Energy Headline (Read This Like an Engineer)
Large AI companies like OpenAI are building massive data centers that require huge amounts of electricity. By October 16, 2025, OpenAI has committed to over 30 gigawatts (GW) of power capacity through partnerships. That's enough power to run about 25 million average U.S. homes continuously.
This level of demand is changing the energy market. Data centers are competing for electricity supply, which can drive up prices for everyone else. As an energy engineer focused on power systems, you'll see how this fits into grid planning and cost allocation.
OpenAI's Recent Deals: Decision Support → Intelligence → Compute → Energy → Costs
These deals show how decision support tools (like our simulations) lead to more intelligence, which requires more compute power, driving up energy use and costs. The past month has seen a rush of announcements, totaling over 30 GW and hundreds of billions in investments.
| Date | Partner | Deal Description | Value/Scale | Power (GW) | Link |
|---|---|---|---|---|---|
| Sep 11, 2025 | Oracle | Cloud computing agreement for Stargate data center infrastructure, including power provisioning for AI workloads | $300 billion over 5 years | 4.5 | OpenAI Announcement |
| Sep 22, 2025 | NVIDIA | Strategic partnership for AI infrastructure deployment, supplying GPUs and systems with integrated power solutions | Up to $100 billion | At least 10 | OpenAI + NVIDIA |
| Oct 6, 2025 | AMD | Multi-year agreement to supply AI chips (Instinct GPUs) for next-generation infrastructure | Tens of billions annually | 6 (starting H2 2026) | AMD Press Release |
| Oct 13, 2025 | Broadcom | Collaboration on custom AI chips and accelerators, supporting scalable data center expansion | Undisclosed value | 10 | OpenAI + Broadcom |
Notes: These commitments exceed 30 GW total, about 2% of the U.S. grid. Suppliers prioritize big buyers, raising costs for households, small firms, and cities. Ukubona's simulations aim for low energy use (<$0.01 per prediction) to stay efficient.
Why Ukubona Focuses on Energy
Ukubona develops computer simulations for kidney health decisions. For example, we model how kidney function might change over 20 years for potential organ donors, using factors like age, blood pressure, and genetic risks. These use data and math to predict outcomes and test "what if" scenarios.
Simulations run on computers and use electricity, like AI data centers. Rising energy costs from big AI demand make simulations more expensive, slowing research or raising prices for hospitals.
- Supply Challenges: Renewables grow, but big AI companies sign long-term clean power contracts first, leaving less for others.
- Cost Impacts: One large data center can add $10-20/month to household bills in areas like Ohio and Virginia by raising grid costs.
- Our Goal: Design simulations to use little power—under $0.01 per prediction—for efficiency and accessibility.
Energy Scales Comparison
| User Type | Typical Power Use (Continuous) | Annual Electricity (TWh) | Impact on Simulations |
|---|---|---|---|
| Household | 0.001 GW (1.5 kW average) | 0.013 | One simulation uses ~0.1 kWh—price hikes affect home health tools. |
| Small Clinic or Firm | 0.0001 GW (100 kW) | 0.0009 | Screening 100 cases: ~10 kWh—delays if costs rise. |
| City (100,000 homes) | 0.1 GW | 0.88 | Modeling 1,000 cases: ~1 MWh—grid strain slows planning. |
| Large AI System (OpenAI Total) | 30+ GW (committed) | 260+ | Projects use billions of kWh—we target <1% per result. |
| Entire U.S. | 500 GW | 4,380 | Hyperscalers take ~2% by 2026, raising costs for all. |
Encoding Complexity in Simulations (Aperiodicity)
A key part of our simulations is capturing how biological signals change over time. Data can have the same average value and spread (variance), but one pattern might be more irregular or "aperiodic" (non-repeating) than another. This irregularity often decreases with age or illness, making systems less adaptable.
For example, heart rate or kidney function traces might look similar at first glance but differ in how unpredictable they are. We use math tools like entropy measures to quantify this complexity—higher entropy means more aperiodic patterns, which can signal healthier adaptability. Studies show aging reduces this complexity in body systems.
In simulations, this helps predict long-term risks: two donors with similar starting kidney function might have different futures based on pattern complexity.
Kidney Disease Risk Factors (Plain English Guide)
Our simulations use common risk factors from medical studies to model kidney health changes. Here's a simple breakdown—no clinical background needed. These are factors that increase the chance of chronic kidney disease (CKD), where kidneys gradually lose filtering ability.
| Factor | What It Means | How It Affects Risk |
|---|---|---|
| Age | Years lived | Older age raises CKD risk as kidney function naturally declines. |
| Sex | Male or female | Men often have higher risk, but it varies by other factors. |
| Race | Ethnic background | People of African ancestry have higher risk, partly due to genetics. |
| Diabetes (DM) | High blood sugar condition | Major cause—damages kidney blood vessels over time. |
| Hypertension (HTN) | High blood pressure | Strains kidneys, leading to scarring and reduced function. |
| Systolic Blood Pressure (SBP) | Top number in blood pressure reading | Higher SBP (>140 mmHg) speeds up kidney damage. |
| Urine Albumin-to-Creatinine Ratio (uACR) | Protein leak in urine | Early sign of damage; higher levels predict faster decline. |
| Estimated Glomerular Filtration Rate (eGFR) | Kidney filtering speed (mL/min) | Lower eGFR (<60) indicates existing damage and higher risks of death or heart issues. |
| Smoking | Tobacco use | Harms blood vessels, worsening kidney strain. |
| Body Mass Index (BMI) | Weight relative to height | Higher BMI (>30) links to diabetes and pressure, raising CKD odds. |
| APOL1 High-Risk Genotype (in people of African ancestry) | Specific gene variants | Increases CKD risk after donation; about 15% prevalence in Black donors, with higher failure rates long-term. |
These factors interact—e.g., high blood pressure plus diabetes doubles risk. Simulations test combinations to predict 20-year outcomes.
12-Month Overview (20 Hours/Week, Big Picture)
Light schedule: Focus on energy tracking in simulations. Monthly goals build skills gradually—no overload.
| Month | Big Picture Focus | Key Activities |
|---|---|---|
| 1 | Setup & Basics | Demo notebook with data patterns & power estimates; outline simulation inputs/outputs; sample data files. |
| 2 | Data Generation | Build tool for 10k synthetic cases; add energy tracking to outputs. |
| 3 | Dashboard Build | Interactive viewer with sliders for factors; show predictions + energy costs. |
| 4 | Complexity Add | Implement entropy metrics; demo aperiodic patterns in traces. |
| 5 | Risk Integration | Incorporate factors like age, BP, genetics; test interactions. |
| 6 | Stress Tests | Add sudden events (e.g., pressure spikes); measure energy bursts. |
| 7 | Outcomes Reporting | Build summaries for long-term risks; track total kWh used. |
| 8 | Calibration Check | Compare sim results to study data; adjust for accuracy. |
| 9 | Energy Optimization | Test scheduling for low-cost runs; reduce power per sim. |
| 10 | Documentation | Write guides on setup, metrics, energy tracking. |
| 11 | Review & Polish | Test full workflow; refine dashboard for efficiency. |
| 12 | Wrap-Up | Final report on energy insights; handoff notes. |