Earlier this fall, a major outage at Amazon Web Services rippled across thousands of apps and websites worldwide. This disruption—which turned out to be the first of three major internet outages in the space of about a month—served as a powerful, public reminder of how reliant global infrastructure is on a small handful of large tech companies.
Science faces the same problem. Much of modern research rests on tools and platforms maintained not by universities or public institutions, but by corporations whose priorities can shift overnight. Knowledge itself is being built on the shifting sand of corporate discretion.
Tech companies know the playbook: attract users with generous free tiers, then lock critical features behind paywalls once dependence is established. Google Colab notebooks were once heralded as a dream for data science instruction, offering free graphics processing unit access and easy collaboration, enabling entire classrooms of students to train real AI models on real data sets—something that used to require expensive computers or cloud accounts.
Over time, many advanced features were quietly locked behind a subscription paywall. Earlier this year, Bitnami, a widely used software provider in academe, announced that several features long available in its free tier would now require a paid license. Slack similarly wooed universities with promises of unlimited free message history, only to remove that promise once adoption was widespread. Mendeley, once a free-spirited reference manager—a software that helps researchers collect, organize and cite academic papers—beloved by graduate students was acquired by Elsevier and reshaped to fit a commercial ecosystem. Even Dropbox, Evernote and GitHub Copilot have gone down this path: get academic users with generous free tiers, then monetize them once they’re hooked.
Well-funded labs at elite universities can afford to absorb new licensing costs or hire staff to rebuild infrastructure when tools vanish. Underresourced institutions, community colleges and researchers in the Global South cannot. The result is a widening gap between those with the means to adapt to licensing changes and those without. Scientific progress stalls not because of ideas, but because of unequal access to tools.
When tools change suddenly, the ability to replicate old experiments evaporates. It’s like your landlord changing the locks without warning: You still live there, but you can’t get back in with the key you’ve always used. Similarly, a data pipeline built on Bitnami or Google Colab, such as the models at the origin of the Alzheimer’s Knowledge base, may no longer run, even if the data still exists. That undermines one of the core principles of science: that knowledge should be accessible, verifiable and reproducible.
Science advances slowly, across decades. Commercial tools, by contrast, advance quickly through mergers, pivots and quarterly earnings reports. Tethering our infrastructure to private companies is like writing knowledge in disappearing ink.
Some argue that proprietary licenses are necessary to fund software development. This framing is misleading. While they have their own challenges, open-source ecosystems repeatedly demonstrate that sustainability does not require privatization.
In fact, many of the most widely used tools in science are open source, thriving precisely because they are not beholden to a corporate balance sheet. The R programming language has been free for decades, supported by a global consortium of statisticians and data scientists. Python, the workhorse of computational research, thrives because no single company controls it. Jupyter notebooks (for coding), PostgreSQL databases (for querying data) and Zotero citation managers (for organizing scientific papers) exist not because they are profitable, but because communities of researchers, educators and developers sustain them. These projects embody the values that science itself depends on: transparency, reproducibility and collective stewardship.
Open-source software also accelerates discovery by removing financial and technical barriers. A high school student in Nairobi can access the same statistical software as a professor at the Massachusetts Institute of Technology. A medical researcher in Brazil can collaborate on genomic analysis via initiatives like BIPMed (Brazilian Initiative on Precision Medicine), which provides open-access genomic data sets for use by scientists across the country. An environmental scientist in California can use open platforms like CliMA or Cal-Adapt to inspect the source code or download climate projection data sets, test alternative parameterizations (e.g., of wildfire risk or sea-level rise) and build upon existing models rather than being locked into proprietary black boxes. Open ecosystems create a level playing field, ensuring that innovation is driven by ideas, not budgets.
Here lies a deeper challenge: Universities are reluctant to fund software engineers at competitive salaries. A skilled engineer can earn five times more in industry than in academia, yet most grants prioritize graduate student labor over dedicated engineering roles. Without proper funding structures, essential research tools remain fragile, often maintained by overworked volunteers.
Academic culture exacerbates this fragility. Tenure and promotion systems undervalue software contributions, treating infrastructure as invisible labor rather than scholarly achievement. Without recognition, the people building the foundation of modern science often leave for industry, reinforcing dependency on corporate tools.
Bitnami’s announcement—which barely registered for most people outside the tech world but landed with a thud for the researchers and academics who quietly build on Bitnami’s offerings to build, test and share their work—should be a wake-up call. If we want a future where scientific infrastructure is reliable, equitable and reproducible, we cannot leave its fate to the market. That requires action on several fronts.
Universities must begin funding engineering staff at competitive salaries and recognizing software development as a legitimate form of scholarship in tenure and promotion decisions. They should also create dedicated funding lines for maintaining open-source infrastructure alongside traditional investments in labs, libraries and archives. Funders can play a critical role by establishing grant programs that sustain essential open-source tools—not just support new features. Researchers and departments, meanwhile, should audit their dependencies on proprietary tools, adopt open alternatives whenever possible and contribute financially or through labor to the projects that make their work possible.
Finally, companies, nonprofits and universities should invest in open-source initiatives as part of their corporate social responsibility strategies, recognizing that a stable, well-maintained research ecosystem benefits industry, too, by ensuring a steady flow of talent and innovation.
We must treat infrastructure as a public good. Instead of each lab or institution fending for itself, universities should pool resources to support shared platforms with transparent governance.
If we continue to rent the foundations of our knowledge from private companies, we should not be surprised when the floor gives way beneath us. The only way forward is to invest in true open-source infrastructure.