This is my thought process:
On average one accident every 2 million miles (today the number is already better than that when using autopilot, assume this will continue to improve in the future, autopilot/FSD will get better, and people will drive more autopilot miles,).
Assume all accidents are Tesla driver's fault (in reality some accidents are caused by the other driver's).
Assume on average insurance payout for each accident is $100k (some accidents should cost much less). Down the road Tesla can fix Tesla cars in their own shop and reduce cost.
Average mileage per car per year 14k miles.
For 1000 cars insured, the premium is $1.8m (using your $1800/year number).
These cars will drive 14 million miles in a year, get 7 accidents. The payout should be $0.7m. So roughly 50% combined ratio. That's $900 cash retained from each car each year. This cash can be re-invested indefinitely, in the end the $900 and all the investment gain all belongs to the insurance company. In the long run, the company could earn $4000 from this one car in one year, isn't that crazy? it could happen because of the investment gains in the next few decades. So insurance business is truly an amazing business if you can get nice combined ratio and as long as the chief investment officer doesn't mess up. Many CIO/CEOs of insurance businesses do mess up sooner or later (see AIG in 2008), but these three stand out: Lou Simpson, Tom Gayner, Warren Buffett. There are a few other good ones, but they may not have seen the real tests yet.
I bet having 50% combined ratio never happened in history, reaching 90% would be a dream business. Tesla is in a unique position to make this happen while also help the car owners with lower rate.
In real world, only a small portion of accidents happen on year one, so there is additional investment gain before payout. If the average accident/payout happens on 4th year, you may gain additional 10% before the payout. That $90 extra gain from each car can turn into significant amount in the long run.
The above is one of the 2 reasons why Berkshire Hathaway stock price grew from $15 to $320,000. It's also why MKL grew from $8.33 to $1061. If Tesla can make the insurance business happen, the potential of that insurance unit is as big as the car business (not counting the Tesla Network). Because when you look back in 30 years, you only get $10k by selling each car (then you have to take care of all sorts of problems, then you have to use that $10k to buy more equipments); on the other hand, you earn $9k by insuring that car for 10 years, plus that $9k turns into $36k in 20 years if you get 7% gain per year, a balanced index approach would give you that result.