Subvectio.AI · Powered by ACU Markets

Price the compute you build. Control the AI you run. Govern the dollars you spend.

Subvectio turns declared inputs into golden-value-tested economics across three control points: data-centre economic analysis, AI agent workload control, and enterprise FinOps governance. One deterministic engine is the source of truth for every money number — GenAI drafts and explains, it never computes. Benchmarked to ACU Markets.

Request a working session See the deterministic engine
1 Build

Data Centre Economic Analysis

Price a site or fleet against global benchmark compute pricing, energy, carbon and fleet efficiency — before capital commits.

For DC financiers, developers, infra investors, CIOs
Explore Vector 1 → 2 Run

AI Agent Workload Control

Route every workload to the most proficient agent on accuracy, economics and speed via Vector AI.

For AI/ML and platform leads
Explore Vector 2 → 3 Govern

Enterprise FinOps Control Tower

Meter spend and utilisation against the engine's expectations. Catch overspend, leakage and bill-shock before they go structural.

For CFOs, CPOs, FinOps
Explore Vector 3 →
The spine

One engine. One source of truth for money.

Every figure across all three vectors is traceable through the same geometry: declared input vectors, one deterministic simulation engine, golden-value-tested outputs, and an independent ACU Benchmark reference. The vectors differ; the spine does not.

01 — DECLARED INPUTS

Explicit unit contracts. Nothing silent.

Every figure across all three vectors starts as declared input vectors under explicit unit contracts — no hidden assumptions, no silent defaults. Bills, contracts, telemetry and engineering prints enter as data, not folklore.

02 — ONE DETERMINISTIC ENGINE

Same inputs, same answer, every time.

Golden-value regression fixtures pin every calculation — across the wire and across releases. Racks, stacks, pathways, routing and spend all resolve through the same simulation core.

03 — VECTOR COPILOT, BOUNDED

GenAI drafts and explains. It never computes.

VECTOR Copilot classifies, extracts, explains and drafts inputs and scenarios from bills, contracts, telemetry and engineering prints — but never produces a financial figure. That rule is non-negotiable.

04 — ACU BENCHMARK REFERENCE

Independent reference points, not vendor claims.

Every economic case links to ACU Benchmarks — the AI compute benchmark — so pricing, sensitivity and negotiation positions stand on an independent reference point.

Vector 1Build · Analyse

Price the compute before you pour the concrete.

For data-centre financiers, developers, infrastructure investors and CIOs.

Turn engineering prints into an investment-grade economic case — benchmarked against global compute pricing, energy, carbon and fleet efficiency — before capital commits.

Global benchmark compute pricing

Every site and fleet case is priced against ACU Benchmarks and observed rental-market surfaces — hyperscaler, colocation, owned and marketplace pathways side by side, on identical terms. Output: $/PFH-style unit economics and a negotiation corridor.

Energy

Power, PUE, energy price and firming enter as declared vectors. The engine computes energy-adjusted economics — useful output per watt, cost of power as a share of delivered compute, PUE and cooling sensitivity, MW forecasts, utilisation break-even.

Carbon

Carbon intensity is a first-class diagnostic, aligned to the ACCARB carbon-index family — carbon per unit of useful output, grid-mix and firming sensitivity — so the case survives ESG and investment-committee scrutiny.

Fleet efficiency

Accelerator mix, GPU efficiency, useful output per accelerator-hour, power density and workload class feed the useful-output denominator — two sites with identical nameplate MW compared on delivered compute, not capacity.

Capital planning

Capacity, stack design and infrastructure capital simulated across pathways with full sensitivity — capex/opex trade-offs, deployment horizon, and a case a financier can take to an investment committee.

Inputs

Power, PUE, energy price, capex, utilisation, demand, fleet mix, firming, region, commissioning horizon, contracting model — declared under unit contracts, with LLM-assisted ingestion from DCIM and schematic-derived exports.

→
Engine

Deterministic simulation of racks, stacks and pathways; full economic case, sensitivities and negotiation corridor.

→
Outputs

Design analysis, sensitivity tables, benchmarked $/PFH-style unit economics, and an investment-grade case linked to ACU Benchmarks.

