What should I do if AI consumes too much energy?Microsoft's answer is to usenuclear energy, a recruitment information was recently posted on its official website to recruit a nuclear technology expert. The full name of this specialist position is "Chief Project Manager, Nuclear Technology", who is primarily responsible for refining and implementing the global energy strategy for small modular reactors (SMRs) and microreactors, and will "lead the technology assessment of SMR and microreactor integration to power the data centers where Microsoft Cloud and AI reside."
Source: Microsoft
In other words, Microsoft will use nuclear power, a new type of energy, to provide the energy needed for the data center where Microsoft cloud and artificial intelligence are located.
In fact, as AI begins to explode, computing power has become the basis for AI development, but in the final analysis,The bottleneck of computing power is still electricity.
Digital Information World once released a report that the energy consumption generated by data centers for training AI models is three times that of conventional cloud work. By 2030, the power demand of U.S. data centers will grow at an annual rate of about 10%.
Obviously, ordinary energy can no longer keep up with the demand for AI computing power.Nuclear energy with higher energy density is beginning to take full advantage.
01
Accelerate the injection of nuclear energy
Using nuclear energy to power its own systems is something Microsoft doesn't think about for a day or two.
In May of this year, Microsoft suddenly announced a news, and nuclear fusion start-up companyHelionEnergySigned a power purchase agreement to purchase from Helion its nuclear fusion power plant in 202850 MWelectricity.
Source: HelionEnergy
What’s even more interesting is that Helion’s early important investor was Sam Altman, CEO of OpenAI, who invested$375 million, it’s himMy biggest investment to date.
It seems that Microsoft is not the only company that wants to connect AI and nuclear energy.
According to the job description on Microsoft’s official website, this nuclear energy technology expert is not someone you can just become.
Requires more than six years of experience in the nuclear industry, engineering, energy markets or related positions; more than three years of experience managing cross-functional or cross-team projects; more than three years of experience working in nuclear energy and related technologies and/or nuclear regulatory matters.
Microsoft will give priority to candidates who also have a master's degree in business, engineering, science, economics, commerce, and other such fields and have more than four years of experience in engineering, energy markets, grid modeling, analysis, or related positions.
Their work includes: helping build and deliver the networks that power the world's largest online services, leading projects on all aspects of infrastructure, maintainingSMR and microreactorsComprehensive roadmap for integration.
In addition, it is also necessary to develop a clear set of performance and standards to introduce and manage nuclear energy technology partners, establish long-term strategic relationships with key suppliers and industry groups, and ensure the technical feasibility and optimal integration of SMRs and microreactors.
mentioned in the job descriptionSMR (Small Modular Nuclear Reactor), is the key to Microsoft's injection of nuclear energy this time.
SMR is a new type of nuclear power generation technology that is generally smaller, more flexible, safer than traditional large nuclear reactors, and of course more convenient to build and operate.
Source: HelionEnergy
This kind of reactor can be mass-produced in a factory and transported to a designated location for installation. It does not rely on building structure and does not require a lot of infrastructure like conventional nuclear reactors.
This also means that the SMR model can bring nuclear energy into commercial power supply faster. Microsoft may use SMR to enter the nuclear energy market.
The next generation of nuclear reactor technology will largely be implemented in small nuclear reactors.
It is true that SMR also faces some challenges, such as regulatory and site safety issues, but this solution is already a practical and feasible nuclear energy solution at present.
02
“Power-hungry” AI needs nuclear power
Consulting organization Tirias Research has predicted that by 2028, data center power consumption willNearly 4250 MW, a 212-fold increase from 2023, and the total cost of data center infrastructure plus operating costs may exceed US$76 billion.
In other words, all the innovations brought about by AI are based onat the high cost of processing performance and power consumption, behind every AI training, a lot of energy is consumed.
Source: HelionEnergy
The development of AI may be unlimited, but it will ultimately be limited by computing power and energy consumption.
What kind of energy supply can ensure increasingly powerful AI training capabilities?
Sam Altman, the founder of OpenAI, gave the answer. He believes that although solar energy and energy storage can also achieve this goal,Nuclear energy is the best way.
Source: CNBC
He also posted a lot of discussions about nuclear energy on the X (Twitter) platform, and also emphasized that nuclear energy has major advantages.
