Amidst the grinding standoff between Moscow and Washington over Kyiv’s alignment with the North Atlantic Treaty Organization (NATO), a choir of official and nongovernmental voices have sung cautionary hymns against the premise that diplomacy can or will steer the Kremlin away from settling the matter conclusive through military means. Yet, for the manifold proclamations about the seeming imminence of an expanded Russian invasion (which some interpret as being part of a deliberate messaging strategy by the US government), comparatively little consideration has been given to what factors might be restraining Russia’s President Vladimir Putin from sending combined arms units beyond the occupied sections of Ukraine’s Donetsk and Luhansk oblasts. Of these factors, the threat of the White House deploying measures intended to hobble the Kremlin’s ability to assert itself on the world stage is perhaps the most compelling.
On Feb. 15, 2022, President Joe Biden warned that the US was ready to “impose long-term consequences that will undermine Russia’s ability to compete economically and strategically” if the Kremlin invaded Ukraine. Implicit in this vow is the threat that Washington can and will disrupt the Russian efforts to develop certain technological capabilities, especially those which might reinforce the Kremlin’s geopolitical objectives.
If one assumes that Putin is a rational actor, then it stands to reason that he will avoid further conflict with the west over Ukraine lest he compromise his own vision of Russia as a global AI power.
Biden’s statement corroborates earlier reports that the US is mulling measures that would, inter alia, deny the Russian hi-tech sector access to much of the global microchip market if Russian forces push beyond separatist-held areas into Ukraine proper. On Jan. 26 2022, US Secretary of State Antony J. Blinken commented that the Biden administration was considering measures that would affect the Russian government’s economic and financial facilities, “including its ability to develop technology for the sectors that it cares most about.” The Washington Post reports that the White House is considering an exceptionally pervasive type of export control known as the Foreign Direct Product Rule (FDPR) an option that, if employed, would prohibit US and foreign companies from sending any product manufactured or designed with US technology and software to end users in Russia. According to sources familiar with the situation, the National Security Council has advised American chipmakers to undertake “preparations to shut down exports to Russia at a moment’s notice.”
Yet, will the threat of these sanctions deter Russia from starting an all-out war with Ukraine?
RUSSIA’S AI POLICY
By imposing FDPR controls on Russia, the US would constrict Putin’s campaign to achieve the strategic economic and military advantages offered by various emerging technologies, of which artificial intelligence (AI) is one of the most significant. Most histories trace the first use of the term “artificial intelligence” to a proposal submitted to Dartmouth College in August 1955 by a group of mathematicians who sought to study whether learning and other activities associated with human intelligence could be simulated by machines. The 1956 summer workshop that ensued from this proposal initiated a decades-long quest to develop and operationalize AI technology. To date, this quest appears to have mostly succeeded in defining what AI is not; where the members of the 1956 Dartmouth summer workshop postulated that computers might be made to mirror the workings of the human mind, today experts have abandoned cerebral similes in favor of understanding AI as a multiplicity of data-driven analytic tools that can help inform human decisionmaking about complex solutions to complex problems.
A decree promulgated the Kremlin in October 2019 provides a similar definition of artificial intelligence as:
“[A]set of technological solutions that makes it possible to simulate human cognitive functions (including self-learning and seeking solutions without a predetermined algorithm), as well as to obtain results during the performance of specific tasks that are at least comparable to the results of human intellectual activity.”
Putin views AI as an instrument with which to contend Russia’s great power status. Addressing an audience of students in the city of Yaroslavl in September 2017, Putin explicitly framed mastering AI as a prerequisite for becoming the “ruler of the world,” adding that Russia should aim to gain saturation in the global AI market in the manner it has already done in the global nuclear energy industry. The Kremlin also plans to employ AI to modernize the Russian state itself; Russia has already taken steps to apply AI technologies to domestic public administration as part of a broader “digital economy” strategy. This is to say nothing of the Russian military’s embrace of AI as a means of guiding cyberoperations targeting adversarial command and control systems, to name just one of many defense applications reportedly under review by the Kremlin.
