Scanned paper

Ingest

Knowledge Compiler

Reads each source once

Blue: proprietary core and human approval

The vault

One per department

Every claim carries its source.

index.md: read first
log.md: append-only

Schema

AGENTS.md: page types, citation rules
policy/: data classes and retention

Query

Agents

File back

memory/: proposals awaiting approval

Owner approves

Lint, nightly

2–3 months

to compile a whole company’s knowledge base, once

The technology

Why compiled beats retrieved

Knowledge that compounds

	Search and retrieval (RAG)	The memory vault
Assembled	On every question	Once, then kept current
What accumulates	Nothing	Every source and every approved answer
Citations	Chunks chosen by similarity	Every claim carries its source
Permissions	Filtered at answer time	Fixed before compiling
Change control	None: the index changes as the files do	Owner approval, one-step rollback
Where it lives	In a search index	Plain files in the client’s environment

The test, in every Proof: the vault against plain retrieval on the same model.

The technology

The Emissary

Answers across departments, never their bases

A person
The asker · in the example, a Sales lead
06Authorisation checked again when an action runs
The Emissary
Carries the asker’s identity on every hop
05A second model, from another family, tries to refute the answer
The other department’s agent
In the example, Finance’s expert agent
03Answers from its own vault; the base is never exposed

01The question
02Under the asker’s rights
04Derived material only
Released after the check

Example

“Can we offer this client a deeper discount?”“That is above the standard limit; it needs Finance sign-off.”

Finance’s approval limits never leave Finance; only the answer does. And only people start these requests, never agents.

How it is proven

Leak tests across every pair of departments, on every build.
Identity on every hop, so an answer can never carry more than the asker may see.
A second department live is the test the Emissary is switched on against, in v2.

The technology

Pragmatic routing

Sensitivity decides first, then difficulty

A request

01
Sensitivitysets the ceiling
02
Difficultypicks the first rung

Department adapter
Inside the walls · tried first
Checked against cited sources
Check fails: escalate
Open model
The client’s GPUs
Checked
Only if the data may leave
Regulated data stops here
Frontier model
An approved provider
Checked

up to 98%
lower cost at GPT-4-level accuracy on benchmark tasks (FrugalGPT, Stanford, 2023)
85%+
lower cost at 95% of GPT-4 quality on MT Bench (RouteLLM, 2024)

Every route, cost and outcome is logged; the router changes only after passing the living benchmark. No single model or routing vendor sits on the critical path.

The OS reports its own mix from the first pilot.

The technology

Pragmatic model routing

The right model for every request

Request

Orchestrator

Reads

01Sensitivity
02Difficulty
03The person
04The modePersonalisedEfficiencyPerformance

Inside the walls

Valor and department adapters
Open models on the client’s GPUs

Approved providers

Approved frontier models

Best output

outcomes refit the routes

Sensitivity decides first: no model receives data above the ceiling set for it.

Personalised
Learns from each person’s activity which routes give them the best results.
Efficiency
For a set budget: the lowest-cost route that delivers an excellent answer.
Performance
Budget is no constraint: the strongest model available.

Up to 98%
lower cost while matching GPT-4
85%+
lower cost at 95% of GPT-4 quality

The technology

Reversible semantic pseudonymisation

The model meets a twin, never the customer

Blue: proprietary core

The client’s environment

The model

Original record
1 · Semantic
finds direct and indirect identifiers by meaning
2 · Pseudonymise
each one replaced by a twin, faithful to what the task needs
Key store
the twin-to-original map never leaves
only the twin leaves
Any permitted model
the answer returns
3 · Reverse
originals restored at the boundary
The user

Illustrative record
Original	Twin	What is kept
Joan Mercer	Edith Hargreaves	a woman’s name
81	83	age band, over 80
Whitby	Filey	small coastal town, same region

The technology

Tested on 500,000 financial-services cases

Models learn without seeing a real identifier

	Keeps meaning?	Restorable?
Redaction
Masking with placeholders	partly
Tokenisation
Reversible semantic pseudonymisation

