Uni-Sine
PhysicsAtoms and particles

Microsoft explains how lead gave Majorana 2 a 1,000x gain

Microsoft's explainer, published on its YouTube channel on 16 November 2026.

On 16 November 2026, Microsoft published an eight-minute video explaining how, by its own measurements, its Majorana 2 quantum chip came to hold its state about 1,000 times longer than the chip before it. The company's answer is a change of materials: lead in place of aluminium, and antimony added to the semiconductor underneath. Physicists outside Microsoft still dispute what the measurements prove.

Microsoft announced Majorana 2 on 2 June 2026 at its Build conference in San Francisco, with a preprint signed by 165 authors. It is the company's second chip built on topological qubits, which store quantum information in the parity of a nanowire: whether it holds an even or an odd number of electrons. Majorana 1, the first, was shown in February 2025.

What changed inside the chip

The Majorana 2 layers pulled apart: lead on top, then a top barrier, indium arsenide, indium arsenide antimonide, a bottom barrier, a buffer and gallium antimonide
The Majorana 2 stack, top to bottom. Diagram: Uni-Sine, after Microsoft's figure and the paper. Not to scale.

Each qubit sits on a stack of crystal layers grown one on top of another. In Majorana 1 the top layer, the superconductor, was aluminium, over a thin sheet of indium arsenide (InAs) where the electrons move. In Majorana 2 the superconductor is a film of lead 10 nanometres thick. Under it the electrons now move in two layers: 6 nanometres of indium arsenide and 2 nanometres of indium arsenide antimonide (InAsSb), the same crystal with a fifth of its arsenic swapped for antimony. The stack is grown on gallium antimonide (GaSb).

Majorana 1 with an aluminium tile beside Majorana 2 with a lead tile and an added InAsSb layer, with the gaps and the time between flips before and after
The two changes Microsoft names, and the numbers the paper reports after them. Diagram: Uni-Sine.

Lead matters because of its energy gap, the energy needed to break up the paired electrons that make a superconductor work. According to the paper, lead's gap is about 1.3 millielectronvolts against about 0.3 for aluminium. The gap that protects the qubit itself, called the topological gap, rose from about 30 to about 70 microelectronvolts by the paper's figures.

What 20 seconds means

The number Microsoft leads with is the parity lifetime: how long, on average, before the parity flips by accident. The paper puts it at 22 seconds, give or take one, from 324 measured intervals, and says some waits between flips ran to minutes. In the earlier aluminium devices it was 1 to 12 milliseconds. One operation on the chip takes about a microsecond, so about 20 million operations now fit between two flips, where 1,000 to 12,000 did before.

A scale from one microsecond to 100 seconds. One operation takes about a microsecond, Majorana 1 lasted 1 to 12 milliseconds and Majorana 2 lasts 22 seconds on average
Each step along the line is ten times longer. Diagram: Uni-Sine, from the paper's numbers.

Why two times better gives 1,000 times longer

It wasn't that we made a thousand x improvement in the underlying properties. It was improvements by a multiple of a few, two to three.
MicrosoftMajorana 2 Explained: The road to 1000x improvement, 16 November 2026source

The video says the link between the two is exponential. Microsoft has not published a formula for it, but a standard simple model shows how that can happen. If a flip needs a burst of heat large enough to cross an energy gap, the time between flips grows as e to the power of the gap divided by the thermal energy:

\tau\propto e^{\Delta/k_BT}τ is the time between flips, Δ the gap, k_B the Boltzmann constant and T the temperature.

At 50 millikelvin, a temperature the paper uses in its own estimates, the thermal energy is about 4.3 microelectronvolts, so a gap of 30 is about 7 times the thermal energy. In the graph, x is how many times bigger the gap becomes and y is how many times longer the wait between flips. A gap 1.5 times bigger gives about 33 times; doubling it gives about 1,000. At the measured 2.3 times, the model gives about 11,000, inside the measured range: 22 seconds is 1,800 to 22,000 times the earlier 1 to 12 milliseconds. This is an illustration, not Microsoft's own calculation.

Times longer between flips, against times bigger gap
11.11.21.31.41.51.61.71.81.92-200200400600800100012001.5 times the gap: about 33 times2 times the gap: about 1,000 times
A gap twice as big makes flips about 1,000 times rarer in this model.Drag to pan, scroll to zoom. x from 1 to 2, y from -211 to 1260.

What physicists dispute

The preprint has not been peer reviewed, and it reports only one of the two kinds of measurement, known as Z and X, that a working qubit needs. Science News reported on 2 June 2026 that physicists were divided. Kartiek Agarwal of Argonne National Laboratory told the magazine the new measurement method was real progress towards showing the states are Majoranas. Henry Legg of the University of St Andrews said the preprint resolves none of the fundamental issues.

Nothing in the presented data proves the existence of a topological qubit or Majoranas in these devices.
Henry LeggUniversity of St Andrewssource
One qubit drawn as an H of two nanowires with an end state at each end, beside two lists: what the paper reports, and what it does not show yet
One qubit is an H of two nanowires. Diagram: Uni-Sine, from the paper and Science News.

The video answers the doubt directly: scepticism, it says, is a valid and important part of science, and some details are proprietary and cannot be shared.

What happens next

Microsoft says it is now targeting 2029 for quantum computers that solve problems nothing else has solved, half its earlier timeline. The video says Majorana 3, 4 and 5 will follow, and that the challenge now is scaling up. DARPA, the US defence research agency, has named Microsoft one of two companies in the final phase of its programme assessing such machines. On arXiv the paper is still listed without a journal.

Guided courseQuantum ComputingQubits, superposition and the computers of tomorrow, from the first principles up.

Sources

Read next
David Gross on stage in front of a slide reading The Future of Physics: What We Don't KnowPhysics · Atoms and particlesGross committee validates OpenAI Yang-Mills existence proofNobel laureate David Gross chairs the committee that checked an AI-written proof that Yang-Mills theory exists. The mass gap, and the prize, stay open.