Choosing a battery size used to be a rough guess based on how many hours of backup a homeowner wanted. Under NEM 3.0, that approach leaves real money on the table, because the battery is no longer just insurance against an outage; it is a financial tool that determines how much your exported solar power is actually worth. California’s net billing rules pay far less for power sent to the grid at midday than utilities charge for power pulled back at night, and the size of your battery decides how much of that gap you can close. This is where AI-based system design earns its place in the process. Instead of applying a flat rule of thumb, an AI model can read your home’s hourly usage pattern, your utility’s time-of-use rate structure, and your roof’s production profile, then calculate the battery capacity that captures the most export value without oversizing the system. This guide walks through how NEM 3.0 changed the math and what actually goes into an AI-generated sizing recommendation.
Why NEM 3.0 Turned Battery Sizing Into a Money Question
What Net Billing Actually Pays for Exported Power
Under the previous net metering structure, exported solar power was credited close to the retail rate a homeowner would otherwise pay. NEM 3.0 replaced that with a net billing tariff that pays a variable, generally much lower, avoided cost rate for exports, while imports at night are still billed at full retail or time-of-use rates. That gap between what you’re paid for sending power out and what you’re charged for pulling power back is the entire reason battery sizing now has a dollar figure attached to it. Homeowners comparing different solar setups can also compare solar financing and pricing options before deciding how much storage makes financial sense.
Why a Backup-Only Mindset Leaves Savings on the Table
A battery sized purely for outage backup will often sit partially empty during the hours when export compensation is lowest, and grid import pricing is highest. Sizing for financial optimization means the battery does daily work instead of just waiting for an emergency, and that daily work is exactly what an AI model is built to calculate.
How Export Value Is Calculated Under NEM 3.0
Time-of-Use Rates and the Value of When You Export
Most NEM 3.0 utilities pay the lowest export rates during the mid-afternoon hours when solar production peaks and pay the highest import rates in the early evening when demand peaks and the sun is going down. A correctly sized battery stores midday production and discharges it during that evening window instead of exporting it for pennies, which is where most of the realized savings come from. You can see how these windows shift by utility and territory using the current electricity rates in California that Axia tracks for its design models.
Avoided Cost vs Retail Rate: The Gap AI Has to Close
The dollar gap between the avoided cost export rate and the retail import rate is the number an AI sizing model is ultimately trying to close. A larger battery closes more of that gap per kilowatt-hour shifted, but only up to the point where the home’s actual evening usage stops needing more stored power, which is why the calculation has a ceiling rather than a straight line.
The Real Cost of Guessing on Battery Size
Oversizing: Paying for Capacity You Never Use
A battery larger than a household’s evening load will spend part of every day sitting partially charged, since there’s no shiftable production or usage left to fill it. That unused capacity is paid for up front and never recovered, which is one of the most common and most expensive mistakes in a manually sized system.
Undersizing: Exporting at the Cheapest Hours of the Day
The opposite mistake, a battery too small for the home’s actual evening draw, means midday production keeps spilling out to the grid at the lowest export rate of the day even after the battery is full. Both mistakes come from sizing off of rules of thumb instead of the household’s actual hourly data, and both cost money for the life of the system. Rate structures also vary enough by state that a size which pencils out well in one service area can miss the mark elsewhere, which is worth knowing given how electricity rates compare state to state across Axia’s coverage area.
How AI Models Your Home’s Hourly Load and Export Pattern
Building an Hourly Profile From Interval Data
Rather than estimating from a monthly bill total, the AI model builds an hour-by-hour profile of the home’s electricity use across a full year, layering in seasonal swings in heating, cooling, and daylight hours. That resolution is what separates an optimized recommendation from a rough estimate.
Matching Battery Discharge Windows to Peak TOU Pricing
Once the usage profile exists, the model overlays the utility’s time-of-use pricing windows and identifies exactly how many kilowatt-hours need to shift from midday production into the evening peak to capture the most value. This is the same underlying approach behind the broader AI solar design process Axia uses for panel layout, extended into the storage side of the system.
Accounting for Seasonal Shifts in Production and Usage
Summer production and winter production rarely match summer and winter usage, so the model has to check the sizing decision across every season rather than optimizing for a single average month, which is where a lot of simplified sizing tools fall short.
