A few datasets analyzed and an invoice for 50,000 euros received: This has now happened to a user who publicly shares their mistake. The problem also shows how quickly one can underestimate data consumption when analyzing datasets.
Our title image is a symbolic image.
What happened to the user? The user explained that they wanted to explore the public Solana dataset in BigQuery and started writing simple test SQL queries using Python and the BigQuery API. According to them, it was only about basic things like searching for transactions by hash or address.
The surprise came after several queries were conducted and after 10 days. They received a bill for 50,850 euros.
To estimate the costs, they had used ChatGPT. A costly mistake, as they explained (via Reddit.com):
I had no idea that experimenting with a public dataset could carry significant financial risks. I had researched how billing works and inquired about the expected costs, among other things, by asking ChatGPT for rough estimates. Based on this, I was confident that my consumption would be within a reasonable range (around 30-50 dollars per month).
However, it is now clear to me that I approached the billing without sufficient caution and underestimated the potential financial risks, leading to an expensive mistake.
BigQuery charges fees based on data volume, which can lead to high costs
Where do the high costs come from? BigQuery is Google’s cloud-based database and analytics platform that bills based on actual data consumption. The fees primarily arise from the scanned data volume during SQL queries. Incorrectly configured or accidentally initiated queries (e.g., over very large tables or repeated attempts) can unintentionally lead to enormous costs.
Sometimes, automated tools/scripts, testing environments, or missing cost limits are used, which can cause high load unnoticed. And that can become really expensive over time.
How did it turn out? The user contacted Google and explained their mistake. After several messages, Google significantly reduced the user’s costs. In the end, the user is still left with a bill of 5,000 euros. This is considerably less than at the beginning, yet still a high amount for the user. They say: Google wants them to pay the 5,000 euros within 10 days. Now the user hopes that they can pay the amount in installments.
Another user also received a hefty bill. Car rental companies are now relying on AI to check rental cars for damages. This was also experienced by a driver who received a bill of around 375 euros after returning their vehicle. Because the AI found a damage: He returns his rental car and gets a hefty bill of 400 euros because the AI snitched on him.
