04/10/2023

Author - Kate Prebble

Investing in Quantum Computing: The Second Wave

Investing in Quantum Computing: The Second Wave

If you used a supercomputer in 1950 you would be using vacuum tubes. Relatively easy to make, large, power hungry and simple, valve systems were code-cracking, radar-driving, missile-testing supercomputers of their day. Valves were the first wave of computer processors and many engineers and investors got wealthy from bringing them to market. Suppliers of tube technology like Raytheon became household names. During World War II and in its aftermath, demand for more processing power and the commercial potential of large-scale computing services in a variety of industries became apparent.

By 1950 the need for bigger rooms and more energy to house tube-driven computers made clear a new type of processor was needed. Within a decade the underlying technology had changed. By 1960 there was not a single computer service coming to market with Valves. The base technology of the last thirty years had been replaced within a decade by a new processing standard - the transistor. How best to build transistors and integrated circuits that could scale took a few years to sort out.

The silicon platform for transistors is so dominant today, that is hard to believe other approaches were considered, tested and built in the decade after 1950. For the first six years of the decade all transistors were made with germanium. Teams globally competed to build transistors in a variety of forms – each claiming to be able to provide a bridge to mass computing, and thousands of processor ‘Bits’. Germanium soon proved to be a dead end – although much easier to work with than Silicon and able to handle high frequency operation, germanium devices had far worse leakage currents in the ‘off’ position, an anathema for computer logic.

They were also restricted to tight temperature ranges for operation which limited their use in rugged operations. In 1954, chemist Morris Tanenbaum at Bell Labs fashioned the first silicon transistor, but did not pursue full commercial application. Texas instruments swept in and claimed credit for the breakthrough a few months later. In 1954 small start-up led by GK Teal, supported by Texas Instruments, created a sensation when he announced that silicon transistors were in production and available for sale.

In 1959 Bell Labs was back in the race again – engineers Atalla and Khang working with silicon had a breakthrough. They overcame the "surface states" that blocked electric fields from penetrating into the semiconductor material. Investigating thermally grown silicon-dioxide layers, they found these states could be markedly reduced at the interface between the silicon and its oxide in a sandwich comprising layers of metal (M - gate), oxide (O - insulation), and silicon (S - semiconductor) - thus the name MOSFET entered the world.

Silicon worked, it could scale, it was simple and relatively easy to manufacture. Best of all it provided a pathway to millions of bits for computers. Moore’s Law was born. The first wave of computer technology was over. Valves were yesterday’s hero; the second wave had begun. By the 1960s silicon had become the industry’s preferred semiconductor material, and the core of all computing platforms since.

This shift between simple, large but expensive systems is happening again in 2023. This time it is quantum computers that are about to experience a silicon revolution. While history does not repeat – it sure does rhyme. Just as vacuum and valve technologies were built in the 40s and 50s by corporations like Raytheon and IBM, so Superconducting Quantum Computers are being built by IBM and Google today. Both companies using huge dilution refrigerators and lots of power to drive physically large qubits (the fundamental switch within a quantum computer – analogous to a bit in classical computing). IBM and Google's showcase images of quantum computers of a few hundred qubits housed in warehouse freezer parks – eerily similar to the room sized vacuum computers shown in photos from the post-war period. You can access these quantum computers today via cloud services with run time costs many thousands of dollars an hour. Money is being made.

Similar also is the way in which start-ups, trading on the hype and promise of quantum computing are getting investors to support non-silicon approaches to making quantum computers – PsiQuantum and Xanadu --  now valued at well over a billion dollars each employ photonics to generate qubits – while smaller than superconducting qubits, and able to operate using light channels etched into silicon, photonic qubit chips still work at millimetre scale, and need power hungry fridges and lasers to drive them. Large scale plants will be required to house these systems if they are ever going to get to a million-qubit scale service (the generally accepted benchmark for a useful quantum computer).

Some labs are even experimenting with germanium again. This first wave of quantum computing is maybe coming to an end. There is a new technology entering the field. It operates at the nanometre level, is power efficient, scalable, and strangely familiar. Silicon, and specifically electron spin on standard silicon has just been proved. The proof points suggest existing MOSFET systems can be used to build qubit processors at the nanometre scales modern classical chips are fabricated at in the billions each year.

A familiar name seems set to lead this next wave in quantum computing. Intel announced on April 14, 2022 that it had created the first ‘silicon qubits at scale’ at Intel’s D1 manufacturing plant in Hillsboro, Oregon. The result is a process that can fabricate more than 10,000 arrays with several silicon-spin qubits on a single wafer with 95% yield. This paper in Nature shows that quantum chips can be produced alongside conventional chips in the same industrial manufacturing facility. Moore’s law might just be about to kick in for Quantum Computers too.

So why is Intel alone at the forefront of this second wave of quantum computers? Just like Texas Instruments back in the sixties they are not the only game in town. IBM and Google have already got large scale defence clients for their superconducting quantum systems, and are busy charging industrial and enterprise clients hefty consulting fees to get their organisations ‘Quantum Ready’. With their deep pockets it will be relatively easy to shift from Wave-One to Wave-Two by buying quantum chips from a fabricator like Intel when they scale up production. Other early stage investors in competing qubit technologies are standing back – having seen good balance sheet growth (but little real revenue) from their early stage on non-silicon quantum processors they are watching from the shoreline.

While Intel steals the PR headlines, there is a competitor who owns and has fabricated silicon qubits with better results than Intel across two years prior. Also cited in nature. A small start-up from down under called Diraq launched in 2022 and have already wooed the ex-CEO of HRL to join the board. Like Intel, Diraq promises billions of qubits on a single chip, and would seem to have a stronger engineering team. Led by heavily cited Australian Laureate Professor Andrew Dzurak, this team have 20 years of IP sponsored to date by the same type of backers that invested in Raytheon back in the 40s – the US Army Research Office.

Diraq has achieved ‘hot-qubits’ providing for electron spins in silicon at temperatures much higher than those used by superconducting and photonic systems – this in turn will allow operation wusing standard colling techniques rather than the huge dilution refrigerators used by all major players today. Intel published in Nature a result of qubit operations above 1Kelvin in April 2020. https://www.nature.com/articles/s41586-020-2170-7. Dzurak group also published this result in Nature achieved in April 2020, but showed operation at temperatures above 1.5K https://www.nature.com/articles/s41586-020-2171-6. This sort of breakthrough will see the current forecasts of quantum computing commercial use collapse firmly into this decade.

Just as the silicon revolution ignited an explosion of computer brand names, newcomers and plucky start-ups in the 1970a (Apple, Compaq, Atari, Compaq); so, the next five years will likely see billions made by new quantum brands. The initial hype of Quantum may be receding, the second wave of quantum technology is just hitting the beach.