Proof: golden-value tested, traceable to declared vectors, ready for investment-grade scrutiny. Scope: Subvectio reads structured DCIM/schematic-derived exports and translates rack, power, cooling, accelerator, utilisation and commercial assumptions into useful-output economics — it is not a DCIM replacement, a CAD/BIM editor, or an arbitrary schematic parser.

Vector 2Run · Control

Every workload, to its most proficient agent.

For AI/ML leads, platform engineering and product owners running agentic or model workloads.

Maximise AI capability per dollar — Vector AI routes every workload to the most proficient agent on accuracy, economics and speed, deterministically.

Deterministic proficiency routing

Vector AI dispatches each workload to the most capable named or open-source agent on three axes — accuracy, economics and speed. Routing decisions are computed, not guessed.

Model–task fit

Match the task class to the model that clears it best per dollar — avoiding over-provisioned frontier-model spend on tasks a smaller or open-source model handles at a fraction of the cost.

Agentic cost expansion, controlled

Agent chains multiply calls and cost silently. Vector AI makes the cost of an agentic workload explicit and routable before it expands into a structural line item.

RBAC-governed front door

Vector Agent is an RBAC-governed entry point for AI-agent workload control — user-facing agent orchestration stays separate from deterministic financial truth.

Deterministic ROI + bill-shock stress

ROI, model-task routing and bill-shock stress are deterministic calc-engine capabilities — exposed as API endpoints, agent tools and interactive UI. "Which agent, at what cost, for what return" is a computed answer.

Inputs

Workload definitions, task classes, model/agent catalogue (named + open-source), accuracy/latency/cost constraints, budget envelope.

→
Engine

Deterministic proficiency scoring and routing; ROI and bill-shock stress modules.

→
Outputs

Per-workload routing decision, expected cost and ROI, capability-per-dollar view, overspend early-warning.

Boundary: VECTOR Copilot supports ingestion, mapping, prompted insight and report generation. Deterministic engine modules remain the source of financial truth — routing economics are computed, never generated.

Vector 3Govern · Control tower

Govern every AI dollar before it becomes a structural cost.

For CFOs, CPOs, FinOps leads and procurement — AI spend intelligence and workload economics, not generic AI FinOps.

A control tower for AI spend — govern budget, utilisation and efficiency by metering reality against what the engine said it should cost.

AI governance

SSO/IAM into role-based access, versioned scenarios, a clean API surface and optional MCP/CLI — built for how enterprises actually govern platforms. Every AI dollar gets a decision basis — computed, never generated.

Spend management

Meter application and agent billing against the engine's expectations to catch overspend, leakage and bill-shock before they become structural line items. Budget exposure is tracked continuously, not reconciled after the invoice.

Utilisation management

Detect under-utilisation across models, agents and infrastructure; surface idle and leaking spend; tie utilisation to the useful-output denominator so "are we getting value from what we pay for" is a measured answer.

Efficiency gains

Model-fit, leakage elimination, right-sized routing (fed from Vector 2) and output-adjusted economics compound into measurable efficiency gains — quantified against the engine's baseline, benchmarked to ACU.

Inputs

Bills, billing exports, invoices, telemetry, contracts, budgets, EDPs, discounts, private pricing — LLM-ingested, deterministically normalised.

→
Engine

Metered-vs-expected reconciliation; overspend, under-utilisation, leakage and bill-shock detection; budget-exposure and output-adjusted economics.

→
Outputs

A live control tower — variance alerts, utilisation view, leakage register, efficiency-gain tracker, auditable decision basis per dollar.

Copilot boundary: the workbench FinOps copilot explains variances and drafts scenarios; the deterministic engine and metered data produce every figure.

The value chain

One deterministic engine. Three control points. The full compute lifecycle.

Not one vector through a geometry — three linked economic control points across the full delivered-compute lifecycle, reconnected on the same spine.

Build →

Subvectio's buy-side layer converts data-centre engineering prints into observed economics (Vector 1).

Bridge →

ACU provides benchmark-based treasury instruments for data-centre and compute exposure — the independent reference the economics link to.

Run →

Vector AI routes and prices every workload to its most proficient agent (Vector 2).

Govern →

The enterprise FinOps control tower meters spend and utilisation against the engine's expectations (Vector 3).