One identifier, on its way through · an example

In the client’s material
Priya Nair
account 40-1187-22 · Leeds
Its realistic twin
Hannah Cole
account 73-5520-09 · Bristol
What the model sees
Only the twin: the meaning holds, the person does not
Restored at the boundary
Priya Nair
account 40-1187-22 · Leeds
with the client’s key

96.2%
answer fidelity after twinning and reversal against 41.3% for masking, 12.8% for tokenisation
1.5%
leak rate under an active extraction attack against 7.8% and 22% for two open-source detectors
100%
of substitutions restored exactly

The technology

The improvement engine

Find the gap, test the fix, keep what passes

01
Exhaust
Traces, requests, corrections and outcomes
02
Gap discovery
- drift
- behaviour
- capability
- funnel
03
Evals
- tests
- judges
- slices
- guardrails
04
Interventions, and the owner approves
- prompt
- tool
- retrieval
- routing
- workflow
- memory
- skill
- adapter
05
Offline replay and verdict
Every fix is replayed on frozen real work before any user sees it
- Test
- Modify
- Discard
- Collect more data
06
Release and measure
A person releases it; the outcome is measured

System knowledge

Every turn leaves a test, judge, lesson or dataset behind.

The technology

Memory first, weights last

Four layers that learn

Memory
Lessons written into the vault once the owner approves
v1
Living benchmarks
A regression suite per department; nothing ships unless it passes
v1
Department adapters
A LoRA adapter per department, trained on golden data
v3
Routing
Routes refit on outcomes
v1

Why it compounds

Models are rented. Everything the system learned is kept.
What learns is the company’s own work: a record competitors cannot buy.
Every change is approved, tested and reversible.

−89%·−71%

Held-out NaturalQuestions F1 after full fine-tuning and after LoRA on new facts

The technology

Continual learning

Memory first, weights last, behind two human approvals

Memory

Every day · undone in one step

01
Agents log the day’s work to the department’s memory folder.
02
Overnight consolidation proposes cited updates to the wiki.
03
Approval 1
The department owner approves.
04
Each update becomes a commit in the vault.

Every day

Only approved, stable knowledge

Weights

Only when stable · versioned, removable

01
Pseudonymised first.
02
Trains a department adapter.
03
Beats the current one on the living benchmark, with no regression on a general test.
04
Approval 2
A person releases it.

+39%
agent performance with file-based memory and context editing, with 84% fewer tokens (Anthropic, Sep 2025)
−89% · −71%
held-out scores after full fine-tuning and after LoRA on new facts (Meta, Oct 2025): why weights come last

The vault stays the source of truth; every adapter can be removed.

The technology

From tribal data to golden data

The engine’s by-product trains Valor

Blue: proprietary core and human approval

Tribal data
Vault pages, traces, corrections, outcomes
RSP
reversible pseudonymisation
The map stays in the client’s key store
Judged and approved
Calibrated judges score it, the owner approves
Golden data
Versioned, frozen, inside the walls
Training
Department adapters, then the client’s own Valor
Valor
the in-house model
Released only when it beats the best permitted frontier model

Golden data, three kinds

Demonstrations
What an expert did
Every correction a person makes
Preferences
Which answer won
Every verdict
Tests
What must never regress
Every gap found

Build and control

Build a workflow in plain words

Described, drafted, tested, inside a governed build space

Workflow studioFinance build space · Invoice approvalsExampleOn a branch · tests passing

1 Describe it

“When an invoice arrives in the accounts-payable inbox, match it to its purchase order, flag duplicates, send anything over the controller’s limit for approval, and post approved invoices to the ERP.”

Plain words, from a Finance builderDraft the workflow

Or start from a template

Invoice approvals
Expense checks
Month-end checklist

Put a front end on it

Form
Page
Dashboard
Case tracker

Routes decide who can reach each one, inside or outside the company.

2 See it on the canvas

edits stay in step with the code

TriggerAn invoice arrives in the AP inbox
Tool · ERPschemaFind its purchase order
CodenumbersFlag duplicates, reconcile the totals
AgentcitationsThe Finance expert reads the approval limit
Approvalsign-inThe controller signs off anything over it
Tool · ERPoncePost the approved invoice

Each step’s output is checked before the next step reads it.