Russia is not the only country intent on realizing AI’s geopolitical potential. Indeed, the Kremlin’s AI programs mirror similar activity already underway by China, the US, and other global actors. In the military realm, some commentators have spoken of an AI “arms race,” with some invoking the specter of AI-guided command and control systems being delegated to manage the conduct of nuclear warfare. AI technologies may also become the engines of a new era of ideological competition: A 2019 report published by the US Air Force’s Air University Press warns, for example, that AI may enable Beijing and Moscow to present their respective models of “digital authoritarianism” as a governance system superior to liberal democracy. Still other, more hopeful observers offer that policymakers should interpret AI technologies as a topic of beneficial diplomatic and economic exchange. The tapestry one weaves from these various threads of thought illuminates the salience of AI technologies to manifold theaters of international affairs.
WHAT FDPR EXPORT CONTROLS CAN DO
Any deployment of FDPR export controls would significantly reduce the Kremlin’s ability to keep pace with the global race to capitalize on AI technologies by throttling Russia’s access to the underlying hardware that make those AI technologies possible. Insofar as AI technologies require prodigious amounts of data to function properly, they are fundamentally reliant on computer microprocessors able to manipulate massive volumes of digital information swiftly and at scale.
The Russian computer chip industry is dominated by two companies, Baikal Electronics and the Moscow Center of SPARC Technologies (MCST), both of which rely on the Taiwan Semiconductor Manufacturing Corporation (TSMC) to fabricate their chip designs. Despite its complicated role in the Biden administration’s plans to secure global semiconductor supply lines, TSMC is an important geopolitical actor in its own right by virtue of its position as the leading manufacturer of the semiconductors used by computer chip companies worldwide. Nonetheless, the FDPR’s statutory basis will likely compel TSMC and most other key points in the global semiconductor supply chain to shutter their dealings with MCST and Baikal Electronics.
In keeping with the logic of import substitution policies (importozameshchenie) adopted in response to the imposition of western sanctions following Russia’s invasion of Ukraine in 2014, the Kremlin seeks to stimulate semiconductor production capabilities to compensate for losing access to TSMC and other overseas chip manufacturers. Russian Prime Minister Mikhail Mishustin’s ambitious roadmap for this effort, first presented in January 2020, aims to curtail Russia’s reliance on international semiconductor supply chains through import substitution, competition for global chip markets, and, ideally, eventual market dominance. Putin’s October 2019 AI strategy also prescribes the rollout of domestically-manufactured microprocessors “not inferior to their global equivalents as far as speed and energy efficiency.”
RUSSIA’S AI INVESTMENTS
A handful of semiconductor fabrication facilities (fabs) already exist inside Russia. For example, in November 2021, fabs operated by Mikron — a subsidiary of Rostec, a state-owned technology company subject to US sanctions since 2014 — reportedly completed an experimental batch of the MK32 Amur, the first ever microcontroller chip entirely designed and produced inside Russia. Although fabs like those operated by Mikron require substantial capital investments over time lest they become obsolete, Russia’s international reserves currently total more than $600 billion — a sum that should enable the Kremlin to subsidize fab development and maintenance for some time. In fact, Russia plans to invest RUB 768.5 million (approximately $9.6 million in US dollars) in information technology infrastructure by 2024. These plans appear to include recruiting foreign talent. For example, in September 2021, Russian media reported that the Ministry for Manufacturing and Trade had partnered with VEB.RF (formerly Vneshekonombank) to recruit several specialists from Taiwanese semiconductor manufacturer United Microelectronics Corporation as part of an effort to relaunch a bankrupt chip plant outside Moscow.
Any deployment of FDPR export controls would significantly reduce the Kremlin’s ability to keep pace with the global race to capitalize on AI technologies by throttling Russia’s access to the underlying hardware that make those AI technologies possible.
