United States: $5 billion to propel science through AI

Genesis Mission: AI becomes an instrument of American scientific power

The Trump administration has unveiled the first funding for the Genesis Mission, an initiative that places artificial intelligence at the center of U.S. research and innovation policy. According to The Verge, which devoted an article titled “The tech-broification of American science has officially begun” to the announcement, the program is set to direct $5 billion toward hundreds of AI-based scientific projects.

The amount, like the political language accompanying the initiative, goes beyond the sole issue of supporting laboratories. It reflects a broader ambition: to make AI a research infrastructure comparable, in its strategic importance, to supercomputers, major scientific instruments, energy networks, or industrial capabilities. The Genesis Mission is therefore not simply presented as yet another funding program. It marks the explicit integration of artificial intelligence into a national strategy for knowledge production, technological competitiveness, and geopolitical power.

The term used by The Verge, “tech-broification,” is deliberately critical. It refers to a now-recurring fear in the United States: that public research, its priorities, and its validation methods will move closer to the interests, pace, and culture of major technology companies. The expression does not mean that funded projects would be private or would no longer fall under public research. Rather, it describes a shift in the balance of power: generative AI, foundation models, computing infrastructure, and scientific data are becoming resources whose control is largely concentrated in a small number of industrial players.

In this context, the U.S. announcement must be read on several levels. First, there is the budgetary dimension: $5 billion is a significant envelope to launch or accelerate hundreds of programs. Then there is the organizational dimension: deploying AI across numerous scientific fields requires coordinating data, laboratories, computing, software, disciplinary expertise, and access rules. Finally, there is a geopolitical dimension: Washington is signaling that the AI race is not limited to conversational assistants, programming tools, or productivity gains in companies. It directly concerns a country's ability to make discoveries faster, test more hypotheses, and turn those discoveries into industrial advantages.

AI-assisted science is already a very concrete field. In materials science, models can help explore chemical combinations and prioritize the most promising candidates before experiments. In biology, they can contribute to the analysis of sequences, images, or complex datasets. In climate science, energy, physics, or engineering, AI can speed up the processing of large volumes of data, build surrogate models that are less costly than some simulations, or assist researchers in identifying patterns that are difficult to isolate using conventional methods.

But the issue is not simply to “put AI” into laboratories. A national program of this scale raises a more fundamental question: who owns the computing capabilities, models, data, and tools needed to make AI a mass scientific method? The American response tends to more closely connect the federal government, scientific agencies, public laboratories, universities, and the technology ecosystem. It is this coordination, far more than the $5 billion figure alone, that could reshape the global balance of research.

An American continuity: from federal research to AI platforms

The Genesis Mission does not emerge in an institutional vacuum. The United States has a long tradition of federal investment in technologies regarded as strategic. The creation of DARPA in 1958, in the context of the Sputnik shock, remains one of the most frequently cited examples: the agency supported high-risk programs, digital infrastructure, and work that contributed to the emergence of modern computing. The National Science Foundation, created in 1950, is likewise a pillar of U.S. basic research funding.

The country also has a network of national laboratories linked in particular to the Department of Energy, as well as very high-level computing infrastructure. These capabilities have made possible a research tradition combining physics, engineering, computer science, and large datasets. Artificial intelligence does not erase this model: it adds another layer to it. Where supercomputers were primarily associated with numerical simulations, AI makes it possible to add models capable of classifying, generating, predicting, and optimizing from scientific data.

This development also comes at a time when AI has become an object of international industrial competition. Large language models and multimodal models have highlighted the importance of hardware resources: specialized chips, data centers, electricity, interconnections, storage, and engineers able to operate these systems. The United States concentrates a substantial share of the most visible companies in this field, as well as cloud infrastructure providers and designers of key semiconductors.

For the U.S. government, supporting science through AI therefore means building on an existing advantage while seeking to extend it. Genesis Mission funding can create structured public demand for scientific AI tools, provide use cases for existing infrastructure, and bring researchers even closer to technologies developed by companies. This logic differs from that of a program limited to funding publications or isolated equipment. It fosters an ecosystem in which discoveries, software, and infrastructure can mutually reinforce one another.

The notion of a “mission” also carries particular political weight in the United States. It evokes programs organized around identified objectives, with the idea that research should not merely produce knowledge but solve problems deemed decisive. This approach may concern energy, health, materials, security, industry, or infrastructure resilience. By placing AI at the heart of such a framework, the Trump administration is treating this technology not as a standalone sector, but as a cross-cutting accelerator of scientific research.

This direction is not without tensions. Public research traditionally operates on long timescales, with peer-review methods, uncertainty, and results that can sometimes be difficult to monetize in the short term. AI companies, especially those active in general-purpose models, operate in a much faster, competitive environment largely based on the capture of scarce resources. Bringing these two worlds together can facilitate discoveries. It can also intensify conflicts over data access, the publication of methods, intellectual property, and reproducibility.