3 Or as code, in Git

id: finance-invoice-approvals
space: finance
consumers: [ap-team]
trigger: { mail: accounts-payable }
steps:
  - { id: po,    kind: tool,  tool: erp.find_po }
  - { id: dupes, kind: code,  check: [reconcile] }
  - { id: limit, kind: agent, agent: finance-expert }
  - { id: ok,    kind: approval, for: send,
      role: controller }
  - { id: post,  kind: tool,  tool: erp.post_invoice }
# caps on steps, spend and time, set by the owner

Who can start it: The AP team
Connectors: ERP and the AP inbox, as the space allows
Credentials: Injected at run time, never shown
Tests: Sample invoices, then the living benchmark

Build and control

Two people for every change

And a person for every action that sends, spends or publishes

Every change

01
On a branch
A workflow, app, skill, agent or fix
02
Tested
Sample data, then the living benchmark
03
The owner approves
With a fresh sign-in
04
A person releases
One-step rollback
05
Live on its trigger
06
The console watches
Runs, failures, spend, savings

Agents propose and draft.
They never approve or release anything.

Autofix and Autotune
propose a fix on a branch

Every action

Approval neededInvoice approvals · step 5 · example

Post this invoice to the ERP?

Above the controller’s limit, so it waits for you.

Supplier: Northgate Facilities Ltd
Amount: £18,400.00 · matches its purchase order
Limit: Approval limit, Expenses policy p. 4
Acts as: You, through one short-lived token

Approve and sign inDecline

Exact arguments. Values from document text are marked; the approval matches them and is used once.
Standing approvals only as owner-signed rules with limits on amount, recipients and volume, logged on every run.

Build and control

Who can do what

Roles come from the client’s own groups;
agents propose, people decide

An open key cabinet and a sign-in ledger under a lamp

The client’s identity groups

Finance-UK
AP-Team
Controllers
Finance-Builders

Bound by group id,
never as a list of people

Roles in the Finance department

Member
Builder
Approver
Releaser
Owner

Plus, install-wide, the install owner in the client’s IT.

	Member	Builder	Approver	Releaser	Owner	Install owner	Internet Labs	Agents
Use the OS, run allowed workflows
Build workflows, apps, agents
Approve a change to go live
Release, or roll back
Approve a send, spend or publish of their own
Connect a system, bind groups
Read a department’s sources
Set AI budgets

Decides, with a fresh sign-in
Does it
Proposes or drafts
Only within limits: own rights, a time-boxed setup grant, or as the person
Never

Build and control

IT and security see everything

Every workflow, app and agent:
who built it, what it touches, what it costs

Operations consoleInside the client’s environmentExample data

Inventory
Monitoring
AI spend
Access map
Duplicates
Savings ledger
Dependencies
AI gatewayv2

Inventory

Every workflow, app and agent across the company

All spaces
Finance
Sales
HR
Legal

Name	Kind	Space	Built by	Touches	Last run	State
Invoice approvals	Workflow	Finance	A Finance builder	ERP · AP inbox	2 min ago	Live
Finance expert	Agent	Finance	Preset	Finance vault	1 min ago	Live
Expense checks	Workflow	Finance	Preset	Expense policy · Jira	14 min ago	Live
Month-end checklist	Workflow	Finance	Preset	ERP	Autofix proposed a fix	Awaiting owner
Discount approval routing	Workflow	Sales	Preset	CRM	38 min ago	Live
Onboarding checklist	Workflow	HR	Preset	Mail	Yesterday	Live
Contract intake	App	Legal	A Legal builder	SharePoint	Not yet	On a branch

Build and control

Inside the client’s walls

Designed for a Fortune 500 security review

Approved model providers
Only the ones the client approves
Model calls
Valor API
Zero retention, if the client chooses it
Requests
Internet Labs
Holds no client data; no shared control plane
No line in