Facts have proved that Altmann is not just talking casually. He has already bet heavily on two nuclear energy start-up companies, among whichHelion works on nuclear fusion, has signed a power purchase agreement with Microsoft and completed its first nuclear power commercial order.Oklo is dedicated to nuclear fission, has been listed through backdoor transactions.
He believes that the relationship between AI and nuclear energy is more likecomplementary relationship, Nuclear energy can achieve AI, and vice versa.
Altmann has said that the two causes of artificial intelligence and nuclear energy in which he is devoted are closely linked. Future artificial intelligence systems will consume large amounts of energy, and nuclear fission and nuclear fusion can help provide large amounts of cheap electricity.
And with the advancement of artificial intelligence, AI will also help improve the design level of nuclear energy systems.
In addition, nuclear energy has always been considered the most ideal ultimate energy source. It can produce nearly unlimited energy and does not produce greenhouse gases.
Source: HelionEnergy
Therefore, using nuclear energy as a power source not only reduces dependence on fossil fuels, but also reduces the carbon footprint of data centers.
However, due to the huge technical difficulties, the answer to the question of when commercial nuclear fusion power generation will be realized is still unclear. Previously, there have been rumors in the industry that nuclear fusion is "close to success".There are still 50 years left forever” argument.
However, with the emergence of small nuclear reactors such as SMR, nuclear energy has begun to advance in tandem with AI. The signing of a power purchase agreement between Microsoft and Helion Energy is a milestone event.
03
The track gradually takes off
Nuclear fission refers to the splitting of atoms and the release of energy, while nuclear fusion is a process that imitates the principle of the sun and releases huge energy by merging two lighter nuclei into one heavier nucleus.
Nuclear fusion produces more energy - four times that of nuclear fission reactions.
Source: HelionEnergy
Theoretically, with just a few grams of reactants, it is possible to produce one terajoule of energy, which is almost as much energy as a person would need in 60 years.
"Nuclear energy is vital, and radiation risks have been exaggerated." Musk has personally supported nuclear energy. Earlier, many billionaires, including Bill Gates, Jeff Bezos and Peter Thiel, spent real money to support the new generation of nuclear energy companies.
As early as 2021, nuclear fusion start-up Commonwealth Fusion Systems (CFS) announced the completion of a US$1.8 billion Series B financing. At that time, it was the largest single financing in the world's controllable nuclear fusion field.
The investor behind this investment is the founder of Microsoftbill gates, and giants such as Tiger Global Management and Google are also involved.
CFS was founded in 2017 and was born at the Massachusetts Institute of Technology in the United States. They plan to build a fusion reactor SPARC in the United States, covering an area of nearly 47 acres, and are expected to achieve commercial nuclear fusion power generation in 2025.
According to data from research firm PitchBook, venture institutions poured $34.5 billion into nuclear energy startups in 2021 alone, setting a record high and exceeding the total of the previous ten years.
Since this year, nuclear energy tracks are blooming everywhere.
Breakthrough Energy Ventures, the "venture capital dream team" owned by Bill Gates, is also interested in the nuclear energy field and has invested in a nuclear fusion energy company established in 2019——TypeOneEnergy.
Google and Chevron jointly led the investment in US nuclear energy start-up TAEtechnologies, spending US$250 million, while Amazon founder Bezos invested in Canadian nuclear fusion company GeneralFusion. Even domesticTencent, also went to the UK to invest in a nuclear fusion start-up company affiliated with Oxford University.FirstLightFusion.
Source: HelionEnergy
It is worth mentioning thatNIOIt also invested in the establishment of a nuclear fusion company this year——Fusion new energy, of which NIO invested RMB 995 million and held 19.9% of the shares, while NIO Capital invested RMB 505 million.
Before this wave of investment, the nuclear power industry had been on the sidelines for decades because of extremely imperfect technology. Since the beginning of this year, with the successful listing of the nuclear energy company Oklo, Helion Energy has also received its first commercial order. Various performances indicate that the commercialization of nuclear energy is approaching.
Moreover, many nuclear energy start-ups are backed by Internet giants.
As these internet companies bet on AI, it’s time for nuclear energy to show its capabilities.