By — Amartya Saldanha
Abstract
When Donald Trump and Mohammed bin Salman shook hands on a $142 billion deal, top AI CEOs were standing right beside them. Their presence drove widespread curiosity. Drones, missile systems, and AI chips were quietly bundled into the same transaction, blurring the lines between hardware and intelligence. But what happens when weapons start their own decision-making process, without any existing algorithmic governance? This piece discusses what is really being traded in Riyadh, and what’s at stake if no one notices in time.
Introduction
On May 13, 2025, Donald Trump and Saudi Crown Prince Mohammed bin Salman signed a series of documents and agreements at Riyadh’s Royal Court that was described by the white house as ‘The largest defencecooperation agreement in history’. The deal was reported at a staggering $142 billion, allocated for all types of weapons and arms technology. Sharing the stage with the heads of the two countries were, interestingly, the chief executives of NVIDIA and AMD, who were present to announce a separate but interlinked deal; NVIDIA would ship its most advanced Blackwell AI chips to Humain, a state-owned Saudi AI startup, to power a new 500-megawatt data centre. Further, along with the broader $600 billion package within which the defense deal was announced, there was also a $20 billion commitment toward Saudi AI data centres, which would be using NVIDIA’s semiconductor chips. These centres were the result of a collaboration between the aforementioned Humain and Global AI, a US-focused data centre firm.
This was not a coincidence of scheduling; in fact, it was a preview. A preview of a new template for how arms, alliances and artificial intelligence now move as one. This article argues that the Riyadh deal exposes a structural blind spot in international arms control. Current frameworks were built to govern hardware, not the algorithms embedded inside them, and no treaty currently distinguishes between selling a missile and selling the intelligence that decides where it strikes.
The Transition To AI & Arms
The Riyadh package can be read as three layers stacked into one transaction. The first being conventional hardware; at the centre of the deal is the MQ-9B SeaGuardian. This is a maritime reconnaissance and strike drone manufactured by General Atomics. Alongside the naval intelligence aircraft, Saudi also upgraded Patriot missile defence systems that have been previously used in combat to intercept Houthi missiles and drones. The second layer is infrastructure investment; these are the AI data centres themselves. Finally, the AI stack itself; Humain received commitments of 18,000 cutting-edge Blackwell chips from NVIDIA. While AMD signed a separate $10 billion partnership to provide additional AI compute capacity for the same data centres.
18 months prior, none of this would have been commercially possible; the partnership was only made possible by the Trump administration’s reversal of Biden-era restrictions on AI chip exports. Georgetown’s Centre for Security and Emerging Technology frames deals of this kind as a new form of geopolitical and economic statecraft, in which multi-billion dollar AI partnerships position American technology at the centre of the Gulf states’ digital ambitions. The weapons, chips and infrastructure are no longer separable line items; instead they are now one strategic offer.
Weapons That Think
The significance of this deal for arms control cannot be overstated; the hardware itself is no longer ‘dumb’. The MQ-9B can independently fly for over thirty hours while carrying up to eight Hellfire precision missiles, giving Saudi forces both persistent surveillance and autonomous strike capabilities over the Gulf. Such systems depend on sensor fusion and AI-assisted target recognition, technology that international humanitarian law has never been asked to certify before a sale is approved.
This is not simply theoretical; the US Central Command runs three dedicated military AI task forces, the Army’s Task Force 39, the Navy’s Task Force 59 and the Air Force’s Task Force 99, leading counter-drone experimentation inside Saudi Arabia itself, under a program known as Red Sands. Therefore, the Gulf is not simply receiving AI-enabled weapons; it is also actively testing them. Independent research on military AI reliability gives reason for caution about how far this experiment has been validated. Wang, Yang, and Jia in ‘Research on Military Artificial Intelligence Risk Governance’ (2025) found that AI target recognition systems in complex urban warfare simulations carried a misidentification rate of 12.3%, far above that of a human counterpart. The weapons are being sold as intelligent, while they are not yet reliably so.
The Governance Void
The multilateral response to this shift has been sluggish at best. In December 2024, the UN General Assembly passed a resolution on lethal autonomous weapons systems with 166 votes in favour and only 3 against; Russia, North Korea and Belarus, yet the resolution carries no binding force. Talks between two of the world’s largest AI superpowers have fared worse; the US and China initially agreed at the 2023 San Francisco summit to hold the first AI arms control talks in 2024. However, China later suspended negotiations, while the US Department of Defense has rejected calls for an outright ban on autonomous weapons. Instead, the US DOD opted for a self-governing framework built around its own 2020 Ethical Principles for AI, a model that asks states to regulate themselves rather than submit to external verification.
The current legal instrument governing conventional weapons transfers, the Arms Trade Treaty, was drafted in an era where such technology may not even have been conceived. There are no such laws pertaining to algorithmic targeting, autonomous decision loops or software updates that help a weapon gain independence after delivery. The repercussions of such actions are not merely hypothetical. In October 2025, it was reported that, based on US intelligence from 2022, G42, a Microsoft-backed AI firm, had transferred flight software technology to Huawei. This software was subsequently used to improve the targeting and flight systems of the Chinese military’s PL-15 and PL-17 air-to-air missiles. If an American ally’s AI infrastructure can be leaked to its biggest rival, the premise that bilateral chip deals can be safely walled off from military risk becomes difficult to defend.
The Riyadh deal does not exist in a vacuum. China opened its first drone manufacturing facility in Saudi Arabia in 2017, producing the UH-4 UAV, a system visually similar to the American MQ-9 Reaper. Therefore, Washington and Beijing are now competing for influence within the same areas simultaneously. What May 13, 2025 signifies is that whoever supplies the AI raw materials inside a state’s defence architecture stands to gain something far deeper than a diplomatic and trade relationship. Visibility into how a state senses, decides and strikes is the real currency being traded at the Royal Court of Riyadh.
Conclusion
At the end of the day, the US-Saudi deal, while monetarily expansive, will not be remembered for its dollar value. Its real significance lies in what it normalised, the fusion of conventional arms, AI chips, and data infrastructure into a single instrument of statecraft. A deal brokering not just arms, but the future of military and defense discourse, all finalised without a single binding international rule governing such trades. The challenge now lies not in reversing that fusion; it is irreversible. Instead, it is to build, with urgency, a governance framework that can understand the gravity and capability of autonomous weapons. The twentieth century took two world wars to produce the arms control regime we currently employ; given the current pace at which intelligence and weaponry are accelerating, the twenty-first century may not be as forbearing.
About The Author
Amartya Saldanha is a 2nd year B.Sc. Economics student at the Jindal School of Government and Public Policy.
Image Source: https://timesofisrael.com