#118 AI Supremacy or Self-Sufficiency? Decoding US Policy
In this edition of Technopolitik, Arindam Goswami writes on the new US domestic AI infrastructure policy. Adya Madhavan follows with a piece on AI sovereignty in the context of the military. Finally, Avinash Shet writes on India’s advancements in Scramjet technology.
Technopolitik: AI, Power, and Politics
— Arindam Goswami
A policy document recently released by the outgoing United States government which outlines its vision for domestic AI infrastructure development is important as a decisive step in tilting the balance of power in the global AI race. This strategic initiative aims at establishing AI data centres across federal lands, but, more importantly, the vision behind this is to use a multi-pronged approach to secure America’s technological leadership while also taking care of environmental and social considerations.
The Five Guiding Principles
At the heart of this policy are five guiding principles. The first principle emphasises national security and AI leadership. It recognises the geopolitical and strategic importance of AI capabilities in global affairs. Technological superiority is the aim, but more than that, it's about ensuring America's continued ability to shape the future of AI development and deployment, and to position itself as the leader, and possibly, the gatekeeper. This principle, while sounding great from a national perspective, may not be the best approach for an interconnected, globalised world, where free markets should determine such economic decisions based on considerations of comparative advantage. Nevertheless, such are the times we live in that this competitive one-upmanship in the garb of narratives of self-sufficiency and technological leadership will continue to proliferate globally. Nations would have to do a comprehensive cost-benefit analysis of such decisions to be able to harmonise national interests with global partnerships.
The policy's second principle focuses on advancing America’s economic competitiveness. It recognises that AI infrastructure isn't just about computing power. It’s a whole-system approach which requires nurturing an entire ecosystem. A vibrant technology sector creates a self-reinforcing cycle of innovation and growth.
Perhaps most striking is the policy's emphasis on clean power for next-generation data centres. This is a significant, promising shift from traditional infrastructure development approaches. It’s good that the United States aims to lead by example in demonstrating that technological progress and environmental responsibility aren't mutually exclusive. However, it would be prudent to take this with a fistful of salt, considering that the Trump government has shown scant regard towards climate change concerns, exemplified by the US withdrawing from the Paris agreement on climate change immediately after its inauguration, among other actions and words from this government.
Coming back to the policy, what makes this approach particularly noteworthy is that it explicitly addresses community interests and economic impacts. It says that AI infrastructure development must proceed without raising energy costs for Americans. The recognition that technological advancement must serve broader societal interests to be truly successful is noteworthy.
The fifth principle's focus on benefiting workers involved in building this infrastructure adds to this aspect of social consideration, acknowledging that the transition to advanced AI infrastructure must create opportunities for American workers, not just technology companies.
Translating Policy into Action
The policy has an ambitious timeline. By February 28, 2025, the Departments of Defense and Energy must each identify at least three federal sites suitable for AI data centres and clean energy facilities. This deadline, coupled with the 2027 target for operational facilities, signals urgency in translating policy into action.
This policy envisages an innovative public-private partnership model. Federal lands are to be leased to private entities for both data centres and clean energy facilities. This combination of government resources and private sector expertise and capital can potentially create a more efficient and responsive development process. By all accounts, the energy needs for AI are going to increase exponentially, even though some recent developments like the DeepSeek model have shown that energy needs need not be as astronomical as was earlier envisaged.
However, the success of this initiative will largely depend on its execution. The sites identified must balance various factors. It has to consider proximity to power sources, environmental impact, community acceptance, and security considerations. A simple policy is generally a more successful policy. Every policy must ideally have a single purpose. Piling up multiple objectives risks jeopardising the success of the policy. It adds complexity to the structure, though surely there are opportunities for integrated infrastructure development.
Going beyond the domestic implications, what is very important is that as countries compete for AI supremacy in the global AI race, the US is taking a comprehensive approach that goes beyond mere computational capability. It also sets a new standard for responsible technological advancement. My sense is that this won’t stand the test of time under the current US government, especially the clean energy requirements of this policy.
Looking at this policy from the perspective of how innovation ecosystems work, the establishment of these data centers could create new tech hubs outside traditional centres like Silicon Valley. As a policy, this is a public-private partnership model that is worth considering by the Indian government. Governments typically have large swathes of land under their control. Leasing these out for AI infrastructure development seems like a low-hanging fruit with immense benefits.
Overall, this policy has good guiding principles, both economically, strategically, and socially. India should take a good look at them, and consider exploring similar policies.
Technomachy: The Military Stakes of AI Sovereignty
— Adya Madhavan
Amidst the broader conversation about Deepseek and OpenAI, Union Minister Ashwini Vaishnaw announced India’s intention to build its own LLM. Apart from wanting to keep up with the global artificial intelligence wave, this is also a significant step towards serving India’s strategic and security needs. Why? This is because countries are increasingly trying to achieve AI sovereignty, and developing a homegrown model is a key step in that direction.
While theoretically, since freely accessible LLMs with advanced capabilities are widely available, countries don’t really have to build their own models. However, using models developed by foreign countries usually serves the interests of said countries. It sometimes can contradict the narratives of other countries, apart from the host of security concerns that come with using foreign models for critical sectors such as financial services and defence. For example, DeepSeek refuses to answer any questions about where Arunachal Pradesh (an Indian state) is located, since China lays claim to it. For the Indian government, this is misinformation that the Indian population will be privy to if they utilise DeepSeek and happen to prompt it on the subject.
