It is not only failing more. It is taking twice as long.
It is not only failing more. It is taking twice as long.
Kenneth · 2026-08-12
Source: LBNL Queued Up 2026 Edition (CC BY 4.0), through end-2025. Analysis my own.
Notes 002 and 003 established that the failure rate rose and that it rose almost
everywhere. This is the other axis: **for the projects that do get built, how
long does it take?**
Only projects that reached service. Cohorts with fewer than 15 completions omitted.
| Filed | n | Median years |
|---|---|---|
| 2003 | 18 | 1.5 |
| 2007 | 115 | 2.0 |
| 2011 | 59 | 2.5 |
| 2013 | 62 | 2.9 |
| 2015 | 67 | 3.8 |
| 2016 | 107 | 4.7 |
| 2017 | 70 | 5.0 |
| 2018 | 55 | 4.9 |
| 2019 | 75 | 4.4 |
**A project filing in 2003 was operating in eighteen months. One filing in 2017
took five years.**
The 2019 and 2020 cohorts show shorter medians — 4.4 and 3.2 years. That is not
an improvement. It is survivorship. Only the fastest projects from those
cohorts have completed at all; the slow ones are still in the queue or have
withdrawn, and they cannot appear in a median of completions.
Read those two rows as a floor, not a trend.
A 2017 filing faced roughly **85% odds of never being built, and five years to
find out.** A 2003 filing faced 64% odds and eighteen months.
**The cost of a queue position is not the interconnection fee. It is the option
value of five years of capital and site control**, spent on a coin that now lands
badly five times in six.
*Method: completions only, grouped by q_year, median of on_date − q_date.
Free at emp.lbl.gov/queues.*
*This is published research. Every figure is a count from a free dataset at emp.lbl.gov/queues — check it yourself rather than taking my word for it.