Whether these investments will immediately translate into chipsets capable of providing the computing power AI technologies require remains an open question. A chip’s computing power is generally measured in terms of the length of its transistor gates, which govern the flow of electricity and thereby control the flow of information through a computer system. The smaller the transistor gate, the more transistors that can fit onto a single chip, thus enabling greater computing power and energy efficiency per individual chip. At the time of this writing, the most advanced mass market chips used for AI applications, such as AMD’s Radeon Instinct family of graphics processing units (GPU) and Huawei’s Ascend 910 processors, feature transistor components measuring approximately 7 nanometers (nm). By contrast, Russian media reports indicate that the components composing Mikron’s MK32 Amur microcontroller measure approximately 90 nm. For context, 90 nm chipsets provided the computing power behind the Sony PlayStation 3 upon that console’s release in November 2006. In this sense, the MK32 Amur, though ostensibly a success in terms of Russian import substitution strategies, has specifications that fall well short of US and Chinese-manufactured AI chips, which suggests that existing Russian fabs are several generations old.
Even if the Kremlin manages to parlay its international reserves into developing more advanced fabs, there is no guarantee that Russian chip companies can produce reliable designs. Russian business daily Kommersant has reported that the Ministry of Internal Affairs, among other key government organs and state-owned enterprises have encountered extensive problems with the latest generation of MCST’s flagship Elbrus family of microprocessors. According to anonymous sources familiar with the situation, the issues with MCST’s chips are so grave as to have convinced major economic actors like Rostec and Sber (formerly Sberbank) to refrain from fully embracing domestically produced microprocessors until the latter’s computing capacities become more reliable without them becoming more expensive than foreign analogues.
Minpromtorg has even gone so far as to file a civil lawsuit demanding that the Isaac S. Bruk Institute of Electronic Control Machines (IECM), MCST’s parent institution, pay back RUB 325.5 million, which is approximately $4.1 million in US dollars as of Feb. 21 2022 in subsidies along with RUB 92.3 million (approximately $1.2 million US dollars) in additional fines. Although limited information is available regarding the legal action under concern, CNews.ru reports that Minpromtorg — which has also filed similar lawsuits against Baikal Electronics and Russian supercomputer firm T-Platforms — originally extended the subsidies as part of a program to develop a massively scalable server system employing MCST’s Elbrus 8S microprocessor chip for use in the kind of data center and supercomputer equipment utilized in AI applications.
Equally debatable is the argument that sanctions will catalyze import substitution policies mitigating Russia’s reliance on imports of chips. The available evidence suggests that this is not the case: Data collected by the International Trade Center, an international development agency run jointly by the World Trade Organization and the UNs, shows that Russian import of semiconductor devices and electronic integrated circuits actually grew by 47 % and 50 % respectively, between 2014 and 2019. Russian security services observer Andrei Soldatov maintains that the Kremlin has actually made considerable progress in hi-tech import substitution since 2019; however, most of the progress that Soldatov references concerns the development of alternatives to western online platforms and other consumer-oriented applications, rather than the sort of high-caliber semiconductors required to support advanced AI functionalities.
The Kremlin could turn to Chinese semiconductor manufacturers for AI-capable chipsets, though it is by no means certain that Beijing would allow domestic chip fabricators to supply Russia — especially if it meant running afoul of FDPR controls and depriving Chinese companies of access to US technology, financing, and markets. Moreover, the notion that using Chinese technology serves the interests of the Kremlin’s AI development strategy, and its much-touted “digital sovereignty” policy more broadly, better than using US-derived analogues may be a hard sell for elements within the Russian security elite, not least given that both of the above scenarios essentially entail reliance on a foreign power for sensitive hardware.
