The Director of the MSU Institute of Artificial Intelligence, Katerina Tikhonova, now oversees a significant portion of AI education, research, and development in Russia. The Kremlin has long been working to consolidate this industry into a single system, and “control over it” can only be entrusted to a select few. However, these technologies are sovereign only on paper and in officials’ speeches. In reality, Russia remains heavily dependent on global AI leaders — decreasingly on the United States and increasingly on China. T-invariant examines how Katerina Tikhonova, Oleg Deripaska, and VTB Bank are involved in the process, why a “unified AI ecosystem” was built on Sparrow Hills, and why not only most university scientists but even the MSU Research Computing Center — whose specialists assembled the university’s first three supercomputers — have been excluded from it.
Previously on T-invariant
Punished for a Prompt: How the State Plans to Control AI in Russia
“Sunny Peak” of Sparrow Hills. How Big Computational Science at Moscow State University Became Secret and What Putin’s Daughter Has to Do With It
“Sovereign” Means Military. How Russia Militarized AI, Drone, and Cryptography Industries
State Corporation “Unified Perimeter”. How Putin’s Daughter and Her Photomodel Friend Decided to Make Innopraktika an Integrator of All High-Tech Companies
On April 21, 2026, MSU hosted the launch of the AI Faculty, marking the completion of the “unified AI ecosystem” on Sparrow Hills. Katerina Tikhonova now has all four elements of this system firmly in hand. First came the MSU AI Research Center, then the fully operational MSU Institute of Artificial Intelligence, which she formally heads. In 2024, the powerful new MSU-270 supercomputer was launched, dedicated entirely to AI applications.
Center, Institute, Faculty
The AI Faculty is led by Ivan Oseledets, who holds a Doctorate in Physical and Mathematical Sciences and is a Professor of the Russian Academy of Sciences, and CEO of the Artificial Intelligence Research Institute (AIRI). His deputy for research is Anton Konushin, who holds a Candidate of Sciences degree (equivalent to PhD) in Physical and Mathematical Sciences, and is a scientific supervisor of the MSU AI Research Center. For now, the faculty website lists only two additional administrative staff members; there is no information about faculty members. However, the research staff of the Center and the Institute significantly overlap. Most likely, they will form the core teaching staff.
“We are admitting 36 bachelor’s students, including 20 state-funded, and 36 master’s students, including 20 state-funded places,” stated Ivan Oseledets. Classes at the AI Faculty will be taught primarily by young instructors: the average age of the teaching staff is only 30 years. Tuition for paying students is about 500,000 rubles [7000 USD] per year.
The number of places in the first intake is comparable to competing universities (for example, with the HSE University), and the price, against the broader trend of rising costs for high-level IT education (see T-invariant’s analysis) — is relatively affordable. The reason may lie with the sponsors.
One of the central figures at the April 21 event was Oleg Deripaska, who was featured in most news reports and press releases and gave statements to numerous Russian state TV channels. Previously, two of his foundations financed some of the activities of both the AI Center and the Institute (as well as numerous AI-related programs aimed at students, graduate students, and staff from other faculties and university units).
Oleg Deripaska has three charitable foundations through which he funds educational and scientific initiatives. Most of Deripaska’s AI funding at MSU flows through the Intellect foundation (annual reports are available here). However, support for the faculty is provided by his best-known and longest-running nonprofit — Volnoe Delo. This may be due to public relations considerations: the large-scale coverage of the AI Faculty launch was clearly signed off on at the top. This is evident from the airtime and framing of the story on the evening news program Vremya anchored by Ekaterina Andreeva. Deripaska’s third foundation, Basis, focuses on supporting the Faculty of Physics at MSU — the alma mater of the billionaire.
Deripaska is not the only financial partner in the MSU AI ecosystem. VTB Bank plays a significant role as well. Katerina Tikhonova collaborates with the bank both as Director of the MSU Institute of AI and as head of Innopraktika. It is hard to distinguish between these two roles. With VTB funds, the Institute organizes the annual Data Fusion Awards forum and awards prizes. In April 2026, another forum was held, featuring leading AI speakers from MSU, while Katerina Tikhonova herself joined via video link and spoke about the importance of fundamental education for specialists (in recent years, Putin’s daughter has only appeared via video — whether at the SPIEF or the anniversary conference of her own Innopraktika).