The client’s environment

Its cloud account · on-prem · air-gapped with the Sovereign edition

Memory vaults
One per department, plain files under Git
Workflows and apps
Code in Git, shown as a canvas
Logs and audit trail
Every write, with one-step rollback
Keys and the twin mapping
In the client’s key store
Adapters and local models
Regulated data reaches only these
The operations console
IT and security see everything built and run

Permissions before content
A source on every claim
Two human approvals
Every step sandboxed
Credentials never visible

Certifications

01
SOC 2 Type I
before the first paid Proof
02
SOC 2 Type II and ISO 27001
within 12 months of it
03
Compliance packs on request
HIPAA, ISO 42001, FedRAMP Moderate

The frontier stack

Field

Forward-deployed engineers, and models clients own outright

01 Internet SI OS
02 Field
03 Research
04 Valor
05 Infrastructure
06 Protocol and platform

01 Deployment pods
Two forward-deployed engineers, a data and platform engineer, a delivery lead.
02 Field models
$8M+ to build one, $2.5M a year to retrain and support.

396
open forward-deployed roles at 13 companies
70%
of enterprises expected to abandon agentic AI built by vendor forward-deployed teams by 2028

The frontier stack

Research

The lab that makes enterprise intelligence learn

01Continual memory under approval
An adapter ships only when its benchmark holds everywhere.
02Private training
Reversible semantic pseudonymisation before client data trains a model.
03Routing and evaluation
Routers refit on each client’s outcomes.
04Valor
Post-trained on compliant open bases first.

01Internet SI OS
02Field
03Research
04Valor
05Infrastructure
06Protocol and platform

−89% · −71%
held-out F1 after full fine-tuning and after LoRA: why weights come last
85%+
lower cost at 95% of GPT-4 quality from routing

The frontier stack

Valor

The frontier model line, trained where the knowledge lives

Valor is post-trained on golden data inside each client’s walls: training data frontier labs never train on.

01Internet SI OS
02Field
03Research
04Valor
05Infrastructure
06Protocol and platform

Three ways to buy

An API from Internet Labs’ regions
Dedicated in-country capacity
A licence on the client’s own GPUs

The pretraining gate

Demand inside the OS pays for the serving fleet
Valor beats the best permitted frontier model

$20M$400M+
Mistral’s annual revenue run-rate in the year to February 2026
$62M$240M
Cohere’s ARR, end-2024 to end-2025
$1B+
a frontier training run by 2027

The frontier stack

Infrastructure

Owned compute under every layer

01Internet SI OS
02Field
03Research
04Valor
05Infrastructure
06Protocol and platform

After Gate 1
B300 and H100 clusters train Valor, sized to a million-token context.
Years 3–5
The serving fleet grows with Valor’s demand.
By year ten
722,000 GPUs serving, 100,000+ training Valor, 65,000 rented out (derived).

$31.5K
a B200 earns a year at $4.50 an hour and 80% utilisation (derived)
60%
CoreWeave’s adjusted EBITDA margin, 13% after depreciation

The frontier stack

Protocol and platform

The standard for how enterprise agents exchange knowledge

01Internet SI OS
02Field
03Research
04Valor
05Infrastructure
06Protocol and platform

The Emissary protocol
An open standard for permissioned, identity-carrying exchange of derived answers.
The application platform
Agents that build the company’s own tools.

11
protocol-level flaws found in A2A, including identity lost across delegation hops
Up to $234B
of enterprise application spend exposed to agentic arbitrage through 2030

The frontier stack

It compounds

Every installation that opts in makes Valor stronger

01
Daily work
in the OS
02
Golden data
demonstrations, preferences and tests, approved by owners
03
Valor
Internet Labs’ model line, post-trained inside each client’s walls
04The five-year aim
The default model
for daily corporate work

Frontier models stay in the mix for the hardest tasks; pragmatic routing decides, request by request.

The frontier stack

The five-year aim

Valor, the default model for running a company

What Claude became for code,Valor becomes for corporate work

After 1,000+ installations, Valor learns from the golden data of those that opt in.

The data
Trillions of tokens, out of reach of every frontier lab.
The fit
Trained on real corporate work in every department.
The cost
Much daily corporate work needs less than a frontier model’s compute.

