The future of war is no longer on the battlefield: or at least, not only. With yesterday’s announcement, it officially enters the laboratories of technology companies. With Project Meridian, announced on 30 September 2026, the Pentagon entrusts Elon Musk, Palmer Luckey and Newt Gingrich with a study on the military capabilities of the coming decades. At its core is one question: what advantage could artificial intelligence give in war?
We are not talking only about missiles and autonomous systems. Superintelligence (it is now official: for Trump it is important to keep calling it that) will represent a crucial advantage over the very opportunities that war can offer. Elon Musk’s ecosystem, Palmer Luckey’s military collaborations and NVIDIA’s race for computing power will help make sense of this interweaving of industry and power.
The day before, Washington asked AI producers to strengthen their controls, while NVIDIA presented tools to contain agents. Three distinct initiatives, one common tension: increasing the autonomy of machines while retaining the ability to stop them.
What Project Meridian is
Pete Hegseth presented Meridian during his “State of the Force” speech at Quantico. The project is to identify the capabilities needed for American technological superiority on future battlefields, examining domains ranging from underground environments to cislunar space, between the Earth and the Moon.
Coordination involves Emil Michael, the Department’s chief technology officer; Musk, Luckey and Gingrich will be co-directors, supported by academic, industry and public policy experts. The expected deadline is 120 days, with a public report and a classified annex.
Meridian is therefore a study designed to guide priorities and the development of technological capabilities for military purposes, with the aim of identifying opportunities that are not yet clear.
Future war begins with those who imagine it
At the project table, Musk brings the perspective of a space and AI entrepreneur; Luckey that of a company producing smart weapons (Anduril); Gingrich adds a political profile to the study’s leadership.
Washington is clearly trying to add industrial speed to its repertoire: people used to developing technologies, raising capital and turning prototypes into organisations. All of this to anticipate opportunities that a complex apparatus would end up recognising too late.
Of course, Musk and Luckey belong to companies with interests in the defence market, and suppliers’ input into public decisions would require transparency and checks on possible conflicts of interest. Whether there will be any, however, is not clear at present.
Autonomy in command
In the same speech Hegseth announced the Autonomous Warfare Command, planned by 1 October 2027. Its establishment will require the Pentagon to work with Congress. Meridian and this command remain separate initiatives: the study looks at future capabilities; the proposed command aims to develop and integrate autonomous and robotic capabilities.
Imagine a system that coordinates sensors, suggests priorities and updates a plan as conditions change. Speed can become an advantage. It can also compress the time available to recognise an error. This scenario is a possibility to be assessed, not a function already announced for Meridian. This is where the future of war meets the problem of control.
The White House pact and the responsibility of boards
On 29 September Donald Trump met the heads of AI at the White House. Or rather, of SI…
The voluntary agreement signed with representatives of Google, Meta, OpenAI, Anthropic, NVIDIA and X-AI covers internal controls, independent external assessments and board-level oversight.
A voluntary commitment, born in the light of recent alarm, which sets out to define verification criteria, auditor independence and the consequences of non-compliance. In short, for the first time Trump is considering AI governance. All of it, provided the lead over other countries is not lost.
And, on this point, NVIDIA also announced the Open Agent Safety Platform: an execution environment with limits and rules. According to the company, it will be able to isolate and stop within milliseconds an agent attempting to cross its permitted boundaries, as already happened in July and, again, more recently in September.
Technological superiority will require super control. The future (of war and beyond) may depend on the ability to connect models, machines and infrastructure. But as Project Meridian shows, what is at stake today is who will manage these connections.
And, let’s be honest: also who stands to profit.