Open Source Software and AI Infrastructure

Overview

Open source software (OSS) has transformed how software is built. Instead of writing every component from scratch, developers now assemble products from a vast ecosystem of reusable infrastructure. The benefits are straightforward: lower cost, faster development, flexibility, and scalability.  As a result, open source is not a niche corner of software. It has become the foundation. Synopsys’ 2024 Open Source Security and Risk Analysis (OSSRA) study found that 96% of commercial codebases contain open source, and on average open source components make up roughly 77% of the codebase. In practice, this means most modern products are assembled from open building blocks. 

Open source works because it spreads before it monetizes. It is largely free to adopt, easy to inspect, and permissionless to integrate. That makes it inherently viral. A developer finds a piece of open source code that solves a problem and adds it to the product. Once it is added, it stays part of the software. When the product grows, new developers continue building on top of the same components. Over time, adoption compounds.

What is changing now is the speed of this compounding. AI tools and agents are already mainstream in software development. Stack Overflow’s 2024 Developer Survey shows that 62% of developers are already using AI tools in their workflow, and nearly 80% of new developers on GitHub use Copilot within their first week. This shift means that software is increasingly being built and operated in environments where agents interact with other systems on behalf of humans. For years, software and the web were optimized for human consumption: interfaces designed for visual navigation, dashboards built for human interpretation, and workflows structured around manual interaction. As AI agents take on more responsibility for retrieval, orchestration, and decision support, those assumptions begin to change. Systems must now be legible to machines as much as they are usable by people. Infrastructure therefore needs to be easy to integrate programmatically, simple to test automatically, and clearly structured so automated systems can interpret and operate it.

Observers of the open source ecosystem have long noted that each shift in computing infrastructure produces a new phase of open source adoption. Peter Levine described this progression in his well-known analysis of “Open Source Through the Ages,” where early open source centered on foundational software such as operating systems, followed by a second wave driven by cloud computing and DevOps infrastructure. The current transition points to the next stage in that evolution. As software becomes increasingly assembled, evaluated, and operated by automated systems, a new generation of open infrastructure is emerging, designed not only for human developers but also for the agents that now participate directly in building and running software. This emerging phase can be described as Open Source 3.0.

Why Open Source Goes Viral in the First Place

Open source spreads differently from traditional software. A proprietary product must be sold before it is used. Open source can be used before it is sold. That simple difference changes everything.  

When a developer encounters a problem, the path of least resistance is often an open source package. It can be installed in seconds, inspected line by line, and modified if needed. There is no procurement cycle and no sales call. This reduces friction at the exact moment of need. This is the first viral loop: problem → install → commit → inherit.

The second loop is ecosystem-based. Once a project becomes popular, it attracts contributors. More contributors mean faster bug fixes, more integrations, and broader compatibility. That makes the project safer to adopt, which attracts more users. The loop reinforces itself.

The third loop is standardization. When an open project becomes the de facto way to solve a problem, other tools build around it. Integrations, plugins, documentation, and tutorials form a layer of social proof. The project becomes a default choice, not because of marketing, but because it is everywhere.

This dynamic explains why open source has for a long time dominated modern software supply chains. As the earlier cited OSSRA data shows, open components are not marginal additions, they are the majority of what runs in production systems. Virality in open source is dependency virality.

The Commercialization Pattern: Open Sets the Standard, Companies Monetize Production

Open source has already proven that viral distribution can translate into durable businesses. The pattern is consistent. First, a project gains widespread developer adoption. Second, it becomes a standard inside teams and companies. Third, enterprises begin to demand reliability, security, compliance, and operational support. Finally, a company monetizes those needs through managed services, enterprise features, or cloud offerings. Several companies illustrate this pattern clearly.

GitLab provides a clear example of open source distribution evolving into a scaled commercial platform. The product became a default inside engineering teams because it reduced friction across the entire software development lifecycle, from code review and collaboration to testing and deployment. Over time, this bottom-up adoption translated into strong enterprise expansion. In its S-1 filing, GitLab reported dollar-based net retention of 152%, a level of expansion typically associated with platform products rather than single-feature tools. The company’s more recent financials reinforce this trajectory. For fiscal year 2026, GitLab reported $955.2Mn in revenue, representing 26% year-over-year growth. It also disclosed 10,682 customers generating more than $5,000 in ARR, including 155 customers contributing more than $1Mn annually. GitLab demonstrates how open distribution can seed developer adoption, which later converts into durable enterprise revenue.

ClickHouse provides a similar example in modern data infrastructure. Originally developed at Yandex for high-performance analytics workloads, ClickHouse spread quickly because it enabled real-time analytical queries at speeds and cost profiles that many traditional systems could not match. Today it is used for large-scale analytics by companies such as Uber, Cloudflare, and eBay. The scale of the project’s community reflects this adoption. The ClickHouse GitHub repository now has more than 46k stars and over 2.8k contributors. As usage expanded, a commercial layer formed around the project. In 2021 the creators launched ClickHouse Inc., raising a $50Mn Series A led by Index and Benchmark to build a managed cloud platform on top of the open database. Subsequent funding rounds have accelerated the company’s growth, including a $400Mn Series D announced in 2026 that valued the company at roughly $15Bn. ClickHouse reinforces the same commercialization pattern: open source distribution establishes the technical standard, while managed infrastructure becomes the monetization layer.