The same declared-input → deterministic-engine → traceable-output geometry runs end to end. A customer can price a proposed 300 MW site, route the workloads that will run on it, and govern the enterprise spend that consumes it — all against one benchmarked source of truth. That is the delivered-compute economic value chain.

Trust & governance

The credibility layer behind all three vectors.

◉ Deterministic source of truth

The engine is the source of truth for money numbers — GenAI never computes a financial figure.

◉ Golden-value regression

Fixtures pin every calculation: same inputs, same answer, every release, across the wire.

◉ ACU Benchmarks

Independent AI compute reference — powered by ACU Markets, not co-branded or endorsed.

◉ Explicit unit contracts

No silent defaults, no hidden assumptions — every input is declared and traceable.

◉ Enterprise governance surface

SSO/IAM, RBAC, versioned scenarios, a clean API, optional MCP/CLI.

◉ Deployment honesty

Restricted-network and true air-gapped modes are distinguished, not blurred — shipped vs. roadmap is always labelled.

Who it's for

Enter at your vector. Leave with the chain.

Vector 1 · Build

Data Center Financier

Developers, infrastructure investors, CIOs

Investment-grade site and fleet economics — benchmarked against pricing, energy, carbon and fleet efficiency — before capital commits.

Data Centre Economic Analysis →
Vector 2 · Run

Enterprise Effort Planner

AI/ML leads, platform engineering

Maximum AI capability per dollar via deterministic proficiency routing across named and open-source agents.

AI Agent Workload Control →
Vector 3 · Govern

Enterprise Budget Planner

CFOs, CPOs, FinOps leads

A control tower governing spend, utilisation and efficiency against the engine's expectations — before bill-shock goes structural.

Enterprise FinOps Control Tower →
Leadership

Built by the team behind ACU Markets.

Decades of technology procurement, enterprise software and corporate expertise — applied to the AI compute economy.

Brandon Combes

Brandon Combes

Founder & CEO

Technology procurement expert in cloud infrastructure sourcing and pricing; founded ACU Markets to fill the market gap for standardized AI compute benchmarks.

Gavin Wall

Gavin Wall

Chief Technology Officer

Gavin is an experienced technology and strategy executive with more than 20 years' leadership experience across cloud architecture, digital infrastructure, enterprise transformation and operational delivery. During nine years at Microsoft, he served as Azure Engineering Director for Australia and New Zealand, where he led regional strategy, capability development and investment across the Azure ecosystem and supported complex enterprise transformation initiatives.

As Chief Technology Officer of ACU Markets and Subvectio.AI, Gavin is responsible for the group's technology strategy and platform development. He leads the technical evolution of ACU Markets' independent AI compute benchmarking and market-data infrastructure, together with Subvectio.AI's enterprise compute-economics and decision-support platform. His experience in building secure, resilient and scalable technology environments is central to the group's objective of establishing trusted infrastructure for the emerging global AI compute economy.

Bryan Drennan

Bryan Drennan

Chief Operating Officer

Two decades transforming and scaling businesses; drives execution and operational excellence, building institutional-grade benchmarks for the AI compute economy.

Clifford Combes

Clifford Combes

Chief Customer & Revenue Officer

25+ years developing, marketing and selling technology, SaaS and infrastructure across Australia and New Zealand; former senior leadership at Oracle, Infor, TechnologyOne and Civica.

Garth Kerr

Garth Kerr

Chief Corporate Officer

Garth Kerr is a senior corporate executive with extensive international experience as a general counsel, director and strategic adviser to global listed technology companies and high-growth businesses. His expertise spans corporate and commercial strategy, governance, complex transactions, risk management, capital raising, institutional engagement and business scale-up.

As Director and Chief Corporate Officer, Garth supports both ACU Markets and Subvectio across their corporate, commercial, legal, risk and governance functions. He works closely with the CEO, Board and leadership team to advance strategic partnerships, investor engagement, capital initiatives and institutional readiness.

Bring your vector. We'll bring the engine.

A working session walks your numbers through the deterministic engine — site economics, workload routing or spend governance — against ACU Benchmark reference points.

Request a working session Explore ACU Benchmarks
Vector 1 · DC economics demo Vector 2 · Vector AI demo Vector 3 · FinOps demo