Pragmatic routing decides which,
request by request

Frontier models: the hardest problems
Valor: the daily work of every company

An office building after dark, one window lit, a lamp on the desk inside

The business

One installed base

Every business selling into it

Business	SAM a year	Year 5	Year 10
Internet SI OS	$556B	$4.6B	$60.7B
Valor	$139B	$1.5B	$26.2B
Infrastructure	$174B	$158M	$2.0B
Field	$10B	$52M	$262M
Programmes, self-serve, partners	$82B	$610M	$5.8B
Consumer edition	$52B	$180M	$4.5B
All businesses	$1.01T	$7.1B	$99.5B

Revenue is the plan, derived. Not a forecast.

The business

How a deployment lands

A paid Proof first, a training month last

10–12 months · value from the first department

Discovery
Shown in person, under NDA
The Proof
Paid, fixed scope, one department; credited against setup
Waves
One department at a time, each live before the next
Handover
About a month training the client’s team
Every year after
License and subscription; departments added at their setup price

Inside the Proof · one department, paid, fixed scope, credited against setup

01Its compiled base
02Its expert agent
03Its first workflows
04A side-by-side test against the current tool
05A baseline for the savings ledger

Thresholds fixed in writing before the test runs; every deployment reports its lift over plain retrieval with the same model.

A pod of four delivers it

Forward-deployed engineer
Forward-deployed engineer
Data and platform engineer
Delivery lead

About one deployment per pod a year.

A ring of old keys beside a closed folder on a wooden table — Handover closes only when the client’s team does every job unaided.

The business

Priced off one number

The setup fee: from $5M, then 70% of it every year

Setup, once

$5M $10M

From $5M for the first ten deployments once P&L case studies exist

Every year after

70% of setup

$3.5M a year on a $5M setup

20% license
50% platform subscription

Setup per department, by edition

Core $300K
The shell, Knowledge Compiler, expert agents, routing, memory vault
Enterprise $450K
Core plus the Emissary, the build loop, adapters, governance and audit
Sovereign $675K
Enterprise plus on-prem or air-gapped, data residency, compliance packs

On top, as it is used

Seats: About $25 a seat a month, realised
Agent actions: $0.07 above the included volume
Voice: $0.12 a minute
Desk resolutions: $1 per request an IT, HR or finance desk resolves
Third-party models: List price plus 5%
Valor: What it costs to serve, plus a margin
Adapter training: Compute at cost

Seats are a planning assumption. Support, compliance, residency and services are priced as add-ons.

The business

A $304B market for the OS

Every company with $1B+ of revenue, at our price

Revenue	Buyers	Package a year	Market a year
$10B+	2,000	$70M	$140B
$5–10B	2,000	$22M	$44B
$1–5B	13,750	$8.7M	$120B
All tiers	17,750		$304B

Price

Setup $5M+, then 70% of setup a year

First vertical

Financial services: regulated, document-heavy, and where the benchmark ran

The business

TAM, SAM and SOM

A $10.3T market, a $1.01T one we can serve

TAM
$10.3T
in 2030
Worldwide IT spending, plus the outsourced work agents compete for
SAM
$1.01T
a year
The OS and its editions, Valor, compute and Field
SOM
$7.1B
year-5 revenue
Year-5 revenue on the plan, every business counted

Not to scale

The business

How the SAM is built

Buyers × price, tier by tier

		Buyers	Package a year
Internet SI OS$556B	$10B+ revenue	2,000	$70M	$140B
	$5–10B	2,000	$22M	$44B
	$1–5B	13,750	$8.7M	$120B
	$500M–1B	12,000	$3.5M	$42B
	$100–500M	60,000	$3.5M	$210B
Programmes, self-serve, partners$82B	National and sovereign programmes	100	$100M	$10B
Programmes, self-serve, partners$82B	Per-seat edition	200M seats	$30 a seat a month	$72B
Other businesses	Consumer edition			$52B
	Valor			$139B
	Infrastructure			$174B
	Field			$10B

SAM $1.01Ta year

Every company with $100M+ of revenue is a buyer: setup is $5M+, then 70% of setup a year.

