● ENGINE LIVESign inStart a free shadow runPricing
Free to prove on your own workload. After that, a share of a saving you have already watched us produce, never a subscription.
Modular and priced as a share of the savings once live. Enterprise when it must run in your building, and hyperscaling across your whole fleet. The few flat lines are listed below, with their prices.
A worked example
With the cost modules alone we save you $84.54 of it. Our share is 35% of that saving, $29.59. Your total is $45.05 instead of $100.00, and $54.95 of the saving stays with you.
Take the performance modules too and the same $100.00 also buys $119.78 of extra capacity, valued at your own rate. That is $204.32 of value in all; our share is 25% of it, $51.08. A lower rate on a wider base: you pay the smaller percentage, we earn the larger amount, and nobody has to lose.
The rates are 35% of the measured saving, or 25% of the saving plus the capacity it frees. The saving in this example is the composed benchmark figure; yours comes from replaying your own workload in shadow, free, before anything is billed.
Built for what you do next
A smaller bill is the start, and it has a limit: you can only save what you spend today. What you do with the headroom has no ceiling. More work on the machines you already own, results that arrive in time, and things that were not possible before.
We can't know what you will build with that, and we don't need to. That is why the 25% rate covers the capacity you gain as well as the money you save. Whatever you do next, it runs on WUBBERY, and we grow only when you do.
Free · Modular · Enterprise · Fleet
START
$0
Prove the saving on your own workload, with nothing to risk.
SELF-SERVE
From $50
Per module, per month. For teams spending under about $250k a year on compute, where measuring the saving would cost more than the saving is worth. No baseline study, no true-up, no salesperson.
GROW
Keep 65–75%
Go live and keep 65% of every dollar we save you. We take 35% of the measured saving — nothing else. Add the performance modules and the rate falls to 25%, charged on the saving plus the capacity it frees.
Most popular
Keep 65–75%
The same rates, running inside your own network. We never store your prompts or responses, and the engine calls home only if you turn reporting on.
FLEET
Keep 65–75%
The same two rates, one engine across every rack in the estate.
No saving, no bill
01
Connect WUBBERY in shadow and it measures what it would have saved, and the capacity it would have freed, on your real workload. Your systems keep serving exactly as they do now. You see the number before anyone talks about price. Free.
02
Go live and WUBBERY takes 35% of what it measurably saves you; you keep the other 65%. Add the performance modules and the rate FALLS to 25%, charged on the saving plus the capacity it frees: a lower rate on a wider base. It works the same on a cloud bill, a data-centre power bill or a fleet you own; the worked example above shows the arithmetic, and your own figures come from your own shadow run. The saving is never a subscription and never per-seat: if it does not save you money, there is nothing to share. The only flat lines are for things no ledger can measure (a gaming console, extra stations, admin seats), and each is listed with its price.
03
The engine is modular. Start with the one result that moves your bill and add the others as they prove out, each scoped to your deployment. You never buy a combined OS to get one capability.
Only with WUBBERY
WUBBERY is a performance engine, not a discount on a bill: the same work that cuts cost also makes the output faster, sharper and more coherent. One of these ships free, because the fastest way to prove it is to let you feel it.
Included free
Your users feel the wait drop within the first 300 requests, about 0.45 ms of traffic, and the results stay exactly the same. We give this one away: once your systems are visibly faster, going back is not an option.
Add-on
Your product arrives already knowing your customer. Better results sooner, and an experience that feels like it remembers them.
Add-on
The system gets sharper the more it runs on your workload, because the work it has seen before stops costing what it cost the first time.
Add-on
Voice, avatar and copy stay coherent with each other and with the work behind them — an entire quality dimension nothing bolted on can reach.
The meter changes, not the deal
When the limit is your own hardware rather than an invoice, the saving stops being avoided bills and becomes freed machinery, and the engine measures that just as directly.
Capacity
The measured 54.5% composed compute saving, read the other way: the same cluster does ×2.20 the work per unit of modelled compute. For a self-hoster the constraint is the cluster, not an invoice. This is the difference between buying the next rack this year or next. The dedicated capacity stack measures further still: ×6.64 the work, 84.95% of the accelerator work avoided, with zero priority violations. A different bench with different levers: the two multiples are never added, to each other or to the cost figure.
Your numbers, our ratios
There is no API price table in this arithmetic. You know what a unit of your compute costs you: hardware, energy, operations. We bring the measured ratios; you bring your own cost basis; freed compute times your number is the value, and the bill is 25% of it. Both halves of the multiplication are checkable: yours because it is yours, ours because every ratio reproduces from a seeded command.
Energy & memory
Energy is a bill you pay directly on hardware you run, and the composed energy estimate is taken across everything we ship, on one population, quoted with its assumptions and with the hostile worst case beside it when you ask for one. The smaller memory pool and the reachable-compute gain land the same way: hardware you did not have to buy.
The principle
The same module is worth hundreds a month to one operation and hundreds of thousands to another. So modules set what you get, and the measured outcome sets what you pay. On a bill you pay someone else, we take a share of what we removed; on your own metal, 25% of the value of the machines you didn’t have to buy, measured the same way and priced from your cost basis, never ours.
Questions
35% of the cost WUBBERY measurably removes, on whatever you run: cloud, data centre, devices or AI. You keep 65%. Take the performance modules as well and the rate drops to 25%, charged on that saving plus the capacity it frees. You see the saving in shadow before going live, so the price is always a fraction of a number you have already verified on your own workload.
Because it aligns us with you. A subscription charges the same whether or not it works. A share of savings means WUBBERY only earns when you are already ahead — and it keeps our incentive pointed at cutting your cost, not growing a seat count.
Every organisation that runs on compute: data centres and cloud fleets, databases and data platforms, games and real-time 3D, edge and robotics, regulated and sovereign systems, and AI in production. If compute is a cost or a constraint, there is a saving or capacity to share. Start in shadow and see it.
They go beyond cost. They lift the quality and speed of the output itself, and the console shows each one off and then on, on your workload. They are how a deployment gets better, not just cheaper.
Enterprise is about CONTROL — inside your own network, nothing leaving the building. Hyperscaling is about SCALE — one engine and one bill across a fleet of racks. Many customers want both, and they compose.
Three things, because a ledger cannot measure them: a gaming console at $10 a month each, stations beyond 3 free at $5 each, and an admin or billing seat at $15 (plain members are free). Station telemetry is kept for months for every account and is never charged for. They appear as their own lines on the same monthly invoice as the share, so there is one bill and every line says what it is for. The saving itself is never flat.
Create an account and run a shadow on your own workload, or talk to us and we will scope it with you first. Either way you see the saving before you commit to sharing any of it.
Run a shadow on your own workload and watch what it gives back, with your systems still serving exactly as they do and nothing to risk in finding out. Or talk to us and we will scope it with you.