Young innovators,
defined concretely.
“Talent is everywhere, compute is not” is the whole thesis. The program targets the people learning by building.
Students
High-school and university students learning AI by building it — coursework, theses, side quests.
Clubs & teams
Robotics clubs, hackathon crews, study groups. Apply as a group; the whole cohort gets capacity.
Classrooms & teachers
Educators running AI-assisted classes get compute plus the educator dashboard and evaluation traces.
Early independent builders
Young innovators outside any institution. If you are learning by building, you qualify to apply.
What “free”
means here.
We would rather under-promise than quietly throttle. This section is the contract.
Admitted-cohort eco-server hours
Real GPU capacity on the Eco Foundry engine: power-capped, carbon-scheduled, never billed.
Always-free browser runtime
WebGPU models run on your own device with no account — that tier is free for everyone, forever.
Not an unlimited GPU faucet
Cohorts are sized so capacity is real. We would rather admit fewer groups than throttle everyone.
Clear selection, honest answers
We review applications each term and email you either way — accepted or not.
The energy story
is the business model.
Free is possible because the engine optimizes learning per watt: power caps, carbon-aware scheduling, and smallest-capable-model routing. Here is what that would cost on demand — and what the levers save.
GPU energy via power caps
for +6.8% time (McDonald et al., MIT Lincoln Lab, 2022)
emissions cut via carbon-aware scheduling
next-day to weekend shifting, no missed deadlines (Wiesner et al., 2022)
training-energy savings range
batch size + power limit co-tuning (Zeus, USENIX NSDI ’23)
H100 on-demand, multi-provider average
up 18% YoY; AWS ~$6.88, GCP ~$11.06, Azure ~$12.29 (GetDeploying, 2026)
one 15-week class, at on-demand prices
30 students at 2 A100-hrs/week (illustrative)
the same class, accepted into a cohort
energy-aware routing is why the math works
Asked often,
answered plainly.
Is the compute really free?
Yes, for accepted cohorts. We admit classrooms, clubs, and student groups in waves so capacity is real; the in-browser playground is free for everyone with no account. We keep costs near zero by routing to the smallest capable model, power-capping GPUs, and scheduling around clean energy.
Who counts as a young innovator?
Students, student clubs and teams, classrooms with their teachers, and early independent builders. If you are learning by building, you qualify to apply — there is no institutional requirement.
What happens after I join the waitlist?
We review applications each term, in cohorts. You will hear from us either way. Accepted cohorts get eco-server hours, memory vaults, cohort sandboxes, and the educator dashboard where relevant.
What does “energy-aware” actually mean?
Every job is power-capped, shifted toward clean-grid windows when it can wait, and routed to the smallest capable model — browser first, edge next, eco servers last. The engine page cites every number.
Something else? Write to research@epochlearn.com.