Circling back to AI sovereignty– to be AI sovereign, countries need to be able to develop, deploy and control their own AI models. First, models must be built using in-house critical infrastructure, such as how DeepSeek R1 runs on Chinese-made Huawei chips. Secondly, to control how the model functions, a country needs to control what training data it is built with. This is not to be confused with saying that it only uses that particular country’s data, but rather that the developers are fully in control of what data it uses, and, therefore, try to tailor it to the contexts and needs of the country it is being developed for.
In the initial days after ChatGPT was first released, AI was more of a gimmick– something novel that captured the public's and government's attention. Since then with layers of improvement to every model and the release of new models every couple of days, it has become much more integrated into state functions across sectors. Most critical of all, perhaps, is the integration of AI in the defence sector, a phenomenon of which the extent went largely unnoticed until Israel burst onto the stage with Lavender and Daddy. Since then, reports have been released of Amazon and Google’s ties with the Israeli government and contracts between the US government and various AI giants such as OpenAI and Meta. AI sovereignty plays a key role in developing military AI, which is where India’s indigenous model is likely to play a pivotal role.
If India were to look across the border, it would see how China is exemplary of how AI sovereignty can play a role in furthering one’s military interests. The Military-Civil Fusion Strategy integrated AI startups and companies into China’s defence ecosystem, allowing defence oversight over AI development and the opportunity for technologies to be tested in the military context. Since then, China has developed multiple AI-augmented technologies for defence purposes, and there are likely to be more, given China’s propensity towards disclosing as little information as possible. To name a few examples, China has developed AI-enabled UAVs such as lightweight Kamikaze drones for surveillance and combat, AI-driven missile guidance, and an ‘AI commander’ for war games and military training.
India's having access to its own model would allow it to adapt and repurpose it for military purposes, as a few countries have done before, and many, no doubt, will do after.
Technopolitik: India’s Scramjet breakthrough
— Avinash Shet
On January 21, 2025, the Defence Research and Development Laboratory (DRDL) successfully conducted the ground testing of the indigenously developed scramjet engine. DRDL, a Hyderabad-based Defence R&D Organisation (DRDO) Lab, demonstrated the Active Cooled Scramjet engine, which ran for 120 seconds, a first for India. This test would be crucial for powering hypersonic missiles, a key modern defence technology.
Scramjet, short for Supersonic Combustion Ramjet. Unlike a jet engine, ramjet engines belong to the airbreathing engine class that lacks moving parts like compressors and turbines. Instead, they rely on aerodynamic forces for compression and thrust creation. However, this works only when these engines fly at Mach 2 or above, which is twice the speed of sound. Ramjet engines are designed to use shockwaves to reduce and compress the airflow before combustion, rendering them effective only at supersonic speeds. Beyond Mach 5, the shock wave gets excessively strong, causing high drag, outweighing the thrust generated, reducing its efficiency. Scramjet engines overcome this limitation by allowing supersonic speed with weak shockwaves. The shockwaves generated produce less drag, achieving more speed and efficiency. This lets scramjet achieve hypersonic speeds.
This is not the first time India has tested its scramjet engine. In 2016, ISRO experimented with its first flight test of a scramjet engine on an Advanced Technology Vehicle (ATV), a modified-sounding rocket. ISRO evaluates the scramjet engine's performance under supersonic conditions in this test. The ATV was carrying a system of two scramjet engines mounted on its second stage. It was launched and accelerated to Mach 6 at an altitude of approximately 20 kilometres. The scramjet engines operated for about 5 seconds. On September 2020, DRDO successfully tested the **Hypersonic Technology Demonstrator Vehicle (HSTDV)**, which is an unmanned scramjet demonstration aircraft. During this test, the vehicle achieved a speed of Mach 6 for approximately 20 seconds. DRDO has leapt by extending the duration of scramjet engine operations with the latest tests.
The first successful ground testing of a scramjet engine was done in the US by Antonio Ferri, an Italian aerospace engineer. During the Cold War, the US and Soviet Union actively developed the scramjet engine to achieve hypersonic speeds for their aircraft and missiles. In the early 2000s, the US started actual flight testing of their engines in unmanned experimental aircrafts X-43A and X51A. X51A achieved a flight duration of over 360 seconds and was hypersonic for around 210 seconds. Australia’s University of Queensland have undertaken multiple scramjet testing in collaboration with NASA. Russia already has two hypersonic missiles equipped with scramjet engines, Zircon and Kinzhal. China is still developing scramjet engines, with the latest ground test running for 600 seconds, 5 times longer than DRDL demonstrated. Japan and the European Union are also actively developing scramjet engines.
Successful development of this engine technology will be crucial for India to counter technologically ahead China. This will change the regional power dynamics in the Indian subcontinent. Neighbouring nations may accelerate their hypersonic missile development programmes, increasing the risk of advanced missile proliferation. On the other hand, India can improve its strategic autonomy by reducing its reliance on foreign systems. This development can also enhance relationships with global powers like the US and Europe.
India has showcased its technological advancement in the Scramjet engine, which can be a deal-breaker for developing hypersonic missiles for the military and Reusable Launch vehicles for civil applications.