Idle GPUs
Datacenter and workstation GPUs sit unutilised for most of their lifetime. CapIX onboards that idle capacity as routed supply — and pays providers in USDC for it.
Investor overview · August 2026Try it live ↗
CapIX · Investor overview
The compute fabric
built on spare capacity.
CapIX aggregates idle GPUs, spare inference capacity, and quantum nodes into one secure, verifiable fabric. Because the supply is spare capacity, we supply compute 40–60% below hyperscaler pricing — and every unit settles with a signed, publicly queryable receipt.
63,000+ public receipts · 40–60% below hyperscalers · SWE-bench 45.4% stable
01 · The fabric
The world’s idle compute,
aggregated into one supply.
Spare inference capacity
Existing inference deployments run below capacity. Spare throughput is aggregated into the fabric and sold as cheap, metered compute.
Quantum nodes
Frontier compute joins the same fabric as it comes online — one aggregation layer across classical and quantum supply.
Spare capacity has no capex to amortise. That is why the fabric can price compute 40–60% below hyperscalers and still pay providers for capacity they were not selling at all.
02 · The operating system
Compute that
proves itself.
Idle GPUsSpare inferenceQuantum nodesProvider capacity
CapIX OSroute · verify
· receiptorchestration layer
· receiptorchestration layer
Routed computeVerified unitsSigned receiptsUSDC settlement
Smart routing
Every task is classified and scored across eligible lanes — price, fit, and harness memory — then placed on the cheapest supply that can do the work.
Verification
Spare compute is only useful if it is trustworthy. The OS verifies execution so untrusted supply can serve production workloads.
Signed receipts
Every unit of compute settles with a signed, publicly queryable receipt — 63,000+ and counting. The bill is evidence, not a claim.
The fabric supplies cheap compute. The OS is what makes spare compute trustworthy enough to sell.
03 · Proof
Live receipts.
Graded benchmarks.
Production activity + public methodology
August 2026
Benchmark honesty, on the record: two identical re-grades of the same patches resolved 106 and 103 — a ~3% environment-flakiness flip rate, disclosed in our public methodology. We quote the stable 103; the ever-resolved union is 110. The published ablation measures harness-vs-raw-model uplift at 2.75× (11/20 vs 4/20 on the identical set) — measured, not claimed.
04 · Apps on top
The first apps
built on the fabric.
These apps are proof the substrate works — production demand running on spare capacity, receipted end to end. They are the evidence, not the pitch.
chat.capix.network
The everything app: build websites, games, and apps — generate videos, images, music, and characters — plus the creator marketplace. One prompt box, one USDC balance.
Consumer surfaceCapIX Code
The agent harness on the fabric: openweight routing, escalation ladders, institutional memory, and graded benchmark discipline — 45.4% stable on SWE-bench Multilingual.
Developer surfaceConsole
Network operations surface: lanes, balances, receipts, and provider activity — the metered view of the fabric.
Ops surfaceIDE
Build environment wired directly into the fabric — route, run, and settle compute without leaving the editor.
Builder surfaceone fabricone balanceopenweight routingsigned receipt per unit
Every app task is routed fabric demand: it buys spare compute, earns a receipt, and makes the router smarter.
05 · Why we win
Cost structure, trust,
and a compounding router.
Cost structure
Hyperscalers price against capex. We price against spare capacity with no capex to amortise — a structural 40–60% cost advantage, not a promotional one.
Trust
63,000+ signed public receipts and a published benchmark methodology with disclosed flakiness. Our claims are queryable; competitors’ are not.
The flywheel
Every routed task teaches the router — placement, pricing, and harness memory compound. Solve rate rises while cost-per-solve falls.
CapIXfabric + OS + receipts
More tasks routedEvery build, fix, and generation
Richer routing memoryPer-repo lessons, pattern library, playbooks
Better placementHigher solve rate at lower cost-per-solve
Cheaper, proven unitsReceipts users and providers can verify
Spare-capacity cost structureReceipts as trust infrastructurePublic benchmark methodologyRouting data compounds
06 · Market
Compute spend is moving
to whoever prices it lowest.
We don’t need to invent a market. AI compute demand already exists and is growing; the open question is who supplies it cheapest with proof. CapIX consolidates four categories of existing spend onto one fabric.
AI inference
Every AI product is metered compute underneath. Today it is bought at hyperscaler prices with opaque billing.