Inside an AI-Generated Battery Sizing Recommendation
Inputs the Algorithm Actually Uses
A sizing recommendation draws on the home’s historical or estimated hourly usage, the utility rate schedule for the address, the proposed array’s expected hourly production, and any known future loads, such as an electric vehicle. Each input affects the output, which is why two neighboring homes on the same rate plan can end up with different battery sizes. You can also get a personalized solar estimate for your home to see how system size, production, and battery needs can vary based on the property.
Why the Output Is a Range, Not a Single Number
The model typically returns a usable range rather than one exact number, since a homeowner’s backup preferences and budget still factor into the final choice. What the AI removes is the guesswork in the financial baseline, so the range being chosen from is already grounded in real export value math instead of an installer’s rule of thumb. You can run your own address through the same underlying logic with Axia’s AI solar calculator.
Where a Human Designer Still Reviews the Recommendation
Every AI-generated sizing output still gets reviewed by a human designer before it reaches a homeowner, since local permitting limits, panel space, and equipment availability can all adjust the final number the algorithm alone wouldn’t catch.
Why Axia Solar’s AI Design Process Gets This Right
One Continuous Model From Roof to Battery to Rate Plan
Axia’s design platform doesn’t treat panel layout, production estimates, and battery sizing as separate steps. The same hourly production model that lays out your roof feeds directly into the battery sizing calculation, so the recommendation reflects your actual system rather than a generic average. That continuity is what the AI-powered solar estimate process is built around.
Built for the Specific Utility Territory Your Home Sits In
Rate schedules, export tariffs, and battery incentives differ by utility territory even within the same state, and the model pulls the specific structure for your address rather than applying a statewide average.
Battery Sizing Considerations Beyond the Algorithm
Backup Priorities Beyond Financial Optimization
Some households want enough stored capacity to ride out a multi-day outage regardless of what the export math says, particularly in wildfire-prone areas with a history of public safety power shutoffs. That preference is a legitimate input, not a mistake the algorithm needs to correct.
Future Loads That Change the Math (EV Charging Etc.)
Adding an electric vehicle, a heat pump, or a pool pump after the system is installed changes the household’s evening load enough to shift the ideal battery size, so the sizing conversation should account for loads planned within the next few years, not just current usage. Local incentive programs can also change the calculation, and homeowners in Pasadena should check Pasadena battery rebate eligibility before finalizing a size, since a rebate can shift which capacity makes financial sense.
Getting the Battery Math Right the First Time
Battery sizing under NEM 3.0 isn’t a backup decision anymore; it’s a financial one, and the difference between a battery sized off a rule of thumb and one sized off actual hourly export math can run into thousands of dollars over the life of the system. An AI model that reads your home’s real usage pattern, your utility’s actual rate structure, and your roof’s production profile removes the guesswork from that decision while still leaving room for a human designer and your own backup preferences to shape the final number. If you want to see what that math looks like for your own address, request a custom AI solar design and get a battery recommendation built around your actual home instead of an average one.
What is NEM 3.0 export value?
NEM 3.0 export value is the credit a California solar customer receives for electricity sent back to the grid under the state’s net billing tariff, which pays a variable avoided cost rate instead of the closer-to-retail rate used under the previous net metering rules.
How is battery size determined under NEM 3.0?
Battery size is determined by comparing a home’s hourly electricity usage against its solar production and the utility’s time-of-use rate schedule, then calculating how much stored capacity is needed to shift midday production into higher-value evening hours.
Does a bigger battery always mean bigger savings?
No. Once a battery is large enough to cover the shiftable gap between a home’s production and its evening usage, additional capacity sits unused and stops adding financial value, so bigger is only better up to that point.
How does time-of-use pricing affect battery sizing?
Time-of-use pricing sets the value of every kilowatt-hour a battery shifts, since exporting during low-priced midday hours and importing during high-priced evening hours is what a correctly sized battery is designed to avoid.
Can AI battery sizing account for future EV charging?
Yes. An AI model can incorporate planned future loads like an electric vehicle or heat pump into the sizing calculation so the recommended capacity still makes sense once those loads are added, not just for the home’s current usage.
How accurate is AI-based battery sizing compared to manual estimates?
AI-based sizing is generally more accurate because it works from hour-by-hour usage and production data across a full year rather than a single monthly average, which is the level of detail manual rule-of-thumb estimates typically skip.