The business

Ten years

One plan, every business

The plan, derived bottom-up. Not a forecast.

2027–2036	Year 1	Year 2	Year 3	Year 4	Year 5	Year 6	Year 7	Year 8	Year 9	Year 10
Revenue	The build	$93M	$635M	$2.4B	$7.1B	$15.8B	$29.0B	$47.3B	$70.8B	$99.5B

Year 10

$99.5B

Internet SI OS$60.7B
Valor$26.2B
Infrastructure$2.0B
Field$262M
Programmes, self-serve, partners$5.8B
Consumer edition$4.5B

The business

One enterprise

And what the whole company takes

One enterprise of 50,000 people: what it pays a year

Year 1
$21.3M
Year 2
$28.6M
Year 3
$41.4M
Year 4interpolated
$54.2M
Year 5
$67.0M

~59%
blended gross margin, every business
$69B
of capital to year ten, most of it GPU-backed

Why we win

Built to win

And why an early lead compounds

A compiled, cited foundation
Retrieval is table stakes; the compiled vault is not.
The Emissary
Departments share intelligence without sharing their files.
Sovereign, on no single vendor
Any cloud, on-prem or air-gapped.
Evolvable after handover
The improvement loop keeps running after the engineers leave.
Every deployment advances the base
The next one starts further along.
Daily use is the switching cost
Leaving means moving the entire company.

Buyers compare us with Palantir AIP, Glean, Cohere North, Microsoft Copilot and the labs’ own deployment arms.

Why we win

Where the frontier labs meet us

Everything we build survives

Not their market, not their intent, and too small to chase while they race for fabs, power and compute.

What Internet Labs builds	If a lab moves in	Why it stays ours
Knowledge compiled inside the client’s walls	Survives	The labs’ deployment arms install their own models
The approval loop and living benchmarks	Survives	Each client governs what its own model learns
Air-gapped and sovereign deployment	Survives	Installs on the client’s Kubernetes with any permitted model
Provenance free of restricted models	Survives	DHS barred a foreign model and products using it from federal civilian agencies
The Emissary	Survives	Too small to move a lab’s revenue
Valor	Survives	Post-trained on each client’s golden data inside its walls

The labs’ game

$500B
OpenAI’s Stargate: 10 GW pledged by 2029, contracted by Apr 2026
$100B+
Anthropic’s ten-year compute deal with AWS, up to 5 GW (Apr 2026)
1 TW
of compute a year: Terafab, the fab of Tesla, SpaceX and xAI (Mar 2026)

Why we win

Designed for trust

Inside the client’s walls, on no single vendor

Sovereign by design: Client data leaves only in calls to approved model providers.
Learning under governance: Pseudonymised in, approved, benchmarked, versioned, reversible.
Agents that ask first: A person signs off anything sent, spent or published.

Compliance

SOC 2 Type I — before the first Proof
SOC 2 Type II and ISO 27001 — within 12 months of it

The plan

Twelve months to build

Three teams in parallel, then one system

01OS core
02Continual learning
03RSP

Month 3
Compiler and vault, shown on a real department’s corpus
Month 9
Three financial-services design partners; lift over plain retrieval measured

Integrate and test

Month 13First paid Proof
Month 18Gate 1
Proof targets met and a second department signed
On Gate 1The next round
Sized to the evidence Gate 1 produces

Internet SI OS is pre-revenue. Product brief, technical design and MVP plan complete; this round builds the OS.

