Top 10 Ikke‑Silicon Computervirksomheder

Fra silicium til nye former for beregning

The computing industry was born when mechanical devices started to perform calculus that was, until then, reserved for the human brain. But it was with vacuum tubes and, later on, transistors that true computers started to be created.

The next revolution was the silicon computer chips, with ever-growing transistor density for ever-growing computation power.

Kilde: Mobile First

Currently, the semiconductor industry is experimenting with increasingly powerful systems to create chips in the 5nm and even 2nm range. This is bringing us ever closer to a problem, as, at one point, using smaller and smaller silicon transistors will not be possible anymore.

One single atom of silicon is a theoretical limit, but practical engineering problems will probably make it happen before that threshold.

So, will computing power stop progressing from here? Probably not.

However, the solution will be to perform computation using entirely new principles. There are actually many potential ways to perform computing without relying on silicon transistors. We can look at the most promising ideas without going into the technical details.

Ikke-silicon halvledere

A semiconductor is a material with the ability to switch between being conductive (transmits electric current, creates a “1” data in binary) or an isolant (blocks electric current, creates a “0” data in binary).

Silicon has been the material of choice for creating semiconductor chips, but plenty of alternatives are now being explored. Any material displaying the property called band gap can be a good candidate.

Kilde: Energy Education

Vanadiumdioxid

For a long time, vanadium dioxide has been seen as a good option to replace silicon. This is because it undergoes a phenomenon known as “undergoing metal-insulator transitions,” which takes only a trilliontedel of a second.

The speed of the metal-insulator transition should allow for faster and smaller electronics compared to classical silicon-based electronics.

Ny forskning har formået at studere vanadiumdioxid aflejret på et substrat af titandioxid.

They also discovered that titanium dioxide can be a semiconductor as well. This discovery could allow for creating neuromorphic chips that could learn at the hardware level, taking inspiration from the brains of living systems with neurons.

Thanks to its very quick Insulator-To-Metal transition, vanadium dioxide with an active substrate of titanium dioxide could be used to create Mott-neuron‑lignende spiksende oscillatorer able to replicate at the hardware level biological neurons.

Grafen

Another good candidate is graphene, a 2D material with extremely high electric conductivity. It is even a potential superconductor and a “wonder material” whose properties are still being discovered in real-time.

You can read more about the first-ever successful efforts to make graphene into a semiconductor material in our article “Grafen‑halvledere – Er de endelig her?”

Organiske materialer

According to a recent discovery, organic material could be forced to form a 2D structure similar to graphene. This could make them as ultra-conductive as graphene, while naturally displaying semiconductor properties, contrary to graphene which has to be “forced to do so”.

You can learn more about this option in “Kan organiske halvledere kombinere fordelene ved grafen & silicium?”

Optimering af halvlederes energiforbrug

An issue with using ever quicker and smaller transistors is the growing power consumption.

An alternative could be to use a technique called “redox gating.” This relies more on a chemical reaction (redox) and could drastically reduce power demand.

If the price of computing starts to rise from power costs more than the chips themselves, this is a solution we might see implemented as well. We explored the latest news on this topic in “Redox‑gating kan føre til nye niveauer af effektivitet i små elektronik”.

Fotonik

Alternative semiconductor materials try to replace silicon. But what if computing was done entirely without using electrons, transistors, and semiconductors?

This is the idea of photonics, looking to perform computing directly with light.

Light is the fastest thing in the universe, so it could be orders of magnitude quicker than silicon and semiconductor-based computing.

In practice, fotonik stadig involvere silicium but could also baseres på krystaller.

Due to light’s wave-like nature, photonics design relies on curves and unique (and somewhat not technologically mature yet) design principles that differ from those used for semiconductors.

Kilde: Synopsis

Kvanteregning

Computing could also be performed by measuring not electric current but the quantum state of particles.

Instead of generating 0 and 1 (no current or current), it uses “quantum bits,” called qubits, where particle data is either 0 AND 1 at once, or 1, or 0.

Because of the fundamental difference in the calculation, quantum computing is not an alternative to “normal” computing but rather a complement.

Standard computing works linearly and struggles with very complex calculations, like climate modeling, cryptography, or the 3D configuration of complex molecules like proteins. And this is precisely the type of calculation that quantum computing is expected to excel at.