There is a distinct possibility that imposing FDPR controls might also compel — however indirectly — Russia’s nascent digital elite to act out in ways inimical to Putin and his Kremlin’s interests. Elements of the Russian elite are stakeholders in the Kremlin’s digitalization efforts. A good example is the Russian data center industry: The son of former Rosneft executive Sergey Bogdanchikov and Prime Minister Mishustin’s former brother-in-law both own controlling stakes in data center companies. Subsidies granted to foster development of AI and related digital technologies stand to enrich politically-exposed elites, whom Putin’s Kremlin has privileged with new “digital rents” in exchange for those elites’ diligence in ensuring that their exploitation of the same technologies serves the interests of official government strategy.
Russian political sociologists Vadim Kononenko and Elenea Ledeneva argue that Putin’s strategy of policymaking via informal patrimonial networks essentially delegates economic and political power to loyal elites to compensate for weak bureaucratic, political, and legal institutions. The result, reasons Kononenko, is a contradiction in which elites “[infiltrate] institutions, in effect merging with the state, while at the same time maintaining their own position as unaccountable to these institutions.” Ledeneva describes this phenomenon as the “modernization trap of informality,” wherein “mobilizing elites and allocating resources to insider networks undermine the fundamental principles of the rule of law, the separation of powers, and the security of property rights.”
WILL THE SANCTIONS WORK?
Were Putin to invade Ukraine and incur retaliatory FDPR controls from the Biden administration, Russia’s digital elites could compensate for the ensuing economic losses by embezzling state subsidies and resorting to legally gray methods of seizing their competitors’ assets. Worse still than an already destabilizing intraelite competition is a scenario wherein those economic elites coalesce with their counterparts in the security services around a shared belief that Putin’s actions have squandered potential rents to be gained from digitalization and even jeopardized Russia’s cybersecurity — a perception which may form a justification to remove Putin from power through various intrigues.
The forgoing arguments assume that sanctions can be used to nudge elites toward influencing the sovereign. Such is the view of economist Anders Åslund and Russian studies scholar Maria Snegovaya, who advocate that US policymakers should design sanctions to throttle the supply of rents available to Kremlin-affiliated elites so as to prompt them to push the Kremlin to agree to concessions with its Western counterparts regarding Ukraine. Yet, analysis of public opinion survey data by political scientists Mikhail Alexeev and Henry Hale indicates that sanctions targeting Russian economic elites following the 2014 annexation of Crimea, rather than invoking political opposition to Putin’s policies among the country’s upper classes, inadvertently provoked a “backlash of the better off” that saw the country’s most well-off citizens rally around the president instead.
The available information thus suggests that the Kremlin will struggle to realize the full promise of AI technologies if Putin incurs FDPR sanctions by adopting military tools to settle the present standoff over Ukraine’s NATO membership. Russia currently lacks the infrastructure to manufacture semiconductors of the quality necessary to support AI computation on a level comparable to the US and China. Although the Kremlin may eventually succeed in developing advanced fabs of its own in keeping with its broader import substitution strategies, advances by the US and China in their respective AI capabilities may render those achievements irrelevant. Moreover, by incurring FDPR controls, Putin may turn members of the Russian elite with material interests in AI and related digital technologies against him.
If one assumes that Putin is a rational actor, then it stands to reason that he will avoid further conflict with the west over Ukraine lest he compromise his own vision of Russia as a global AI power. Nonetheless, Putin’s strategy in the present crisis remains a matter of conjecture for many analysts, such that no one can say for certain what his next move might be. It is entirely possible that Putin believes the utility gained from employing armed forces to squash Kyiv’s NATO aspirations is greater than whatever potential utility might be as a consequence of export controls encumbering the Kremlin’s AI development efforts.
In any event, one thing seems clear: If Putin truly believes that AI technologies are key to Russia’s long-term geopolitical competitiveness, as he suggested he did in Yaroslavl back in September 2017, then it follows that he will seek to exhaust all diplomatic options at his disposal to resolve the present crisis.
Allen Maggard is currently a graduate student at American University’s School of International Service. Previously, he worked for the US Department of Justice’s Foreign Corrupt Practices Act unit, TD International (a due diligence and political risk advisory firm), and the Center for Strategic and International Studies.