When you deploy a Lektra BusinessEdge server at your solar property, you aren't just buying hardware. You’re activating a high-yield asset within Lektra’s distributed network. Your site becomes a critical node, supplying the raw compute power required for the world’s most advanced AI and media applications.

We manage the complexity. You capture the margin.

Lektra handles the heavy lifting: cloud orchestration, developer demand, secure billing, and 24/7 workload monitoring. By combining your low-cost energy with our optimized software layer, you generate consistent, hands-off revenue from your existing solar infrastructure.

Think of a GPU like a high-end rental property. To make the most money, we don't just wait for someone to show up; we use four different "rental layers" to ensure your hardware is working—and earning—nearly 24/7.

1. On-Demand (The "Hotel Booking")

What it is: Developers who need power right now for a specific project. They pay a premium hourly rate, similar to a guest booking a hotel room for a few nights.

Why it’s great: These are our highest-margin hours. We charge more for the convenience of "instant access."

2. Reserved (The "Long-Term Lease")

What it is: A tech company signs a contract to "rent" your GPU for many months.

Why it’s great: This is guaranteed income. Whether they use the GPU for 1 hour or 24 hours that day, they pay for the whole day. This provides the steady "base-load" revenue for your facility.

3. Spot (The "Last-Minute Filler")

What it is: When a GPU isn’t being used by a lease or a hotel guest, we don’t let it sit idle. We offer it at a discount to researchers running background tasks (like "training" an AI).

Why it’s great: It ensures your equipment stays productive. It’s always better to earn a discounted rate than to earn $0 while the power is still on. Spot is interruptible, meaning it can be paused or moved if a Reserved or On-Demand customer needs the GPU.

4. EdgePulse (The "Digital Toll Road")

What it is: Instead of one person renting the GPU for an hour, our EdgePulse technology can process thousands of tiny AI "tasks" per second, like generating an image or a chat response for users all over the world.

Why it’s great: It allows the GPU to make money in the tiny gaps between other jobs. It’s like a vending machine that never sleeps, collecting "micro-earnings" every time someone uses an AI model in their app.

Optimized Utilization
By blending these four types of rentals, Lektra aims to keep your GPUs "occupied" 70% to 90% of the time. This specialized mix ensures you aren't just waiting for the phone to ring. You are running a high-yield digital power plant.

AI developers and enterprises rent compute power from Lektra using a professional cloud model. Because our structure follows the same standards used by major global cloud providers, your hardware is instantly familiar and attractive to a wide range of high-value customers.

There are four ways developers utilize GPUs on Lektra:

1. On-Demand (The "Hotel Booking")

Maximum Flexibility, Maximum Margin. On-Demand is the premium tier of the Lektra network. Developers get instant, zero-commitment access to your silicon at the full market rate. Because they can stop anytime, they pay a premium for that agility, putting the highest revenue-per-hour directly into your pocket.

The Revenue Sprint
While utilization may fluctuate more than Reserved plans, the higher hourly rate makes On-Demand a massive driver of monthly ROI.

Example: 70% Utilization (504 active hours/month)

On-Demand Rate: $1.84/hr (RTX 6000 Ada)

Gross Revenue: $927.00 per GPU / month

Net to Host: $742.00 per GPU / month (After Lektra’s 20% Ops Fee)

For a standard 8-GPU Server:

Monthly Net Payout: $5,936.00

Why this scales:
In big cloud, on-demand usage often comes with extra charges that add up fast. Lektra keeps it lean. With no egress fees, developers can run and iterate without worrying about surprise costs, which makes them more likely to keep workloads running on your hardware longer.

2. Reserved (The "Long-Term Lease")

Reserved contracts are where you de-risk your hardware deployment. Developers commit to 3, 6, or 12-month terms in exchange for a discount. In return, you get a guaranteed revenue "floor" that covers your overhead, regardless of market volatility.

Typical discounts (industry standard):

3 months: ~10% off

6 months: ~15% off

12 months: ~18-20% off

The Math of Certainty
Reserved capacity isn't just about a lower rate; it’s about 100% occupancy. While On-Demand fluctuates, a Reservation is a "set it and forget it" revenue stream.

Example: 6-Month Reservation (Full Occupancy)

Reserved Hourly Rate: ~$1.56/hr (based on $1.84 base)

Gross Revenue: $786.00 per GPU / month

Net to Host: $629.00 per GPU / month (After Lektra’s 20% Ops Fee)

For a standard 8-GPU Server:

Monthly Net Payout: $5,032.00

Why this scales:
The tradeoff is simple: you trade a slice of margin for absolute predictability. Reserved contracts allow you to project your earnings with total accuracy months in advance. It turns your rack from a "side hustle" into a professional infrastructure business. Lock in the floor, cover your costs, and use the remaining capacity for high-margin sprints.

