Interviewer view · keep this screen to yourself

Profit per GPU for a GPU cloud in Singapore

You run the case. Read the prompt, answer questions from the notes below, and share data only when the candidate asks for it or gets stuck. Score at the end.

Case timer

00:00

Target length:

1. Read the prompt aloud

Read it slowly, then pause. Let the candidate ask questions before they structure.

A GPU cloud (a neocloud) in Singapore rents out AI GPUs by the hour and reports in USD. Each GPU costs USD 35,000 all in, including its share of the server and network, and is written off over five years. It sells GPU time at USD 2.50 an hour and expects the GPUs to be busy 70 percent of the 8,760 hours in a year. Each GPU draws 1.25 kW including cooling, at USD 0.10 per kWh, around the clock. Colocation rent and support cost USD 2,000 per GPU a year. What is yearly profit per GPU before interest and tax, and at what utilisation does it break even? (Figures are illustrative.)

2. Answers to clarifying questions

This case has no scripted clarifying answers. Answer from the prompt, and say "assume what you think is reasonable" if the prompt does not cover it.

3. A model structure

Compare the candidate's structure with this one. A different split can be just as good if it is clean and fits the problem.

Profit per GPU = hours sold x price minus (depreciation + power + facility)
- Hours sold = 8,760 x utilisation
- Depreciation = GPU cost / useful life
- Power = kW x 8,760 hours x price per kWh
- Break-even utilisation = total costs / (8,760 x price)

4. The working, step by step

Each step shows how a strong candidate works it out. Share a new fact from it only when the candidate asks or is stuck, and let them do the math: the result in the dark box is what they should reach.

Step 1: Hours sold
What a strong candidate does: 70 percent of 8,760 hours.
GPU hours sold a year: 8,760 × 0.7 = 6,132
Step 2: Revenue
What a strong candidate does: 6,132 hours at USD 2.50.
Revenue per GPU a year (USD): 8,760 × 0.7 × 2.5 = 15,330
Step 3: Depreciation
What a strong candidate does: USD 35,000 over five years.
Depreciation per GPU a year (USD): 35,000 ÷ 5 = 7,000
Step 4: Power
What a strong candidate does: 1.25 kW for 8,760 hours at USD 0.10 per kWh (power is drawn even when idle, a simplification).
Power cost per GPU a year (USD): 1.25 × 8,760 × 0.1 = 1,095
Step 5: Profit per GPU
What a strong candidate does: Revenue minus depreciation, power and USD 2,000 of facility costs.
Profit per GPU a year before interest and tax (USD): 15,330 - 7,000 - 1,095 - 2,000 = 5,235
Step 6: Break-even utilisation
What a strong candidate does: Total costs of USD 10,095 divided by the revenue at full use, 8,760 hours at USD 2.50.
Break-even utilisation (percent): (7,000 + 1,095 + 2,000) ÷ (8,760 × 2.5) × 100 = 46.1

The recommendation to listen for

At the end, say: "The CEO walks in. What is your recommendation?"

The GPU cloud makes money at these numbers, about USD 5,235 per GPU a year, but it should lock in long contracts, because profit is very sensitive to price and chip life. First, it breaks even at about 46 percent utilisation, so a fall in demand hurts quickly. Second, if the price drops to USD 1.80 an hour at the same 70 percent utilisation, profit falls to under USD 1,000 per GPU, and if the chips must be written off over three years instead of five, depreciation rises to about USD 11,700 and profit falls to under USD 600 per GPU. The risk is that newer chips make these GPUs cheaper to rent within two or three years. As a next step, compare the length of customer contracts with the life assumed for the chips.

Risks a strong answer names: Newer, faster chips push down the price customers will pay for older GPUs; Interest on the debt used to buy GPUs is not in this profit figure; A single large customer leaving would cut utilisation sharply.

Score the candidate

Score each criterion from 1 to 5. A 2 or a 4 sits between the descriptions.

This case has no exhibit. Score Exhibit reading on how the candidate used the data you gave them: did they pick out the number that matters and say what it means?

Structuring

1: No clear structure, or a memorized framework that does not fit the problem.
3: A sensible structure, but with some overlap, a gap, or no clear priority.
5: A structure built for this problem, with no overlap and no gaps, a stated hypothesis, and a clear first step.

Math

1: Frequent errors, no units, or needed heavy help to finish.
3: Right answers but slow or messy working, or one error caught late.
5: Accurate and clearly narrated, with units, a sense check, and what the number means.

Exhibit reading

1: Read the chart aloud without finding the number that matters.
3: Found the key number but gave only part of the so-what.
5: Pulled the right number quickly and said what it means for the case.

Synthesis

1: Summarized what was done, with no clear answer.
3: Gave an answer, but with weak reasons or no next step.
5: Answer first, then two or three reasons with numbers, the main risk, and next steps.

Communication

1: Hard to follow: long silences, no signposting.
3: Clear most of the time, with some rambling.
5: Calm and concise, signposted each step, and checked in with the interviewer.

Total

0 out of 25

Score all five criteria to see the band and the feedback template.