Pricing

Everything you need to write, run and test Sparsr code is free and always will be. You pay only for hardware.

  • Cloud FPGA

    Coming October 2026

    Running your kernels against real Sparsr hardware on AWS EC2 F2, billed by the minute. Write and test your kernel against the free emulator now, and the same kernel runs here when it opens — the instruction set is the same one.

    • FPGA acceleration of Sparsr
    • Per-minute billing
    • Scalable performance
    • Technical support

    Tell me when it opens

  • Custom work

    Quoted per project

    Instruction set extensions, memory and parallelisation work, and RTL built around your algorithm rather than around ours. This is where a workload that needs a primitive Sparsr does not have yet gets it — and, for a project that justifies it, where the design goes to silicon rather than staying on FPGA. It starts with a kernel study, so the quote rests on a measurement of your code rather than on an estimate.

    • Custom wide instructions at 4096 bits
    • Custom memory and compression formats
    • Custom parallelisation technologies
    • Algorithm optimisation on demand
    • Silicon manufacturing
    • Dedicated support
    • Flexible enterprise pricing

    Start with a kernel study

Why the free tier is genuinely free

Not a funnel step. The emulator exists because we need it ourselves, and the marginal cost of your using it is zero.

  • The SDK is always free

    Write your host application and your kernels on your own systems, at no cost, with no account and no usage cap. If you never touch Sparsr hardware, you never pay us anything.

  • Software emulation is always free

    Your kernel runs against a real implementation of the Sparsr instruction set on your own CPU. That makes local development and automated testing in CI possible without hardware access — which is the only honest way to let someone evaluate a processor they cannot yet rent.

Start with the free SDK

Everything on this page except the hardware is available right now, and the hardware is not what you need in order to find out whether Sparsr suits your workload.

Get the Sparsr SDK See what it fits