Monopoly is always the best business.
The singularity is approaching, and the world is changing at an ever - accelerating pace.
At the GTC conference in October, Jensen Huang held up a demonstration diagram of the NVQ Link interconnection architecture and started making predictions again:
It took 30 years for GPUs to change the world. Quantum computing may not take that long.
This statement is most likely true. The development of quantum computing has never been linear.
Looking at the data, the annual investment in quantum computing in the industry has increased from $1 billion in 2020 to $5 billion in 2025... Who knows, maybe one day there will be a sudden technological breakthrough, and the entire industry will explode.
So, Jensen Huang, with an ecological alliance of 17 quantum enterprises and 8 US national laboratories, is determined to be the "connector" between quantum and classical computing.
Why the connection?
The official line is "connecting computing power with the future." The real reason, of course, is to grab a ticket to the future. In the next five years, continue to be the "shovel seller" and keep making easy money.
And how to make the connection?
Two words: monopoly and ecosystem.
Technology is always the most powerful productive force.
And monopoly is always the best business.
01
From "Theoretical Toy" to "Computing Power Beast"
Quantum computing is not hype but a slowly - burning revolution that is finally heating up.
As early as 1981, physicist Richard Feynman pointed out the inherent limitations of classical computers in simulating quantum systems during a speech at the California Institute of Technology and first proposed the concept of a quantum computer.
In 1994, mathematician Peter Shor invented a quantum algorithm for factoring large prime numbers, which for the first time made the industry realize: this thing can crack the bank's encryption system!
However, at that time, quantum bits were still very "fragile." The decoherence time couldn't even last for 1 second, and it could only be a decorative item in the laboratory.
Thanks to the development of hardware, it wasn't until 2012 that Google started investing heavily in superconducting quantum computing. In 2019, it introduced the 53 - qubit "Sycamore," claiming to have completed a task that would take a classical supercomputer 10,000 years to finish, and for the first time demonstrated "quantum supremacy."
Subsequently, in 2020, IBM launched the 127 - qubit "Eagle" processor, and in 2021, it was upgraded to the 433 - qubit "Condor," refreshing the record at a doubling rate every year. The R & D investment soared from $320 million in 2018 to $870 million in 2021.
Almost at the same time, in 2020, China's "Jiuzhang" photonic quantum computer used 76 photons to boost the computing speed to 100 trillion times that of a classical supercomputer; in 2021, the 66 - qubit superconducting chip of "Zuchongzhi II" increased the complexity by six orders of magnitude.
At this time, the core of the competition was simple: the more qubits, the more powerful.
But soon, people found that just increasing the number of qubits was useless. Fidelity (the accuracy of quantum operations) was the key.
Differences between quantum computing and classical computing, Source: CB Insights
So, from 2022 to 2024, global tech giants quickly embarked on five technological routes.
1. Superconducting route: Most similar to traditional chips, IBM, Google, and the University of Science and Technology of China are betting on it. In 2024, Guodun Quantum delivered a 176 - qubit whole - machine, and the dilution refrigerator was completely domesticated.
2. Photonic quantum route: Not afraid of high temperatures. China's "Tianguang Quantum Brain 550W" can operate at room temperature, leading the world with a scale of 550 qubits.
3. Ion trap route: It has the highest fidelity (up to 99.99%), but it's difficult to increase the number of qubits.
4. Neutral atoms: Low cost. Google uses lasers to manipulate atoms for computing.
5. Silicon semiconductors: Intel's favorite. It can be compatible with existing chip production lines, but unfortunately, the progress is the slowest.
Although there are huge differences in the routes, and it's still far from the stage of making money through implementation.
However, the global market, especially the markets in China and the United States, has already started to seize application scenarios.
Logistics: JD.com uses quantum algorithms to optimize warehousing. Test data shows that the cost has been reduced by 15%, which means saving 2 million yuan per large warehouse per year.
Meteorology: The Anhui Meteorological Bureau cooperates with China Telecom Quantum. Using the "Tianyan" platform to predict rainfall, the accuracy is 8% higher than that of the classical model.
Medicine: The Hefei Institute of Quantum Medicine uses quantum computing to simulate protein structures, shortening the new drug R & D cycle from 10 years to 6 years.
Finance: Goldman Sachs uses IBM's quantum cloud for option pricing, and the computing speed is three times faster than that of the traditional model.
...
The latest progress in the commercialization of quantum computing, Source: Guojin Securities
At this moment, the industry has finally reached a consensus: Quantum computers can't work alone. They have to team up with GPUs and supercomputers.
The crucial turning point is in 2025, that is, this year: IBM says it will achieve fault - tolerant computing with 200 logical qubits by 2029.
Nvidia's NVQ Link has become the key hub.
Although quantum computing can break through the limits of classical computing power, it still highly depends on classical computing power for support in data pre - processing, error correction, result analysis and other aspects.
By deeply integrating GPUs with quantum processors (QPUs) through NVQ Link to build a hybrid computing architecture of "quantum + GPU", it is equivalent to providing a computing power operating system for all quantum hardware manufacturers.
Letting the AI model filter the data first and then sending it to the quantum computer for calculation is like assigning several assistants to the quantum computer, increasing the efficiency by 10 times.
