Other answers have addressed the "why", but let's talk about the "what's real" regarding lumens.
Before we get into calculation the actual output of this light (which is definitely nowhere near 2000 lumens), we need to understand what lumens are. Lumens is the SI unit for luminous flux, which is essentially the total amount of light in the visible spectrum emitted by a source in all directions. Lumens does not depend on distance from the light source, because it must be measured in all directions. A related unit, lux, measures light at a point, and is dependent on distance. 1 lux = 1 lm/m2. If you were to measure the lux at every point on an imaginary sphere around a light source, add them up, and multiply by the surface area of that sphere, you would get the total lumens output of the device. No matter what the diameter of the sphere, you would get the same number of lumens. As you increase the diameter, the lux measured at each point decreases, but the surface area and number of points you must measure at increase.
It turns out this is almost exactly how you measure lumens in real life - with a scientific instrument called an integrating sphere. An integrating sphere uses a highly diffuse coating to evenly disperse light inside a sphere. All light entering through the sample port bounces around so many times that the brightness is nearly even at any point on the surface of the sphere. A photodetector, typically a radiometer, measures the intensity of the light received at the exit port across all visible wavelengths, then sums the intensities to determine the total luminous flux. Below is a photo of the integrating sphere I used to conduct testing of bike lights:
The ANSI FL1 Standard specifies measurement procedures and reporting criteria for measuring the output (lumens), beam distance (calculated from lux), battery life, and other properties of flashlights, headlamps, bike lights, and other personal lighting equipment. This provides a repeatable, verifiable, and trustworthy way for lighting manufacturers to report device specifications. Most reputable personal lighting manufacturers follow this standard, but random generic manufacturers - like those selling the light mentioned in the question - do not. These generic sellers tend to drastically overstate and even outright lie about specifications, claiming insanely high lumen values with no data to back them up.
So, now let's prove why this light doesn't put out 2000 lumens.
It appears this light uses two 18650 batteries. The Panasonic NCR18650 PDF is one of the most common 18650 lithium battery cells, and is the cell used inside of many repackaged batteries sold by flashlight companies and inside many devices with integrated batteries. Its rated capacity is 2700 mAh and the nominal voltage is 3.6 V. For two batteries, that gives 9.7 Watt-hours.
Let's look at the Cree XM-L T6 emitter, the LED specified for the product you mentioned. Cree's Data Sheet PDF gives luminous flux (lumens) data for up to 2000 mA current draw. The maximum forward current for this LED is 3000 mA. At 3000 mA the forward bias is 3.35 V, so the light would draw just about 10 W. With the batteries mentioned above, the light would last just under 1 hour, if we ignore the power consumed by the control electronics and any efficiency losses.
How much light does the LED put out at 3000 mA? I plotted the luminous flux vs. current draw for the light, and fit a linear function to the data:
At 3000 mA current, the maximum current for this emitter, the extrapolated luminous flux is 1000 lumens. To achieve 2000 lumens, the LED would require over 6000 mA, more than twice the maximum for this emitter. Keep in mind that the lumen values here are calculated estimates based on the emitter's specifications. They don't take into account losses due to reflector design and lens material. LEDs also drop in output when they get hot, and most cheap lights have poor temperature management.
I run a bike light testing and review site called The Bike Light Database, and I've tested similar generic headlights like this one. I measured a nearly identical light with a Cree XM-L T6 emitter and the actual output was 600 lumens, half of its claimed 1200 lumens and less than a third of this headlamp's claimed 2000.
So, to answer your question rather long-windedly, why a 2000 lumen headlamp? Well, it's simply not a 2000 lumen headlamp.