The plan

Twenty-three features, three gated phases

Each client moves up only after passing its gate

13v1 · Core
- Knowledge Compiler
- OS shell
- Company knowledge base
- Hard-copy digitisation
- Department expert agents
- Access that travels with the knowledge
- Multi-model routing
- Memory vault and living benchmarks
- Workflow studio
- Secure runtime
- Operations console
- Build spaces and skills
- IT and Security presets
Gate 1
7v2 · Connected
Gate 1 opens v2: the Proof’s targets met
and a second department signed.
- The Emissary
- Agents that suggest skills
- Graph engineering
- A shared agent space
- Apps and cases
- Autofix and Autotune
- AI gateway
Gate 2
3v3 · Self-building
Gate 2 opens v3: no permission breach,
and every living benchmark holding.
- Self-improving build loop
- Department model adapters
- Partner and MSSP tenancy

The plan

The founding team

Abraham Vincent Philip
Founder and CEO
Design engineer; founder of Internet Labs. Has worked for Netflix, Meta, Microsoft, Airbnb, Uber, Canal+ and Chanel, and on Stranger Things. Leads product, design and engineering, and runs the company.
Vipul Sharma
Founding CTO
Previously Meta.

Hiring rule

Staff level and above only — fourteen people by month six.

First hires

Founding ML research lead
Enterprise sales lead, financial services
Forward-deployed engineer

The plan

What this round funds

Raising

$20M

Eighteen months to Gate 1, plus three months of breathing room.

Instrument and terms open

Use of funds, 18 months
People		$7.6M
Hiring		$0.7M
Compute		$2.1M
Go-to-market		$2.4M
Company		$2.2M
Contingency, 15%		$2.2M
Breathing room		$2.7M
Total		$19.9M

Reaches

OS built
SOC 2 Type I
First paid Proof
First deployment live
Gate 1

The plan

How the capital is staged

Each round is sized to the evidence the one before it produces.

Pre-seed · now
$20M
18 months: Internet SI OS to Gate 1.
Seed · on Gate 1
$250M
Valor v1, the first serving fleet and more pods.
Series A
$2B
The fleet, in equity and GPU-backed debt.

The plan

The exit

A public listing as $INT

The plan reaches $15.8B of revenue in year 6, 2032.

$INT

2027the build
2028$93M
2029$635M
2030$2.4B
2031$7.1B
2032$15.8B

$5B
$10B

Now: $20M pre-seed
Month 18 · Gate 1: The next round
Years 3–5: growth equity and GPU-backed debt
Year 6: IPO

Capital for the fleet
Public equity alongside GPU-backed debt.
Trust for regulated buyers
Audited public reporting.
Liquidity and currency
Liquidity for every investor and employee; shares as currency.

$965B
Anthropic’s Series H valuation, days before its IPO filing
$1.5B
raised in CoreWeave’s Nasdaq IPO
$477B
Palantir’s market value, about 77× trailing revenue

Frontier superintelligence for enterprises

Built where the knowledge lives

internetlabs.si · hi@internetlabs.ai

Internet Labs · Pre-seed

Thank you

We’re raising $20M to build Internet SI OS and reach Gate 1.

Fund this idea

hi@internetlabs.ai

Follow the build

Appendix

Sources

Full citations sit in each slide’s notes; the plan’s derivations in the market-sizing notes (26 Oct 2026)