So, while maybe not replacing silicon, quantum computers could perform better tasks that were previously almost impossible for silicon chips.

You can read more about the latest news in quantum computing in our article “Den aktuelle tilstand for kvanteregning”.

Biologiske organoider

Our brains are essentially supercomputers, at least when it comes to processes like pattern recognition, language, etc. And very efficient at that, consuming barely a few dozen watts.

En schweizisk startup, FinalSpark, har nu udviklet en 0,5 mm stor kugle (organoider) lavet af 10.000 menneskelige neuroner. And uses it to perform computation. The service will even be accessible through the cloud.

This is a very new field, and it is unclear yet how far it will go. But who knows, maybe one day our self-driving devices will run on neurons instead of chips.

Top 10 Ikke‑Silicon Aktier

1. International Business Machines Corporation

International Business Machines Corporation (IBM) var den ledende kraft bag kommercialiseringen af den første mainframe-computer. Dog er den faldet bagud i produktionsvolumen i forhold til andre teknologigiganter som Apple, TSMC og NVIDIA.

Den er dog i frontlinjen for udviklingen af kvantecomputere. For eksempel udviklede den sin 127‑qubit “Eagle” kvantecomputer, som blev fulgt af et 433‑qubit system kendt som “Osprey”.

Og dette er nu fulgt af “Condor”, en 1.121 superledende qubit kvanteprocessor baseret på cross‑resonance gate‑teknologi, sammen med “Heron”, en kvanteprocessor i spidsen af feltet.

IBM er involveret i de fleste af de andre banebrydende innovationer inden for beregning og halvlederindustrien. Disse inkluderer ledende organiske materialer, neuromorfisk beregning, fotonik, osv.

Til en vis grad er IBM blevet et “patentfirma” med ekspertise i at udvikle nye beregningsmetoder og licensere dem til industrien.

Indtil videre ser det meget beslutsomt ud at holde så mange nøglepatenter i alle de ikke‑silicon beregningsmetoder, den kan få, og dermed gentage sin tidligere succes med at bidrage massivt til udviklingen af halvlederindustrien til den gigant, den er i dag.

2. Microsoft Corporation

Allerede en leder inden for “normale” cloud‑tjenester, Microsoft er en pioner i at tilbyde kvante‑cloud‑tjenester med Azure Quantum.

Det er fuldstændig muligt, at størstedelen af kvanteberegning i fremtiden vil blive udført “fjernbetjent”, ved at stole på cloud‑tjenester som Microsofts, i stedet for direkte adgang til en kvantecomputer.

Dette er især sandsynligt, da de fleste kvante‑applikationer vil blive forsket af biokemikere, materialvidenskabs‑eksperter, klimaforskere og andre specialister uden specifik baggrund i kvanteberegning.

Så at stole på dedikerede fagfolk hos firmaer som IBM, Microsoft eller Google til at håndtere beregningsdelen giver mere mening end at ansætte eller uddanne folk uden erfaring inden for området.

Microsoft‑tjenesten tilbyder “hybrid computing”, der kombinerer kvanteberegning med traditionel cloud‑baseret supercomputer‑service.

Kilde: Microsoft

I stedet for vertikal integration har Microsofts tilgang til kvanteberegning været at etablere partnerskaber med ledere inden for feltet, der dækker praktisk talt alle teknologier, der er mulige for at opnå kvanteberegning, som IonQ (IONQ), Pasqal, Quantinuum, QCI (QUBT), og Rigetti (RGTI).

Kilde: Microsoft

Microsoft etablerede også i slutningen af 2023 et samarbejde med Photonic, et firma der arbejder på at kombinere kvanteberegning og fotonik.

Microsoft har også arbejdet på analoge fotoniske chips til finanssektoren.

Kvanteregning er ikke centralt for Microsofts forretning, i hvert fald for nu. Det er dog en central aktør i sektoren og kan udgøre et “sikrere” aktievalg i forhold til direkte at købe aktier i deres kvante‑partneres børsnoterede selskaber, som QCI eller Rigetti.

3. Alphabet Inc.

Google er meget aktiv inden for kvanteberegning, primært gennem sit Google Quantum AI‑laboratorium og Quantum AI‑campus i Santa Barbara.

