The way in which the co-founder and chief executive officer of Nvidia, Jensen Huang, sees it, the role of his business in the field of quantum computer still emerging is no different from that of other industries where AI and accelerated IT roles. Nvidia is a catalyst and facilitator. It makes things happen.
“NVIDIA does not make quantum computers, but we are devoting ourselves to the creation of accelerated computer batteries to allow quantum computers,” said Huang during the inaugural day of the GPU manufacturer at the end of his recent week -long GPU technical conference in San Jose, California. “We do the same with autonomous cars. As you know, Nvidia is probably more integrated into the world of cars and autonomous vehicles and we work with everyone in one way or another to advance autonomous vehicles, and yet we do not build cars. ”
The same can be said of robotics – Nvidia has its hands in industry, offering models and libraries, but it does not build robots.
“We do not build quantum computers, and yet we are Deeply integrated into the quantum IT industry And we create libraries, “said Huang.” Cuda-Q is the programming model for the accelerated hybrid quantum. We have Cuquantum, libraries that help you simulate quantum circuits, and DGX-VENTUM, to make an error correction in quantum computers. We associate ourselves with them, we support them, we help them. »»
As he noted, Nvidia was active in quantum space For several years, a partnership with an increasing number of myriades of suppliers of the industry and to lend its AI and an accelerated IT expertise for reference architectures, libraries and tasks such as the design of fleas and systems.
Huang also obtained a lesson earlier this year on the way in which industry can be volatile, when he declared at the Mansa Public Electronics show in January that useful quantum systems were still 15 years to 30 years, sending quantum stocks plummeted up to 60%.
The first quantum day of the company – there will be a second next year – partly a means to appease hard sensations about this shift and this Blip and to allow Huang a chance to mea culpa: “This is the first event in history where a CEO of the company invites all the guests to explain why he was wrong.”
A quantum research center
The quantum day was not the only place of the quantum computing in GTC. Sprinkled among the news and discussions on advances in AI and Nvidia’s latest effortsGPU alignments extended to product roadmaps to AI initiatives for the company, have been the next steps for its quantum plans. The key of them is that Nvidia plans to build a research center for the construction of advanced technologies for quantum computer science.
The NVIDIA Acceleraiteated Quantum Research Center based in Boston, or NVAQC – Huang said that the company had chosen the city to be closer to research universities like Harvard its quantum initiative in the MIT engineering systems group, from Quantum and Quantum research with NVIDIA and other technologies and meet challenges such as Correction of errors in qubits.
The center will house a NVIDIA GB200 NVL72 Rackscale Supercalculator This will include 576 GPU Blackwell-that is to say four racks, or half a superpod-and will use the company Infiniband Quantum-2 of the company to wedge racks together. It will include quantum architecture DGX and will allow researchers to orchestrate what will be quantum hybrid-classic systems.
“While quantum computers continue to develop, they will Integrate with AI Supervisors To train accelerated quantum supervisors capable of solving some of the most difficult problems in the world, “wrote Nicholas Harrigan, director of product marketing for Quantum at Nvidia.” Integration of quantum treatment units (QPU) in AI supercalculators is the key to developing new applications, helping to unlock breach and control of peripherals. »»
Many partners
There have also been a number of announcements in and around the GTC in terms of quantum IT companies to take advantage of Nvidia technologies. Pasqal said that he integrates his quedes with neutral atoms and his cloud platform with the quantum platform CUDA-Q open to give organizations more tools to develop and execute quantum applications. Last week, Pascal announced It will build Euroqcs-ItalyA quantum computer that will be hosted in Italy. The French company has obtained a contract for the supply of the High Performance joint European company (EuroHPC JU).
Seeqc said that it had created the first digital interface to connect quantum chips to GPUs to allow an error correction in real time. The technology includes the quantum product of SEEQC single flow integrating with Cuda-Q.
Quantum rings have integrated its quantum circuit simulation technology with Cuda-Q, Quera worked with NVIDIA to develop an AI decoder based on a transformer for error correction via CUDA-Q and the data from the quantum chip with neutral atoms of Quera, and Q-CTRL associated with NVIDIA and quantum circuits to reduce calculation costs that have a deletion of errors using its fire opagen software.
Breakthroughs are underway
The quantum day of Nvidia came in the context of several progress in the succession of such critical challenges such as error correction, which is difficult to achieve due to the fragility of qubits, which can easily be torn by light, vibrations and other disturbances. However, quantum heavy goods vehicles Google,, MicrosoftAnd Amazon web services Each methods described that they have developed which, according to them, approach the obstacle to the correction of errors.
In addition, researchers led by a professor at Queen Mary University in London suggest that the superconduing quantum systems could operate at room temperatureWhich is important because it was assumed that these computers should operate at very cold temperatures, which could limit their use.
Such progress has contributed to rolling the waters of the current debate on the estimated arrival time for useful and tolerant quantum computer science in which Huang entered in January. When the company’s majorana quantum chip was announced in February, Microsoft managers said that it could reduce time to useful quantum computer science from decades to years.
During a large round table with managers of a certain number of quantum companies, Huang asked this question, some at a high level noting that this could be a decade or more, but added a certain number of warnings, including the probability that quantum systems operate side by side with conventional supercompants, essentially taking these workloads that the supercomputers cannot manage. Alan Baratz, CEO of D -Wave – who has his own conference this week and his own announcements – noted several times during the round table that his business already has his Recuit quantum computers on the market.
However, several people have noted that applications already work on some of their systems and the number of qubits they grow there.
Peter Chapman, executive president and former CEO of IONQ, said that the industry was to mature again, which will include the adoption of two or three methods – which currently include superconductive, trapped, neutral atoms, photonic qubits and topological qubits, among others – and to browse mergers and other consolidation.
There are more common points among the different quantum companies than expected and you can expect it and there is hope that there will be more sharing and work together in the future to stimulate the development of quantum computer science.
“New ways of building qubits are found every day, but I think that in the past few years, we will start to merge to probably two, three different approaches and that some of us will probably meet because we are essentially sharing underlying technology,” said Chapman, noting that it takes a lot of time to spend a startup to a Nvidia stature company. “It is good for quantum industry if it will be ten to fifteen years to arrive where Nvidia and all the other giants.”
That said, he added: “It’s just not going to start then. It starts today.”