Our documents: Founder’s plan (24–25 Oct 2026) · Market Research (2 Oct 2026) · Product Brief (2 Oct 2026) · Continual learning (2 Oct 2026) · MVP plan (10 Oct 2026) · Technical Design (16 Oct 2026) · Founder’s designs (Oct 2026): the improvement engine, pragmatic model routing, reversible semantic pseudonymisation · Red-team memos (24 Oct 2026) · Internal pseudonymisation benchmark (500,000 financial-services test cases)
Our models and team: The plan (26 Oct 2026) · founder’s aim and assessment (26 Oct 2026) · Internet SI OS investor deck (25 Oct 2026): derivations, pricing ladder, price ramp and margin profile, engine model, bottom-up budget · salaries set by the founders (25 Oct 2026) · 2026 list prices · team backgrounds (24 Oct 2026): internetlabs.ai, staciechan.com, public profiles
Market sizing: Gartner, AI-ready data (26 Feb 2025) · Gartner (2026): 3Q26 Market Databook, sovereign cloud IaaS (9 Feb), agentic spend (1 Jul, 2 Sep), CIO planning (31 Aug), AI forecast (16 Sep), agentic abandonment (29 Sep), AI-optimised IaaS (14 Oct), IT spending (21 Oct), 2030 forecasts for models and consumer agentic AI · Software Strategies Blog (Aug, Sep 2026) · Grand View Research, BPO (2026) · BCG Applied AI Index (Sep 2026)
Enterprise adoption: KPMG AI Pulse (Sep 2026) · McKinsey (Aug 2026) · S&P Global (Jan 2025) · Deloitte (State of AI 2026, Jan 2026; Aug 2026) · EMA (Aug 2026) · a16z CIO survey (Jan 2026) · Menlo Ventures enterprise surveys (2025) · Rung (26 Oct 2026) · Fortune (26 Sep 2026) · TechCrunch (Sep 2026)
AI companies: Mistral via FT (Feb 2026), Reuters (Sep 2026), Series D (8 Sep 2026) · Cohere memo via CNBC (Feb 2026), Sacra, Reuters (May 2025) · OpenAI (2023; 22 Sep 2026), Stargate (Jan 2025; Apr 2026), Deployment Company via Reuters (May 2026), The Information (Jul 2025) · Anthropic (29 Sep 2025), AWS deal (Apr 2026), Series H (28 May 2026), IPO filing (1 Jun 2026), via FT (Oct 2026) · Terafab coverage (Mar 2026) · Google Cloud (Aug 2025)
Compute and comparables: Palantir 10-Q and Q2 2026 call · The Motley Fool (25 Oct 2026) · CoreWeave IPO pricing (27 Mar 2025), FY2025, Q2 2026 outlook (11 Aug 2026) · Epoch AI (Jun 2024; Feb 2026) · server listings via gpuperhour.com (Sep 2026), Lambda, getdeploying.com (Oct 2026) · ECA International · Copilot, Gemini and ChatGPT list prices
Research and protocols: Meta, sparse memory finetuning (Oct 2025) · FrugalGPT (TMLR 2024) · RouteLLM (LMSYS, Jul 2024; ICLR 2025) · A2ABreak (9 Sep 2026) · MCP adoption (Dec 2025, Jul 2026) · A2A adoption (Apr 2026)
Security and law: OWASP agentic threats (Feb 2025; v1.1, Dec 2025) · FDA PCCP guidance (Dec 2024; Aug 2025) · EU AI Act Art. 43(4) · CJEU EDPS v SRB (Sep 2025) · EDPB Guidelines 01/2025 · DHS Order 2026-01 (24 Jul 2026; Federal Register, 20 Oct 2026)

Internet Labs

Frontier models have read the internet

The knowledge

The system

The model

Contents

Enterprise AI spending is up a third on a year

Spending rises; results don’t

Knowledge

Trust

Ungoverned building

After the vendor leaves

Why now

Internet Labs

Internet Labs

Internet SI OS

Compiledmemory

Expertagents

The Emissary andthe Orchestrator

Continuallearning

One system for the whole company

Memoryvault

Pragmaticmodel routing

Reversible semanticpseudonymisation

Continuallearning

How it works

Daily work read bottom to top

Improvement read top to bottom

One install, every department

HR

Finance

Sales

Marketing

Design

Engineering

Legal

IT

Security

Any other department

The Knowledge Compiler

Expense approvals

The memory vault

Knowledge Compiler

The vault

Agents

Owner approves

Why compiled beats retrieved

The Emissary

A person

The Emissary

The other department’s agent

Pragmatic routing

Pragmatic model routing

Orchestrator

Reversible semantic pseudonymisation

Tested on 500,000 financial-services cases

The improvement engine

Exhaust

Gap discovery

Evals

Interventions, and the owner approves

Offline replay and verdict

Release and measure

System knowledge

Memory first, weights last

Memory

Living benchmarks

Department adapters

Routing

Continual learning

Memory

Weights

From tribal data to golden data

Build a workflow in plain words

Two people for every change

Who can do what

IT and security see everything

Inventory

AI spend · this month

AI gateway

Compiled
memory

Expert
agents

The Emissary and
the Orchestrator

Continual
learning

Memory
vault

Pragmatic
model routing

Reversible semantic
pseudonymisation

Continual
learning