The Verge specifically emphasizes this cultural shift. Its title does not merely describe a budgetary policy; it suggests that American science could be shaped more by the norms of the technology industry. In this context, the question is not whether researchers will use AI tools — they already do in many fields — but under which rules, with which technical dependencies, and for the benefit of which players.

What the $5 billion funds: the industrialization of assisted research

The first funding announced for the Genesis Mission is intended to support hundreds of scientific projects using artificial intelligence. The wording is important: it emphasizes the multiplicity of projects rather than a single flagship program. The objective appears to be to spread AI throughout the American scientific landscape, creating a scaling effect. Rather than reserving the most advanced tools for a few centers of excellence, such an approach aims to accelerate their adoption across a broader range of disciplines and institutions.

This spread should not, however, be confused with automatic democratization. Deploying AI in research requires resources that remain unevenly distributed. Laboratories need usable data, computing skills, computing resources, personnel able to assess results produced by models, and robust procedures for verifying conclusions. A national funding envelope can reduce some of these obstacles, but it does not eliminate them: institutions that are already best equipped are often the most able to respond quickly to complex calls for projects.

The $5 billion figure nevertheless indicates the scale being sought. This is not marginal funding intended for a few demonstrators. The Genesis Mission places AI among the instruments the U.S. government wants to use to steer scientific innovation. In an economy where spending related to advanced computing and data centers is rising rapidly, public funding can play several roles: absorb part of the risk, create standards, support specialized datasets, and make viable research that would not be immediately profitable for the private sector.

The portfolio logic, with hundreds of projects, has another benefit. Scientific research rarely produces uniform results. Some approaches fail, others take longer than expected, and the most important advances do not always match the initial hypotheses. By supporting a large number of projects, a public mission can spread risk and allow unexpected methods or applications to emerge. This is especially relevant for scientific AI, where performance depends heavily on data quality, problem definition, and integration with real experimentation.

However, it would be misleading to present AI as a complete substitute for the scientific method. Models can produce useful predictions, but they can also reproduce biases in available data, deliver results that are difficult to interpret, or fail when confronted with situations that are underrepresented in their training data. In experimental disciplines, a hypothesis suggested by an AI system must always be tested against measurements, experiments, controls, and human expertise.

The success of the Genesis Mission will therefore depend less on the ability to multiply announcements than on the quality of this connection between computation and validation. A model can speed up the sorting of possible avenues; it does not replace instruments, protocols, measurement infrastructure, or the work of scientific teams. The risk would be to confuse the speed of hypothesis generation with the speed of producing reliable knowledge. The scientific value of an AI system is measured not only by its ability to produce a result, but by its ability to contribute to a verifiable and reusable result.

This distinction is central to the debate on research openness. The most powerful models are often costly to train and operate. When they are accessible mainly through proprietary services, researchers can benefit from high-performing tools while losing some control over the technical conditions of their work. Conversely, open approaches can improve auditability and reproducibility, but they also require funding, skills, and lasting governance. The Genesis Mission, as reported by The Verge, crystallizes this tension between efficiency, scientific sovereignty, and transparency.

Between technological acceleration and dependency: governance questions

The rise of AI in research is not solely a funding issue. It also raises governance issues that will determine the real scope of public programs. Who chooses priority fields? Under what criteria are projects assessed? What data can be shared? What results must be published? And under what conditions can private companies take part in research funded with public money?

These questions are not abstract. In contemporary AI, access to computing capabilities can be as decisive a barrier as access to a major scientific facility. Foundation models often require infrastructure that few universities or laboratories can fund on their own. As a result, a program such as the Genesis Mission can either strengthen dedicated public capabilities, increase dependence on commercial providers, or combine both. The choice between these paths determines part of the future autonomy of research.

Dependency can take several forms. It can be material, when researchers depend on chips, cloud services, or data centers controlled by a limited number of companies. It can be software-related, when essential tools rely on closed models whose parameters, training data, or operating methods are not accessible. Finally, it can be economic, when a laboratory becomes locked into pricing, an interface, or a provider for which it is difficult to substitute an equivalent solution.

For a publicly funded program, these issues are particularly sensitive. The United States has an objective advantage in AI infrastructure and industry, which can make reliance on private players both effective and difficult to avoid. But this advantage also fuels the criticism made by The Verge: when technology companies become indispensable to conducting science, the boundary between public support for research and the consolidation of private positions becomes harder to draw.

The issue of publication is equally important. Science relies on the ability to verify, criticize, and reproduce results. Yet large models, especially when they are closed, sometimes complicate this requirement. A researcher can obtain a convincing output without being able to fully examine the conditions under which it was obtained. Successive modifications to a model available through an online service can also make it more difficult to reproduce an experiment exactly several months later.