Quantum + AI is undoubtedly the core growth point of the industry.
Nvidia's cuQuantum software can efficiently simulate quantum circuits, and the quantum large - scale model developed in cooperation with OpenAI can improve the stability of quantum bits by 40%.
This "AI optimizes quantum control, and quantum empowers AI algorithms" two - way cycle allows Nvidia to transfer its technological accumulation and customer resources in the AI field to the quantum track, forming a collaborative barrier that is difficult to replicate.
Even more impressively, Nvidia has brought in 17 quantum enterprises, from Rigetti in superconducting technology to IonQ in ion trap technology, basically covering the mainstream technological routes. It has also jointly established the "Accelerated Quantum Research Center" with 9 national laboratories under the US Department of Energy.
Nvidia's cooperation in the field of quantum computing, Source: Minsheng Securities
Jensen Huang said bluntly at the GTC conference: "In the future, every Nvidia GPU supercomputer will be hybrid," which means that the computing power demand in the quantum era will still be realized through GPUs, thus extending its existing AI computing power advantage to the next - generation computing paradigm.
To put it simply, Jensen Huang wants to strengthen the positioning of the "shovel seller" role.
Or rather, continue to occupy the monopolistic ecological niche.
02
Three - Step Monopoly Strategy
The current industry consensus is that quantum computing will develop in three steps, and the giants have already included the dominance of each step in their plans, with a clear rhythm and definite goals.
In the short term (3 - 5 years), reap the dividends of equipment and software.
This is the most certain opportunity, and Nvidia and IBM have taken the lead.
Hybrid computing: The market scale of Nvidia's NVQ Link ecosystem is expected to exceed $10 billion in 2027, three times larger than the pure quantum computer market. The chip interface licensing fee alone can reach $200,000 per unit.
Software charging: IBM's Qiskit plans to launch an enterprise version in 2026, with an estimated annual fee of $100,000 per customer. Based on the current scale of 12,000 enterprise users, the annual revenue can reach $1.2 billion.
Equipment iteration: Google's next - generation superconducting chip "Sequoia" is expected to be mass - produced in 2026, with the number of qubits reaching 256, the error - correction code distance increasing to 9, and the error rate decreasing by another 50%.
During this period, the global quantum computing market scale will increase from $5.04 billion in 2024 to $20 - 30 billion in 2028, of which the hardware and software related to hybrid computing account for more than 70%; the domestic market scale is expected to reach 3 - 5 billion yuan, mainly driven by the growth of superconducting quantum processors.
The proportion of Nvidia's quantum - related business revenue will increase from 8% in 2025 to more than 15% in 2028.
Nvidia's quantum computer architecture diagram, Source: Bank of Communications International Securities
In the medium term (5 - 10 years), monopolize high - value scenarios.
The key node is in 2029 when IBM will launch a 200 - logical - qubit fault - tolerant computer. At that time, the giants will fully control the high - value fields.
Financial security: IBM has jointly developed a quantum encryption communication system with JPMorgan Chase and plans to launch a "quantum security suite" in 2030, charging a protection fee of 0.3% of the annual revenue from global banks.
AI training: The quantum large - scale models developed by Google and OpenAI are expected to be commercialized in 2030, with a training speed 1000 times faster than that of deep learning, and the cost of a single model training will be reduced from $12 million to $800,000.
Material R & D: Nvidia has jointly built a quantum material simulation platform with Dow Chemical, which can accelerate the R & D of new energy battery materials, aiming to occupy 60% of the global high - end material R & D computing power market.
It is estimated that by 2030, the global market scale will reach $50 - 80 billion, and the domestic market will reach 30 - 50 billion yuan.
Among them, the market scale of quantum communication is 2 - 5 times that of quantum computing, mainly relying on investments from financial institutions and state - owned enterprises.
The proportion of software and services will exceed that of hardware for the first time. The annual fee revenue of IBM's Qiskit enterprise version is expected to exceed $3 billion, and the scale of Nvidia's hybrid computing ecosystem will reach $50 billion.
In the long term (10 - 15 years), directly rewrite industry rules.
When the number of quantum bits exceeds 10,000 and the fault - tolerance rate reaches 99.999%, the high - tech field will witness three major changes.
First, scientific research: Google plans to build a "quantum universe simulator" for cutting - edge research such as black hole collisions and dark matter, charging scientific research institutions a computing power fee of $100,000 per hour.
Second, industrial disruption: IBM's general - purpose quantum computer will replace traditional supercomputers, reducing the energy consumption of data centers by 90%, aiming to occupy 75% of the global supercomputer market share.
Third, standard setting: The "Quantum Industry Alliance" established by Nvidia in cooperation with IBM and Google has formulated 17 quantum interface and security standards and submitted them to the International Organization for Standardization for review.
According to BCG's prediction, the global quantum computing market scale will reach $85 billion in 2040, of which software and services account for more than 60%; the domestic market is expected to exceed 300 billion yuan, and the quantum - classical hybrid supercomputing market will account for 45%.
The computing power fee of $100,000 per hour for Google's "quantum universe simulator" will become the pricing benchmark for high - end scientific research services, and IBM's general - purpose quantum computer will occupy 75% of the global supercomputer market share.
