Warppool is a new way to fund the open web. Instead of selling reader attention to advertisers, publishers embed one script tag and visitors lend idle GPU cycles to power real scientific computation. No surveillance, no data brokers, no middlemen taking 50%.
Warppool began in 2026 as a class project in the University of Virginia's Master of Science in Data Science program. The question started as technical — what if the GPU in every browser tab could do useful scientific work? — but quickly became economic: what if that compute could replace advertising as a way to fund the web?
Digital advertising is a $680 billion industry built on surveillance. Publishers get pennies on the dollar after the ad network takes its cut. Readers get tracked, slowed down, and manipulated. The bargain is bad for everyone except two or three companies at the top of the stack.
Warppool is building the alternative: a monetization layer where sites earn revenue from useful computation instead of from selling attention. Today we're a company operating on a B-Corp-style charter (certification in progress), with a mission to make the ad-free web economically viable.
The web doesn't need ads. It needs a better business model. Idle GPUs are that model.
Researchers run free. Publishers earn per GPU-millisecond contributed by their readers — no middleman, no revenue share to a platform. The same traffic that funds ads can fund science instead.
Ad networks track people without meaningful consent. Warppool asks visitors once, in plain English, and respects the answer. No visitor's GPU runs a single instruction without explicit opt-in. One click to stop, zero persistent state.
Ad networks take 50–70% of every dollar. Warppool takes nothing from publishers. Revenue is earned per completed GPU-millisecond, paid directly, with no platform tax and no opaque auction.
The compute that visitors contribute powers climate models, protein dynamics, ML inference — real scientific work. Researchers with institutional affiliations run free, no cap. The web funds discovery instead of data brokers.
The greenest data center is the one you never had to build. Every result Warppool returns is scientific compute that didn't require pouring concrete, manufacturing new servers, or evaporating cooling water — because it runs on hundreds of millions of GPUs that already exist, already powered on, already cooled by their own fans. A data center does a unit of work by adding new demand to the world. Warppool does the same unit on hardware the world already built.
≈ 0 liters
Personal devices are air-cooled by fans that are already spinning. A single hyperscale data center can evaporate millions of gallons of water a day; cloud compute consumes roughly 1.8–2 liters of water per kWh directly — and several times that once you count the water used to generate the electricity. Warppool adds none.
~1.0 PUE
Data centers spend an extra 10–50% of their electricity just on cooling and power conversion (a PUE of 1.1–1.5). Warppool's compute piggybacks on hardware that's already running and already cooled, so almost every watt goes to the actual science instead of to overhead.
None
A large share of computing's lifetime carbon is embodied — locked into manufacturing the chips, racks, and buildings before a single calculation runs. Warppool manufactures nothing. It reuses devices already made for everyday use, so the embodied carbon of its "fleet" is effectively zero.
A few watts
At idle priority a contributing device sips only the marginal watts above what it was already using — no new grid load for a new building, no diesel backup generators, no land cleared, no transformers, no fleet of trucks hauling water.