This is not an argument for excluding commercial tools from research. Proprietary software is already present in many scientific sectors, and public-private partnerships have historically contributed to major advances. But the widespread use of AI requires rethinking the safeguards associated with public funding: data preservation, model documentation, traceability of computations, access to infrastructure, and conditions for reusing results.

The Genesis Mission can thus be interpreted as an institutional test. The United States is seeking to accelerate science through technological capabilities over which it has broad dominance. The question will be whether this acceleration comes with a strengthening of scientific commons — accessible data, tools, infrastructure, and results — or whether it durably establishes research that is more dependent on industrial platforms. Both dynamics can coexist, but their balance will not be neutral for universities, small research teams, and other countries.

Europe facing a strategic decision on scientific sovereignty

For Europe, the U.S. announcement cannot be reduced to a comparison of budgets. It highlights a structural issue: the United States can combine substantial public funding with a private sector that concentrates a large share of the models, cloud infrastructure, and most visible semiconductors in the AI market. The European Union has leading research centers, recognized universities, specialized industrial players, and a tradition of scientific cooperation. But it still needs to bring these strengths together with computing, funding, and deployment capacity at the necessary scale.

Horizon Europe, the European framework program for research and innovation for the 2021-2027 period, has a budget of €95.5 billion. This envelope, however, covers a very broad field: health, climate, digital technology, mobility, industry, basic sciences, and many other priorities. It is therefore not a direct equivalent of the $5 billion from the Genesis Mission dedicated, according to The Verge, to hundreds of scientific projects using AI. The relevant comparison concerns less the gross amount than the ability to rapidly concentrate resources around a specific technological ambition.

Europe has also developed instruments dedicated to high-performance computing, notably around the EuroHPC Joint Undertaking. This infrastructure is essential, because a scientific AI strategy cannot rely solely on regulations, ethical principles, or fragmented grants. It requires machines, electricity, networks, operations teams, software, and access arrangements suited to researchers. Sovereignty does not necessarily mean self-sufficiency; it means being able to choose one's dependencies rather than endure them.

France is directly concerned. Its ecosystem combines major research organizations, universities, engineering schools, high-performance computing players, and an AI start-up community. Scientific applications of AI can concern medicine, materials, energy, climate, and industry alike. But the existence of national expertise does not by itself guarantee control of the most expensive technological layers. Models, hardware accelerators, and computing services remain globalized resources, in which American players occupy a decisive position.

The European regulatory framework adds another dimension. The AI Act, whose entry into force began on August 1, 2024, aims to regulate risks related to AI systems in the European Union. This text is not a scientific or industrial policy in the strict sense, but it influences the environment in which tools will be developed and deployed. Europe's difficulty is to avoid a simplistic opposition between regulation and innovation. Demanding governance can strengthen trust, provided it is accompanied by infrastructure, skills, and funding capable of giving European researchers and companies credible alternatives.

The Genesis Mission is a reminder that the battle is now being fought over this combination. Rules without technical capabilities can leave European users dependent on solutions designed elsewhere. Capabilities without rules can weaken the acceptability of uses and data protection. Funding without a strategy for access to infrastructure risks supporting projects whose created value will be concentrated among model and cloud providers. The European response therefore cannot be merely defensive or regulatory: it must link basic research, scientific data, computing, and the conditions for disseminating results.

A long race, beyond budget announcements

The real scope of the Genesis Mission will not be measured solely by the announcement of the first funding. It will depend on the ability of supported projects to produce verified advances, train researchers, create reusable tools, and maintain financial continuity beyond the first phase. In scientific AI, the most significant effects may take time: models must be adapted to disciplines, tested against experiments, and integrated into the daily practices of teams.

Nevertheless, the $5 billion signals a shift in priority. The Trump administration is treating AI as a direct lever for national science, and therefore as an element of competition among powers. This approach could foster an acceleration of research in fields where the United States already has data, computing, and industrial capabilities. It could also attract more talent, partnerships, and capital to American institutions, reinforcing a cycle in which infrastructure attracts projects and projects justify new infrastructure.

For French-speaking and European ecosystems, the challenge is less to mechanically replicate the American model than to define the resources to preserve and build. Open research, public computing infrastructure, quality data, auditable models, and European cooperation can constitute strategic advantages if they are funded at a sufficient scale. Conversely, lasting dependence on foreign platforms could limit laboratories' ability to set their own scientific priorities and control the conditions under which their results are produced.

The Genesis Mission thus opens a period in which AI policy will no longer be played out solely in digital ministries, competition authorities, or debates over generated content. It will also be played out in laboratories, computing centers, funding bodies, and computer hardware supply chains. The country that masters these different layers will not merely have better digital assistants: it could potentially accelerate its ability to produce discoveries, industrialize them, and shape the standards of tomorrow's science.