F. Electricity cost and solar for the owned nodes#

Status: Plan Last verified: 2026-09-27 Canonical for: the Gemini Deep Research prompt for the electricity and solar workstream

The problem#

Our docs disagree on what electricity costs us, and neither figure comes from a real bill:

"Malaysia is an advantage" and "owning our power (solar) is a long-term goal" are founder positions; see benchmarks/CLAUDE.md. This research is meant to give them solid figures, not to argue with them.

The four questions:

  1. What we actually pay per kWh at our real consumption.
  2. How that compares like for like around the world.
  3. Where it goes over the next three years.
  4. What rooftop solar would cost and cover for a load that runs day and night.

Paste everything in the block below. Save the result as results/f-electricity-and-solar.md. Then reconcile the strategy doc and synthesis.md against it and against a real TNB bill.

CONTEXT
I run a small AI inference business in Peninsular Malaysia (image and video generation). It owns consumer GPU servers: currently three PCs, each with one NVIDIA RTX 5090, a Ryzen 9 9950X and a Seasonic 1200 W PSU. I plan to buy several more soon, so assume 3 to 8 nodes. They run in a residential premises on an ordinary TNB domestic supply (single- or three-phase, low voltage), not in a data centre. The business is permanent and owning its compute is a long-term strategy. Owning its own power, starting with rooftop solar, is a long-term goal. I am not asking whether to own; I want exact electricity economics.

Load assumptions (not yet metered):
- Per node: about 0.5 kW average at moderate utilisation, up to about 0.8–0.9 kW sustained at full GPU load, and about 0.1 kW idle.
- The nodes run 24/7. Load follows user demand, which is spread across time zones, so it is not concentrated in Malaysian daytime.
- The household also has its own base consumption. Treat this as a variable: show results at 300, 600 and 1,000 kWh a month of household use.
- Heat from the nodes may need air-conditioning in a tropical climate.

My planning horizon is October 2026 to September 2029 for tariffs, and 10–25 years for solar.

TASK 1: WHAT WE ACTUALLY PAY (TNB, Peninsular Malaysia, RP4 framework from July 2025)
1. Give the exact current domestic tariff structure: energy charge by consumption tier (including the 1,500 kWh threshold), capacity charge, network charge, retail charge, the Automatic Fuel Adjustment (AFA) surcharge (every monthly value you can find for 2026, and any published forecast), the Energy Efficiency Incentive, service tax (SST) rules and thresholds, and any exemptions (for example, below 600 or 800 kWh). Quote the official TNB or Energy Commission (Suruhanjaya Tenaga) figures.
2. Is there a domestic Time-of-Use option under RP4? Give its peak and off-peak rates and windows, and say whether it would be cheaper for a flat 24/7 load.
3. Compute the full monthly bill and the effective RM/kWh, both blended and marginal, for total monthly consumption of 1,000, 1,500, 2,000, 3,000, 4,500 and 6,000 kWh. Show the arithmetic.
4. Compare with the low-voltage commercial tariff (Tariff B or its RP4 equivalent) at the same consumption, including any capacity or maximum-demand charges. Which is cheaper for this load, and at what consumption does that flip?

TASK 2: LIKE-FOR-LIKE INTERNATIONAL COMPARISON
Give the current effective price per kWh in US cents, with the exchange rate and date used, for comparable users:
- Residential (typical and high-consumption tiers): US national average plus California, Texas and Virginia; Germany; United Kingdom; Singapore; Japan; Thailand; Indonesia; Vietnam; Malaysia.
- Small commercial, the same countries where available.
- Separately, and clearly labelled as NOT comparable to us: typical data-centre and industrial contract rates in the US (for example Virginia and Texas), in Johor (the Malaysian data-centre tariff), and in Singapore.
Say where Malaysia's retail rate ranks and why: fuel mix, subsidies, regulation.

TASK 3: WHERE IT'S HEADING (to September 2029)
1. What is known or proposed for Regulatory Period 5 (RP5, from 2028): base-tariff direction, and capacity and network charges?
2. Subsidy rationalisation: any plans to reduce the protection for domestic users, especially high-consumption domestic users?
3. How much of AFA movement is driven by fuel prices (gas and coal) and the ringgit? What do current fuel forecasts imply?
4. Does the data-centre boom (data centres pay their own tariff) raise costs for other users through grid investment passed through in RP5?
5. Give a base, low and high path for our effective RM/kWh through 2029.

TASK 4: ROOFTOP SOLAR FOR A 24/7 LOAD
1. Which schemes are current in 2026: Solar ATAP (which replaced NEM 3.0), any successor, and any self-consumption rules? Give the exact terms: how exported energy is credited (at what rate, netting period, caps), system size limits relative to the supply and the bill, application process, and eligibility for residential premises running business equipment.
2. Installed cost of residential or small-commercial rooftop PV in Peninsular Malaysia in 2026, in RM per kWp, for 5, 10, 15 and 20 kWp. Also any financing schemes.
3. Specific yield in Peninsular Malaysia (kWh per kWp per year), degradation, and inverter replacement.
4. For a flat 24/7 load of 1.5, 3 and 6 kW (3, 6 and 12 nodes at 0.5 kW), how much of the energy can solar cover:
   (a) through direct daytime self-consumption;
   (b) with export crediting under the current scheme;
   (c) with batteries, using current LFP battery cost per kWh in Malaysia and the system size needed.
   Give the payback period and levelised cost of energy (RM/kWh) for each against the TNB rates from Task 1.
5. What limits a residential premises: roof area per kWp, supply capacity, TNB approval?

TASK 5: COOLING
For a tropical indoor environment, what is the extra air-conditioning electricity per kW of IT heat (typical split-unit and inverter air-conditioner efficiency, COP or EER, in Malaysian conditions)? Give a rule of thumb, for example "every 1 kW of servers adds X kW of air-con", and when ventilation alone is enough.

SOURCES
- Today is late September 2026. Prioritise sources published in the last 12 months. For tariff structure, use the official TNB, Energy Commission (Suruhanjaya Tenaga), SEDA and Ministry of Energy Transition and Water Transformation (PETRA) documents and gazettes, and quote them.
- For international prices, use national statistics or regulators (for example the EIA, Eurostat, Ofgem, Singapore's Energy Market Authority) rather than aggregator blogs where possible, and date every figure.
- Solar installer quotes and marketplace prices are fine for cost, labelled as such.
- Where credible sources disagree, show both positions with their reasoning. Do not average them.
- Give a URL for every sourced claim.

OUTPUT FORMAT
1. Answer: at most 12 bullets answering the tasks directly, with numbers.
2. The bill calculation table (Task 3 of TASK 1) with arithmetic shown.
3. The international comparison table in US cents per kWh, with date and exchange rate.
4. Evidence table with columns: claim | number | as-of date | source (URL) | observed / forecast / rumour | leading / lagging.
5. Scenarios to September 2029 for our effective RM/kWh: base, low and high, with rough probabilities and the early indicator for each.
6. Solar options table: system size, cost, share of the 24/7 load covered, payback, levelised cost of energy, under the current scheme.
7. Unknowns: what you could not find or verify, especially anything that depends on the actual bill or premises.

Do not give generic investment advice. Keep the focus on the questions above.