Googles kvantecomputer skrev historie i 2019, da Google hævdede at have opnået “kvantesupremati” med sin Sycamore‑maskine, som udførte en beregning på 200 sekunder, der ville have taget en konventionel supercomputer 10.000 år.

Men måske er Googles største bidrag inden for software, en aktivitet hvor de har langt bedre resultater end hardware (søgning, G‑Suite, Android osv.). Googles Quantum AI stiller allerede en suite af software til rådighed for at hjælpe forskere med at udvikle kvantealgoritmer.

Google er også en aktiv støtter af fotonik‑virksomheder som Lightmatter.

Google vil sandsynligvis blive et af de firmaer, der sætter standarderne for kvante‑software &‑programmering, og giver en privilegeret plads til at påvirke, hvor feltet vil udvikle sig i fremtiden. Deres stærke netværk og VC‑aktivitet vil også sandsynligvis give dem en plads i enhver anden ikke‑silicon‑baseret form for beregning.

4. Intel

Intel er en stor chipproducent og ser ud til at udnytte denne styrke i kvante‑computing arenaen.

De udgav for nylig “Tunnel Falls”, den “mest avancerede silicium‑spin‑qubit‑chip”. Det bemærkelsesværdige er, at den ikke er en prototype, men en chip bygget i skala med 95 % udbytte på waferen og spændings‑ensartethed. Dette baner vejen for masseproduktion af kvante‑chips, noget der indtil nu har været svært i en spirende og hurtigt foranderlig industri.

Kilde: Intel

Trofast over for sine rødder udvikler Intel også software til at udnytte sine chips, med udgivelsen af Intel Quantum SDK. Dette giver retningslinjer for programmører til at udvikle software til kvanteberegning kompatibel med Intels kvantechip‑design, som historisk har været en meget stærk & profitabel forretningsmæssig fæstning for Intels konventionelle chip‑forretning.

Kilde: Intel

Ankomsten af skalerbar kvante‑chip‑produktion kan blive så revolutionerende for industrien som enhver anden teknisk videnskabelig gennembrud, sænke omkostninger og sætte fælles programmeringsstandarder og chip‑arkitekturer.

I slutningen af 2023 besluttede Intel at afstå fra sin fotonik‑forretning til Jabil (JBL).

Samlet set gør Intel fremskridt inden for kvanteberegning og ser ud til at have en klar strategi om at fokusere på dette emne frem for fotonik og andre alternativer.

5. Nvidia

The leading manufacturer of graphic cards and, more recently, cryptocurrency mining rigs and AI chips has now truly evolved from a PC parts manufacturer to one of the global tech giants.

Nvidia is also active in the quantum computing space, with its NVIDIA DGX Quantum combining normal chips and quantum computing using the newly open-sourced CUDA quantum software platform.

Kilde: Nvidia

Looking to reinforce its lead in AI, Nvidia has also released its QuantumX-800 for AI-optimized networking in data centers.

When it comes to photonics, Nvidia has forged a partnership with TSMC and Broadcom.  It will look to create a single module through co-packaged optics (CPO) integrating classical silicon chips and photonics.

Overall, Nvidia’s success is tightly linked to the current AI boom, and quantum computing and photonics come second. However, it will benefit from the growth of these sectors as well and seems to be holding on to stay in the race.

6. Quantinuum / Honeywell

Quantinuum er resultatet af fusionen mellem Honeywell Quantum Solutions og Cambridge Quantum (og, som nævnt, en partner af Microsofts kvante‑cloud‑computing).

Quantinuum ser for nuværende ud til at fokusere på segmenter, der er mindre udforskede af andre kvante‑computingsystemer, især finansielle og forsyningskæde‑relaterede analyser, gennem sin Quantum Monte Carlo Integration (QMCI) motor, lanceret i september 2023.

QMCI applies to problems that have no analytic solution, such as pricing financial derivatives or simulating the results of high-energy particle physics experiments, and promises computational advances across business, energy, supply chain logistics, and other sectors.

Ligesom Microsoft er kvanteberegning ikke den centrale del af Honeywells forretning, som er mere centreret omkring produkter inden for luftfart, automation og specialkemikalier og materialer.