Fabric supplyVibe coding
Prompt-to-app building. Proven paid demand at subscription price points; today sold without receipts.
CapIX Code · IDEContent generation
Short-form video, images, and music tooling — sold per-seat, single-lane, with no creator economy attached.
chat.capix.networkAgent labor
Autonomous coding agents command premium prices, closed and expensive. Our harness runs openweight at ~$8 per near-miss resolution.
Agent harnessBottom-up pricing: per-use lanes at $0.005–$0.50, subscriptions at $20 and $39, and a 5–8% marketplace fee on every creator sale — all settled as receipted fabric compute.
07 · Business model
Buy spare.
Sell metered. Keep the spread.
01Providers supply spare computeIdle GPUs and spare inference join the fabric and earn USDC for capacity they were not selling at all.
→02The OS meters every unitEach task is routed, verified, and settled at its lane price with a signed receipt on a real-money, idempotent ledger.
→03CapIX captures the spreadSpare-capacity supply costs less than the lane price. The difference is margin — plus subscription and marketplace revenue.
Compute margin — the spread
Lane prices hold while spare-supply costs fall as provider recruitment scales. Margin widens with supply.
Subscriptions — Pro $20 / Creator $39
Payable from balance or card. Autopilot channel subscriptions at ~$15/mo each.
Per-use + marketplace
Lanes at $0.005–$0.50 per unit. Marketplace fee of 5–8% — sellers keep 92–95% of every sale.
Both sides are paid in the open: providers earn USDC for idle compute, sellers keep 92–95%, and every charge carries a receipt.
08 · Traction
Live, metered,
and backed.
What is live today
August 2026
Fabric is settling
63,000+ signed receipts on production workloads, publicly queryable by anyone — including this deck’s live 24h counter.
Apps are shipping
chat.capix.network, CapIX Code, the console, and the IDE run on the fabric today — builds, video, images, music, characters, marketplace.
Backed
$250K pre-seed led by Atuin.Ventures (Peter Hopton).
09 · Go-to-market
Six campaigns,
one flywheel.
Provider recruitment
Grow idle and spare supply behind the OS so unit costs keep falling and margin widens.
Campaign 01Receipts Not Promises
Lead every channel with the public receipts feed — 63,000+ signed receipts anyone can query.
Campaign 02First 100 Deployments
Get the first hundred user-built sites, games, and apps live and screenshot-worthy.
Campaign 03Weekly Drops
A recurring marketplace drop cadence that gives creators a reason to ship and buyers a reason to return.
Campaign 0460-second cloud videos
Prompt-to-product in one minute, recorded end to end — the demo is the ad.
Campaign 05Marketplace seeding
Penny-priced packs, bounties, and NFT-backed ownership to bootstrap the two-sided economy.
Campaign 06Every campaign feeds the same loop: more supply lowers unit cost, more tasks sharpen the router, more receipts build trust.
10 · Team
Built for this
exact problem.
Ruqaiyah Iqbal brings enterprise cloud experience from Oracle, quantum-infrastructure DevOps experience from Oxford Quantum Circuits, and founder-level ownership across every layer of CapIX — the fabric, the OS, the harness, and the receipt ledger.
EnterpriseOracleCloud systemsFrontier computeOxford Quantum CircuitsDevOps infrastructureZero to oneCapIXFounder & builder
Cloud infrastructureDevOps & automationDistributed systemsAI orchestrationAgent harnessesPayments & ledgers
RI
Ruqaiyah Iqbal
Founder · CapIX
- Oracle · enterprise cloud
- Oxford Quantum Circuits · DevOps
- Queen Mary University of London
- 822 LinkedIn followers · 500+ connections
Founder-market fit: a career spent running other people’s compute infrastructure — enterprise cloud and frontier quantum — now applied to aggregating the world’s spare capacity.
11 · The ask
Scale the fabric.
Supply first.
CapIX is raising to scale provider supply behind the OS and distribution in front of it — on top of a fabric, product surface, and receipt ledger that are already live and settling.
More idle GPUs and spare inference behind the OS, so unit costs keep falling and the spread widens.
Keep pushing verification depth, cost-per-solve down, and the stable resolve rate up — publicly, with receipts.
Receipts Not Promises, First 100 Deployments, weekly Drops, 60-second cloud videos, marketplace seeding.
Round detailsShared in partner meeting
Let’s talk — try the product first ↗