China and Processors
At the 2026 Data Fusion Awards, the MSU Institute of AI received the special VTB Science Grand Prix “for its significant contribution to the implementation of joint projects on the introduction of AI technologies… and cooperation on a new AI project in China.” Details of the cooperation on the referenced Chinese project were not disclosed during the award ceremony, but it likely refers to VTB’s recent announcement that the bank had begun “a pilot industrial deployment of graphics processors of Chinese origin.”
Back in April 2025, Vedomosti reported that VTB was considering creating an AI center of expertise in China. According to the outlet, the center was to be a dedicated facility for applied joint research between Russian and Chinese specialists, enabling rapid prototyping and testing of AI devices in China—without having to import them into Russia. It is notable that in 2025 the emphasis was not on processors at all. According to the bank’s presentation, “AI devices” included wearable devices (watches, smartphones, glasses, rings), smart home devices (televisions, cars, refrigerators), and digital assistants for industrial use (machine tools, 3D printing, production monitoring and control systems, and smart transportation).
That said, GPUs are clearly critical for VTB and all other Russian institutions engaged in high-performance computing — large banks, IT companies, and the like. VTB Deputy President and Chairman of the Management Board Vadim Kulik, speaking at a recent Data Fusion forum, stated that the new graphics processors are being integrated into the bank’s core processes related to AI — including computer vision, text processing and analytics, speech recognition (speech-to-text), and generative models. The first installations of these GPUs began in March 2026. “Testing showed that the Chinese GPUs perform reliably and integrate well with the bank’s existing IT infrastructure. Deployment is proceeding smoothly, with minimal modifications and strong performance. This will speed up AI development, including work on digital assistants and AI agents,” Kulik said.
“Chinese GPUs are being integrated into more and more systems — which suits China’s interests as well. But there are, of course, difficulties with adapting software. Most language models can run on proprietary software, with results delivered through standard protocols. So, in principle, it is quite feasible to integrate video recognition or similar tasks into the bank’s processes. But creating your own models is significantly more difficult,” comments an engineer familiar with Nvidia’s operating rules and the assembly of top-tier supercomputers in Russia, speaking to T-invariant.
T-INVARIANT REFERENCE
In 2024, sources told The Wall Street Journal that Chinese internet giants and telecommunications operators were testing Huawei’s latest Ascend 910C processor. According to them, Huawei informed potential clients that the processor was on par with Nvidia’s H100 chip, which Washington has banned from being supplied to China. Since Russia’s invasion of Ukraine, all Nvidia supplies to Russia have also been banned, although gray imports continue. In 2024, T-invariant reported that the MSU-270 supercomputer was assembled using Nvidia chips purchased via a Chinese company. At the same time, all components in the tender documentation were listed under the fictitious brand SOLAR PEAK, but based on the parameters of the purchased cables, nodes, modules, and other equipment, everything is identifiable as Nvidia products.
Drones, Robot Dogs, and Gait Analysis
The new AI Faculty building is set to house a “robot park for developing control systems for robotic platforms.” “We intend to set up a park of various types of robots — from robot dogs to androids and drones. We already have a list prepared,” said Konushin (quoted by TASS). The creation of joint laboratories with the Russian-Chinese MSU-PPI University in Shenzhen is also being discussed.
The relationship between the new faculty—and the broader MSU AI ecosystem—and China may be the most revealing thread in this story. “Chinese colleagues visited our department yesterday. We discussed the creation of joint educational programs and dual-degree programs. We will certainly do this,” Konushin said.
Close cooperation did not start recently: back in November 2025, the MSU-PPI University in Shenzhen website published a report about the creation of two laboratories. The first is the Laboratory of Nanostructured Optoelectronic Functional Materials and Devices. “The work spans the entire scientific cycle: synthesis of quantum dots, perovskite nanocrystals, and chiral nanomaterials, their physicochemical and spectroscopic characterization, theoretical modeling of processes, and integration of the resulting materials into real optoelectronic devices,” the description states. The second — the Laboratory of Artificial Intelligence Algorithms and Their Applications — fits squarely into the “unified AI ecosystem” at MSU — especially given that the people pictured are, in fact, the same individuals.
T-INVARIANT REFERENCE
The joint project between the Beijing Institute of Technology and MSU is the first university jointly established by China and Russia. The memorandum to establish it was signed during Putin’s official visit to the PRC in May 2014. Instruction began at the university in September 2017. The number of students is expected to reach 5,000 in 2026.
Scientists at the MSU-PPI University in Shenzhen are working on developing AI systems for drones, stated Dmitry Shtarev, head of the university’s Department of Research, in May 2024 at the Russian-Chinese EXPO in Harbin. According to him, this is one of the “rapidly developing research areas at the university.” “At the EXPO we showcased several drone models that we are developing. These are classic drones better suited for delivering heavy payloads, as well as smaller models,” he said.
Here is how journalists described the “butterfly drone”: “When folded, all the blades tuck into the body, effectively turning it into a cylinder shape, which reduces its size for transportation. It is launched in an unconventional manner — by being thrown upward.”
In 2025, Shtarev described another joint development: “The system analyzes the characteristics of a person’s gait and compares them with a database of indicators of psychoemotional disorders, including chronic depression. Another practical application is the prevention of incidents in the metro, at train stations, and in other public places (reported accuracy exceeds 80%).”
Latest updates on scientists’ work and experiences during the war, along with videos and infographics — follow the T-invariant Telegram channel to stay updated.
Who Is the “Ecosystem” For?
Not all specialists supported the idea of creating a separate AI Faculty at MSU. “Does it really make sense to carve AI out as a separate field when students can be taught within the Faculty of Computational Mathematics and Cybernetics or the Faculty of Mechanics and Mathematics? Why create a new faculty instead of a department? It is unclear to me. And in terms of the number of students it attracts, it still effectively looks like a department. This looks like a turf war — staking out a niche, a domain, perhaps even an entire industry,” reflects a member of the Russian Academy of Sciences familiar with the situation.
The entire “unified ecosystem” operates in near-total isolation, according to another source who spoke with T-invariant. “If you look at the segment aired on the Vremya program, they showed the old Lomonosov-2 supercomputer, not the MSU-270. This suggests the film crew was not allowed inside. And the Lomonosovs have nothing to do with Tikhonova’s operations. There is a clear division. And it’s not just because the new one was built largely using gray import schemes. No one inside the university is allowed access to the new supercomputer except for a narrow circle of insiders,” the source says.
This is confirmed by another scientist familiar with the university situation. “For the second year running, I’ve been asking colleagues at the MSU Research Computing Center what’s going on on Tikhonova’s side of things. They just shrug — they are not allowed access. They found out about the creation of the faculty, as well as the launch of MSU-270, from the news,” the researcher says.
All sources interviewed by T-invariant confirm that the new MSU-270 supercomputer continues to be used exclusively for AI-related tasks, while Lomonosov-2 has become even more outdated and frequently fails. “Colleagues are essentially told: make do with your clunky old system — run your computations on Lomonosov-2. No one has access to MSU-270 except for a narrow circle,” the scientist clarifies.
Identifying People by Their Gait
”The new university division will train researchers who will be able not only to work with existing neural networks but also to create new ones. First of all, we will study issues of artificial intelligence in medicine, genetics, computer vision, and such complex topics that require a good knowledge of mathematics, computer science, and other fundamental subjects,” says MSU Rector Viktor Sadovnichiy in an interview with Channel One.
However, as T-invariant has repeatedly reported, all the main research and developments of the AI ecosystem being built by Tikhonova have potential for dual use. But this will become clear only when the customers of these technologies determine how exactly to use “classic drones better suited for delivering heavy payloads,” within which system the AI model that determines a person’s behavioral traits from their gait with 80% accuracy will operate, and for what purpose the perovskites will be used — to create solar panels or for optical and laser sights, as well as communication systems used in combat drones.
Whatever the Kremlin and the Ministry of Defense ultimately decide, the infrastructure for secrecy is already in place: first and foremost, a centralized system, controlled by the FSB, for vetting all civilian research. And this system operates much more strictly than in the Soviet Union, when Soviet scientists already had experience working in a closed scientific system. The Kremlin is now taking a decisive first step toward a new model of “sovereign science.”