— Jets.hunt at English Wikipedia, CC BY-SA 4.0, via Wikimedia Commons [cropped]
Before the digital display retrofit, the U-2 airspeed indicator shown above would have had the hand at the three o'clock position* in high-altitude cruise (around 110 knots). Any control response to the hand's up/down trend would have been ingrained already, so it would have made sense to keep the higher numbers low (low-to-high configuration).
* The hand's nominal o'clock position is very commonly designed for.
Another plane with this style is the Gulfstream IV:
— Honeywell (L); YouTube (R)
The GIV first flew in 1985. Two years later NASA Langley Research Center published a human factors research that looked into that exact question, and concluded that the now-common high-to-low gives the least workload:
The major factors considered were actual or reference tape centering, high-to-low or low-to-high airspeed tape orientation, trend information available or not, and basic or alternate display configurations [...] lower airspeed values at the top of the tape (low-to-high) [...]
Among the alternate display formats, it was perceived that there was lower work load when the airspeed tape had a high-to-low orientation than when it had a low-to-high orientation [...]
Subjective comments indicated that there was lower work load and better performance when the airspeed tape had the high numbers at the top instead of the low numbers at the top.
— Abbott, Terence S., Mark Nataupsky, and George G. Steinmetz. "Integration of Altitude and Airspeed on a Primary Flight Display." Published April 1, 1987. https://ntrs.nasa.gov/citations/19870010832 [emphasis added]
(Also to consider in today's instruments is the speed trend arrow.)