Proof · Tested on the past · For Solvay

In 1998, the data already knew which light would win homes, shops and offices

The data picked white LED in 1998. Philips announced it would split off its lighting business 16 years later.

Rooms were lit by filament bulbs, then by fluorescent lamps, where a coating of rare-earth phosphors turns the glow of mercury vapour into white light. The white LED is a small blue chip with a phosphor coating that turns part of its light yellow, so the mix looks white. For each we measured how much better per dollar it got each year, using only the patents that existed at the time.

Computed from patents published by each date · nothing here is hindsight
01 · The reading

How fast each light source improved

% better per dollar each year. One reading a year. White LED line starts at its first reading, 1998.

1
2011
US 60-watt bulb prize goes to an LED bulb
2
2014
Philips to split off its lighting business
3
2015
Osram to carve out its lamps business
4
2020
LED is 47.5% of all US installed lighting
2.3×
White LED was improving 2.3 times as fast as fluorescent lamps in 1998 · 29% a year against 13
16 years
From the signal to Philips announcing it would split off its lighting business
02 · At La Rochelle

New phosphors, and far less rare earth per lamp

What the data showed.

From its first reading in 1998, white LED was improving faster than both kinds of lamp: about 29% better per dollar each year, against about 13% for fluorescent lamps and 10% for filament bulbs. The gap widened to 2007, and white LED has led at every reading since.

What that meant at La Rochelle.

The early white LED inventions were mostly about the coating: almost half improve the phosphor, and half aim at the colour of the light. The new light still used rare earths, but new phosphors, and ten to a hundred times less of them for the same light.

Early white LED inventions
Improve the phosphor~½
Aim at the colour of the light~½
Less rare earth for the same light10–100×
New phosphorsStill rare earths
03 · Then the market caught up

Sixteen years from the signal to Philips splitting off lighting

1998

Our data: white LED improving more than twice as fast as fluorescent lamps.

Signal
2011

A US government prize for a replacement of the 60-watt bulb goes to an LED bulb.

+13 yrs
2014

Philips announces it will split off its lighting business.

+16 yrs
2015

Osram decides to carve out its lamps business.

+17 yrs
2015

Solvay’s results report lower rare-earth sales in lighting.

+17 yrs
2020

LED is nearly half of all lighting installed in the US.

+22 yrs
2023

The EU stops new fluorescent tubes from being sold.

+25 yrs
04 · What a 1998 reading would have been worth

The window to plan lamp volumes down and move lab hours to LED phosphors

It would not have said when LED bulbs would get cheap, or which lamp makers would last. It would have said white light would come from LEDs, with new phosphors and far less rare earth per lamp, sixteen years before the biggest lamp makers began splitting off their lamps.

For a lamp phosphor supplier, that is the window to plan lamp volumes down and move lab hours to LED phosphors and other rare-earth uses.

16 years
From the signal to the biggest lamp makers splitting off their lamps
10–100×
Less rare earth for the same light

In 2025 La Rochelle opened a rare-earth line for magnets. The same reading can show which separation route pulls ahead now.

The method, in one minute

What is TIR?

TIR — the Technology Improvement Rate — measures how fast a technology gets better each year, computed from global patent data. Not who is filing, or how much: how fast the route is actually moving. Historically, the fastest-improving technology always wins.

Signal 1 · Cycle time
How many years between one generation of inventions and the next.
Shorter cycles = faster progress toward newer generations.
Signal 2 · Knowledge flow
How much each new generation is cited by later innovations.
More citations = bigger, more important step forward.
One number
Together they give a single, objective measure of speed — taking the guesswork (and politics) out of R&D bets.
Read more about the method →
Methodology co-developed with MIT — 30 years of peer-reviewed research.

Provenance. Every patent carries a publication date, so the record can be replayed: every improvement rate on this page uses only the patents that were published at the time. Chart data: GetFocus platform exports, October 2026.

Your next reading

Bring us the choice you are weighing now

Bring us one live technology decision. Within 48 hours you get a free preliminary Decision Brief: a first-pass radar, improvement rates and suggested triggers.

Free · 48 hours · preliminary scoping. Validated radars come with a subscription