3. Spot (The "Last-Minute Filler")

Spot instances allow you to squeeze revenue out of the hours your GPUs would otherwise spend gathering dust. It’s the ultimate efficiency play. Perfect for non-urgent, high-compute workloads like batch processing.

Standard Discount: 50–60% off On-Demand rates.

The Math: Recovering the Sunk Cost
If your GPUs have 504 hours of high-tier bookings, you’re left with 216 hours of "dead air" per month. Spot turns that silence into a revenue stream.

Example: 50% Spot Fill (108 additional hours)

Spot Hourly Rate: ~$0.83/hr (at a 55% discount)

Gross Revenue: $90.00 per GPU / month

Net to Host: $72.00 per GPU / month (After Lektra’s 20% Ops Fee)

For a standard 8-GPU Server:

Monthly Net Payout: $576.00

Why this matters:
Spot ensures that even your "leftover" capacity is working for you, filling the gaps between Reserved and On-Demand contracts. It’s the difference between a rack that just "covers costs" and a rack that generates a true surplus. If your silicon is on, it should be earning.

4. EdgePulse (The "Digital Toll Road")

Instead of waiting for a single customer to book a block of hours, EdgePulse taps into a global stream of micro-requests (images, LLM chats, embeddings). Your GPUs act as a high-frequency processing hub, earning revenue from every active second of traffic.

EdgePulse Rate (RTX 6000 Ada): $0.000664 per GPU-second
Target Yield: ~$2.39 per active GPU-hour

The Math: Turning Idle Time into Income
A standard month has 720 hours (2,592,000 seconds). Even with modest utilization, the serverless stream adds up fast.

Example: 11.6% Utilization (300,000 active seconds/month)

Gross Revenue: $199.20 per GPU / month

Net to Host: $159.36 per GPU / month (After Lektra’s 20% Ops Fee)

For a standard 8-GPU Server:

Monthly Net Payout: $1,274.88

Why this scales:
Most clouds still waste capacity, leaving GPUs underutilized between workloads. EdgePulse is the engine that fills those gaps with real-time inference traffic. As we add more models and deeper integrations, your server earns from more use cases, more often. Higher utilization means stronger earnings per month.

Bottom line

On-Demand drives top hourly revenue. Reserved smooths and stabilizes income. Spot helps monetize otherwise-empty hours. EdgePulse adds high-volume inference traffic that can lift utilization over time. Combined, they’re designed to keep your server earning more consistently across the month.

What developers pay (gross) and energy hosts receive (net).

Assumptions used in the table:
1. Lektra operational fee: 20% → Host payout = Gross × 0.80
2. Reserved shown as 6-month plan at 15% off on-demand
3. Spot shown as 55% off on-demand (interruptible)
4. EdgePulse uses a serverless rate: 1.3× the on-demand anchor, billed on active GPU time (shown as $ per active hour).

Your real-world earnings will vary based on how many hours your GPUs are used and the mix of usage across On-Demand, Reserved, Spot, and EdgePulse.

High-performance AI doesn't just need compute; it needs lightning-fast access to massive datasets, model weights, and checkpoints. Every Lektra server utilizes its high-speed NVMe overhead to provide developers with the stateful storage they require.

Developers pay for storage based on how much NVMe space they use each month:

Storage Rate: $0.10 per GB / month (Active Storage)

Storage demand scales with your hardware tier. While an RTX 6000 node might primarily handle fine-tuning and light-weight inference, a B200 or B300 powerhouse is often utilized for heavy training and fine-tuning, requiring massive data throughput.

Storage is billed monthly, similar to renting disk space in a traditional cloud.

NVMe drives use only a few watts, making storage a stable, recurring revenue stream that adds almost pure profit on top of your GPU earnings.

For realistic, real-world power budgeting, a full system draws more than just 8× the GPU wattage. Additional server overhead includes: CPU, memory, cooling fans, NICs, power conversion loss, etc.

Net revenue per kWh reflects host earnings after Lektra’s 20% operational fee, any applicable distributor fee, and estimated electricity and internet costs required to operate the server. Internet costs can vary significantly based on bandwidth, redundancy requirements, location, and the scale of deployment. For the most accurate pricing, Lektra will work directly with you to model connectivity costs for your specific site and deployment size.

Multi-server deployments may require additional networking and infrastructure costs (switches, racks, electrical work, backup power, and additional cooling or ventilation). Actual results may vary based on market demand, system configuration, site conditions, uptime, and pricing fluctuations.