Dog, idet alle disse forretningssegmenter kunne drage fordel af kvanteberegning, er det ikke svært at se forretningscasen for Honeywell at blive involveret.

Dette gør Honeywell både til en leverandør af kvante‑computing‑tjenester og til et af de virksomheder, der kan drage fordel af anvendelsen af kvantecomputere i virkelige forretningssager, noget som integrationen af Quantinuum i koncernen bør hjælpe med at fremme hurtigere end dets industrielle konkurrenter.

7. Synopsys

Ethvert fotonisk system skal integreres så gnidningsløst som muligt med silicium‑systemer, i det mindste i starten. Synopsys kan hjælpe med dette.

The company is a specialist in silicon design and verification, meaning its software is used to design new chips, including ultra‑advanced 5nm chips and below.

The company also offers software for photonics described as “The industry’s only seamless design flow for photonic devices, systems, and integrated circuits”. This allows for handling the design and simulation of new photonics devices.

Kilde: Synopsis

The company has also developed a joint venture with Juniper Network to create OpenLight, a photonics company using indium phosphide.

8. Juniper Network

Juniper hævder at tilbyde den #1 cloud‑native trådløse løsning og det eneste AI‑drevne Wi‑Fi‑netværk. Dette placerer dem direkte i konkurrence med ældre og mere etablerede giganter som Cisco. Junipers teknologi, Juniper Mist, siges at være mere skalerbar, fleksibel og bedre til anomali‑detektion end Ciscos tilsvarende tilbud.

The company’s solutions rely heavily on AI, with its AI engine “Marvis” used at all network levels, from user to data center.

Kilde: Juniper

Regarding security, Juniper also shows outstanding results on firewalls, threat defense, and defense against exploits, outperforming most vendors like Fortinet, Palo Alto, Zscaler, etc.

Juniper also offers Photonic Integrated Circuits (PICs), which are currently mostly used for data transmission and sensors. They are expected to be an integral part of future photonics-based computers.

Kilde: Synopsis

9. Rigetti Computing, Inc.

Riggeti er et kvante‑computing‑firma, “ejer kritisk IP til vores banebrydende multi‑chip‑processor og den hybride kvante‑klassiske tilgang, der er blevet den dominerende kvante‑computing‑arkitektur.”.

The company is integrating all the steps needed for quantum computing, from chip design and manufacturing up to cloud delivery of the computing power.

Kilde: Rigetti

The company focuses less on adding as many qubits as possible (as giants like Intel are doing) but on perfecting their existing product and achieving a very high level of fidelity and speed, making it a more reliable commercial product.

Its latest iteration, the 84-qubit Ankaa-3, is expected to be revealed in the second half of 2024. Based on the Ankaa concept, the company aims for a 336+ qubit system in the long run.

Kilde: Rigetti

In december 2023, Rigetti started sales of the 9-qubit system Novera, a “mini quantum computer” selling for “only” $900,000 and a 4-6 weeks delivery.

The first clients included Fermilab’s SQMS Center, the Air Force Research Lab, and Horizon Quantum Computing.

The company announced in spring 2024 that it would join the Russel 3000 Index.

10. IPG Photonics

IPG er en laserproducent, der producerer praktisk talt alle typer lasere, inklusive fiber, diode, UV, and deep UV lasers. With 6,200 employees, it ships 42,000+ laser devices per year.

Its specialty is in fiber lasers, with high levels of precision and the ability to do laser pulses as short as a femtosecond (one quadrillionth of a second).

IPG lasers are currently used for:

• Avancerede videnskabelige anvendelser (spectroscopy, microscopy, interferometry, optical trapping, etc.)
• Fremstilling af batterier og elektriske motorer til el‑biler
• Materialebehandling, especially metal cutting, carving, cleaning, and laser 3D printing.
• Laser‑mikrobehandling, where lasers are used to create ultra-small structures.

While advances in photonics chips will be required to create entirely photonics-based computers, we already know that it will integrate a lot of a specific and already common component: lasers.

The light for photonic computing needs to be based on very stable light emitted by the laser. So leaders in the laser industry, like IPG, would benefit from a boom in laser demand from the semiconductor industry switching progressively to photonics.

And in that nascent segment, ultra-short laser impulses can be turned into ultra-fast computing power.