Open-source AI is already widely used, and many organizations say it can cost less to deploy. But those findings show adoption and perceived value—not how much open-source AI has already added to GDP. The most prominent productivity estimates cover AI overall and model possible future gains. The evidence points to meaningful economic potential, with benefits that depend on whether organizations and regions can access the infrastructure, data, and skills needed to use it.
What does the evidence show about adoption?
Linux Foundation Research’s 2025 report says 89% of organizations use some form of open source in their AI stack, while 63% of companies use an open model. These are different measures: a company can use open-source components in its AI infrastructure without using an open model. Neither figure measures productivity growth or proves that open-source AI caused economic gains.
In Europe, a December 2025 European Commission summary of the European Open-Source AI Landscape reports that over half of developers regularly rely on open models, datasets, and tools. It separately reports that 14% of EU firms used AI in 2024. Developer reliance and firm-level AI adoption have different populations and meanings, so the figures should not be read as competing estimates of the same thing.
The Commission’s summary also says the number of publicly released models has more than doubled since 2022 and inference costs dropped by more than 99% in two years. Those are findings in the report’s context, not a guarantee of equivalent savings for every provider, model, workload, or business.
Recommended Free Tools
#1 Best Overall
Do organizations actually save money with open-source AI?
There is evidence of perceived cost advantage, but it is survey evidence rather than a universal price comparison. In a May 21, 2025 announcement summarizing a Linux Foundation Research study it commissioned, Meta said two-thirds of surveyed organizations believed open-source AI was cheaper to deploy than proprietary models, and nearly half cited cost savings as a reason for choosing it. These results indicate what respondents reported; they do not establish that every open model is cheaper in practice.
For an organization—or an investor evaluating a company’s AI plans—the relevant figure is total cost for a particular use case, not just the cost of obtaining a model. A fair comparison should account for:
Rank #2
- Capability: Does each option perform the required task reliably enough? A cheaper model that needs extensive correction may not be cheaper overall.
- Deployment and operation: Include compute, integration, inference, monitoring, and maintenance costs. Actual costs depend on the workload and implementation.
- Control and customization: Available components and the terms of their licenses affect how an organization can adapt, host, and distribute a system. “Open” does not remove the need to check the specific license.
- Security and governance: More control can also mean more responsibility for updates, safeguards, and oversight.
- Readiness: Relevant data, skilled staff, and the ability to integrate and govern the system affect whether a theoretical saving can be realized.
The Linux Foundation Research report also associates open-source AI with faster and higher-quality development of tools and models. That is an account of reported evidence, not a guarantee that every organization will see the same result. Meta’s announcement is the commissioning organization’s summary, so its survey findings are best treated as reported perceptions rather than independent proof of realized savings.
What do the productivity forecasts say—and what do they not say?
An OECD working paper by Francesco Filippucci, Peter Gal, and Matthias Schief, published November 22, 2024, models AI’s possible productivity effects over a 10-year horizon. It estimates annual aggregate total-factor productivity growth of 0.25–0.6 percentage points and annual labor-productivity growth of 0.4–0.9 percentage points. These are modeled estimates assembled from micro-level performance, task exposure, likely adoption, and economy-wide linkages—not observed outcomes.
Crucially, the OECD estimates concern AI broadly, not open-source AI specifically. They cannot tell us how much open-source models have contributed to productivity or GDP so far. That distinction matters when interpreting headlines: adoption data show that organizations are using open technologies, while a model of AI’s future contribution is a projection about a much wider category.
Productivity gains also do not translate automatically into higher wages, lower prices, or more secure employment. The Linux Foundation Research summary says AI may complement jobs more than replace them, but that is not a promise for every occupation or worker. The report identifies healthcare, agriculture, construction, manufacturing, and energy as sectors where effects may differ. For a household, the eventual impact will depend on how AI changes particular tasks and how employers distribute any gains.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Who can participate in the gains?
Access is uneven. The World Bank’s Digital Progress and Trends Report 2025 says high-income countries lead in AI innovation, compute infrastructure, and startup funding; adoption is rising in middle-income countries but remains very limited in low-income economies. It identifies connectivity, computing capacity, locally relevant data, and digital skills as foundations for participation.
Open technologies can help local businesses and institutions adapt existing systems to their needs, but access to model components alone does not provide reliable electricity, affordable internet, computing capacity, useful local data, or trained workers. The World Bank’s phrase “Compute is the new electricity in the AI era—essential but unevenly distributed” is an analogy for this infrastructure constraint, not a measured statistic.
Free tools Windows power users keep installed
One-click scans. No signup required.
Best Value
The European Commission summary likewise identifies compute access as a constraint and describes EU AI Factories and EuroHPC as efforts to improve access. These initiatives address part of the ecosystem; their existence does not mean that every organization already has the resources or skills to deploy AI.
What is the most defensible conclusion?
The data support a qualified claim: open-source AI is becoming part of how organizations build and use AI, and surveyed organizations often perceive cost advantages. That is evidence of uptake and expected or perceived value. It is not an independently verified estimate of open-source AI’s realized contribution to aggregate GDP. The available OECD productivity figures are projections for AI overall, while the World Bank evidence shows why any gains may be distributed unevenly.
For readers assessing the economic story, keep three questions separate: Are organizations adopting open AI components? Do they report that those tools help with cost or development? Is there measured evidence that the technology has raised economy-wide output? The sources cited here provide evidence for the first two; they do not yet establish the third for open-source AI.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →
