STEP 01
Choose your workload
Explore inference, fine-tuning and agent workloads. CORELOT aims to connect GPU capacity with Solana-native access and clear usage costs. This page is a product concept, not a live private-compute service.
Privacy-first AI.Built for Solana.
ExploreAI compute onSolana. A privacy-first direction forevery AI workload, with clear costs and user-controlled access.
view the compute activityIAD-4 · RACK 07 / SLED 03RENTED
86%rented, 7 days
--:--:--last signed, UTC
Pick a sled
55 GPUS IN THE COMPUTE2 DATA CENTERS
From your ideato private AI.
You choose the workload. CORELOT is exploring privacy-first GPU infrastructure with Solana-native coordination, so you stay in control of how your applications use compute.
How it works
STEP 02
Define your privacy
Choose what should stay private: prompts, files and model outputs. Encryption, execution isolation and data-retention controls are design goals that require implementation and verification.
STEP 03
Settle on Solana
The planned payment layer uses Solana and USDC for compute access. Published costs should describe resource usage without exposing workload content. Payments are not live yet.
Inside the compute
Openthe box.
Every workload needs a clear boundary. Inspect the hardware concept while privacy protections are being designed.
- SITE
- IAD-4, Ashburn VA
- CUSTODIAN
- Equinix
- HARDWARE
- 8× L40S 48GB, 2× 32-core CPU
- OWNERS
- Concept preview
- 01Workload isolationSeparate execution environments are a design requirement.
- 028× L40S 48GBGPU capacity shown here is illustrative.
- 03Metered at the boardResource accounting should not reveal prompts or outputs.
- 04Rack 07, IAD-4Illustrative infrastructure, not a verified deployment.
Distributed compute.Solana coordination.
A private workload should not mean opaque infrastructure. Public resource metrics, private application data, with Solana as the planned coordination layer, so visibility does not require revealing your inputs.
- Sites
- 2
- GPUs
- 64
- Ledger
- 1
Where the machines live
Drag to spin
Custody
The map illustrates a distributed GPU topology. Locations, operators and utilisation are illustrative data, not a live CORELOT network.
Scale your workloadyour way.
Choose a single GPU sled or explore pooled capacity. Different workloads need different resources, without changing the privacy-first direction.
Choose your compute
A distributed capacity profile for scaling AI applications. This interface does not route or execute real workloads.
Capacity scale
8 Sleds, 64 GPUs, 2 Data centers
Registry
Solana-native resource coordination is planned. No live compute registry or settlement is provided by the current interface.
Private workloads.Visible resource costs.
The architecture aims to separate resource accounting from workload content, so compute usage can be visible without publishing your inputs. Solana is the planned settlement layer. The activity below is illustrative.
Compute meets Solana
Compute activityLive
HourMachineGPU-hkWhNetSignature
Illustrative · Solana Base58 signature format. Illustrative records, not on-chain transactions.
01
Controlled access
The design prioritises user-controlled access to prompts, files and outputs. Access controls are not implemented by this current interface.
02
Minimal disclosure
The goal is to disclose resource usage, not workload content. No claim of zero logging or encrypted execution is made.
03
Solana-native direction
Solana is the intended network for compute payments and coordination. The signatures shown are examples, not on-chain transactions.
Understand the costof compute.
Review the cost model before launching a workload, with every component visible in this illustrative breakdown.
Illustrative $40 compute budget: power $5.60, space $3.60, maintenance $2.40, hardware cover $1.20, coordination $3.20 and execution $24.00.
$40.00, stacked to scaleHover a line
Illustrative $40 compute budget. These figures are not a price quote, an earnings forecast or a live billing schedule.
Know the designbefore you build.
The questions people ask first. The docs go deeper on privacy goals, compute and Solana.
read the docsCORELOT is a privacy-first AI compute concept with a Solana-native direction. It explores how GPU access, workload boundaries and transparent resource accounting can work together. This website is a interactive interface, not an operating compute network.
No. This is a concept preview. Private execution, workload routing and payment integration require implementation, security review and a documented launch.
Solana is the planned network for wallet-based compute access, USDC payments and resource coordination. Ordinary Solana transactions are public; using Solana alone does not make payments or workloads private.
Prompts, uploaded files, model outputs and application secrets are the intended privacy boundary. Exact protections and retention policies must be defined before a production launch.
Not currently. Confidential execution requires a verified technical implementation. CORELOT does not currently claim MPC, FHE, TEE integration, end-to-end encryption or zero logging.
No. GPU counts, utilisation, activity signatures and dollar values illustrate the product interface. They do not prove deployed hardware, customer usage or on-chain settlement.
No. This concept preview does not accept compute orders or deposits. Solana and USDC payment integration is planned, not live. Never send funds based on this interface.
Explore privacy-first compute
Build your AIwith privacy in mind.
Explore one sled or scale across a pool. Review the concept before any production integration.
- 01One sledA focused compute profile for inference and experiments. Concept preview.from $100 USDC
- 02A whole rackRack-scale capacity for larger AI workloads. Concept preview.from $500 USDC
- 03A poolDistributed capacity for parallel jobs and agents. Concept preview.from $50 USDC
Controlled accessPrivacy-first designSolana-native direction