Supabase also demonstrates the same commercialization dynamic in developer platform infrastructure. The company built an open source alternative to Google’s Firebase that allows developers to launch applications quickly using a full backend built on Postgres. The approach resonated strongly with developers who wanted the convenience of managed infrastructure while retaining control over their underlying database. The Supabase GitHub repository has accumulated more than 70k stars, making it one of the fastest-growing open source developer platforms in recent years. As adoption expanded, Supabase introduced a managed cloud platform that provides hosting, scaling, and enterprise capabilities on top of the open source stack. In 2025 the company raised $100Mn at a $5Bn valuation, highlighting how quickly widely adopted open source infrastructure can translate into a large commercial platform.

Vercel shows the same dynamic at the application layer, where developers actually build and ship products. The company maintains Next.js, the open source React framework that became a default way to build web applications not through enterprise sales but through developer preference, spreading because it removed the routine work of routing, rendering, and build configuration that every team would otherwise have to build from scratch. That adoption created the commercial opening. Teams running Next.js in production need hosting, edge delivery, preview environments, and deployment tooling that behave predictably at scale, and Vercel monetizes precisely that operational layer. The company closed a $300Mn Series F co-led by Accel and GIC at a $9.3Bn valuation. Vercel is a useful case because the open project and the commercial product are deliberately decoupled: Next.js runs anywhere, which keeps adoption unconstrained, while the managed platform competes on operational quality rather than lock-in.

Across these examples the pattern is consistent. Open source drives distribution and standardization. Monetization emerges where operational complexity and enterprise risk increase.

The scale of this opportunity is visible in venture investment activity. Over the past two decades, venture funding into Commercial Open Source Software (COSS) has expanded dramatically. What began as a small niche in the early 2000s has become a major category of infrastructure investment. According to recent industry analyses, venture investment in COSS reached 211 deals in 2024 alone, totaling ~$25Bn in funding. The chart below illustrates how both the number of deals and total capital invested have grown over time, reflecting the increasing importance of open source as the foundation of modern software infrastructure.

The growth of venture investment in commercial open source software has also begun to translate into measurable liquidity outcomes. As the diagram below illustrates, roughly 850 venture-backed COSS companies have been created, and about 110 have reached an exit through either IPO or acquisition, representing roughly 12% of companies that entered the venture funnel. Of these outcomes, 24 companies have gone public, while 86 have exited through acquisitions. The timing and capital profiles of these exits show that open source companies can scale into meaningful venture outcomes, with IPO companies typically raising larger amounts of capital and reaching higher valuations, while acquisitions often occur earlier in the company lifecycle.

Exits

Over the past two decades, many widely adopted open source projects have evolved into critical infrastructure layers, making them attractive acquisition targets for major technology firms. The chart below highlights some of the largest commercial open source software (COSS) M&A transactions. These deals span multiple categories, including core infrastructure, data platforms, security, and developer tooling. The largest transaction remains IBM’s $34Bn acquisition of Red Hat, which demonstrated that open source platforms can anchor global enterprise cloud strategies. Other major outcomes have emerged in the data infrastructure layer as well. Cloudera, built around the open source Hadoop ecosystem and co-founded by an Egyptian engineer - Amr Awadallah, became one of the earliest large-scale commercial open source companies and grew into a multi-billion-dollar platform before being taken private in a ~$5.3Bn transaction in 2021. More recent transactions, including IBM’s agreement to acquire Confluent for ~$11Bn, further illustrate how companies built around widely adopted open technologies can become foundational components of modern data and AI infrastructure.

IPO Exits

Several companies built around widely adopted open source technologies have also reached the public markets. MongoDB, built around its open source document database, went public in 2017 at a valuation of ~1.6Bn and grew into one of the most widely used developer databases in modern applications. Elastic followed in 2018, listing at a valuation of approximately $4.9Bn, built on the open source Elasticsearch engine that powers search, observability, and analytics workloads across enterprises. A new wave of IPOs arrived in 2021, including Confluent, the company behind Apache Kafka’s real-time data streaming ecosystem, debuted with a valuation of ~$9Bn. GitLab also went public that year, listing at a valuation exceeding $11Bn with its widely adopted open DevOps platform. HashiCorp followed soon after, reaching the public markets at a valuation of ~$14Bn, built around infrastructure tools such as Terraform and Vault that had already become standards across cloud environments.

The OSS Shift - Open Source 3.0 (agent-first infrastructure)

The pattern described above explains how open source became commercial infrastructure. What is changing now is the pace at which that process happens. Software is no longer written and operated exclusively by humans. Increasingly, it is built, tested, and executed by AI systems and autonomous agents interacting with software at machine speed. As Zhang from a16z noted, this shift moves infrastructure from “human-speed” workloads to “agent-speed” workloads, where a single task can trigger thousands of automated queries, tool executions, or system calls within seconds. Systems originally designed for predictable human interaction must now support massively parallel execution, higher concurrency, and near-zero latency across complex workflows.

This acceleration creates demand for a new generation of infrastructure designed to operate reliably at machine scale, platforms focused on coordination, orchestration, real-time communication, and automated operations.

This shift is already visible in how leading AI companies are moving closer to the tools and workflows that power software development. Just this year (2026), OpenAI agreed to acquire Astral, founded by Charlie Marsh, the team behind some of the most widely used open-source Python tools today. These tools, including Ruff and uv, are designed to replace slower, fragmented parts of the Python stack with faster, simpler alternatives, and are now used across millions of developer workflows. OpenAI’s goal is to integrate this tooling directly into Codex, its coding agent, so that the agent can write, run, test, and manage code inside the same environments developers already use. This matters because it reflects the exact shift described above, software is no longer just written by humans, but increasingly executed and maintained by agents that need direct access to tools, not just models. OpenAI has also brought in Peter Steinberger, the creator of OpenClaw, an open-source agent framework that reached over 160k GitHub stars in a matter of months, to help build the next generation of agent systems. Taken together, these moves show that even the most model-driven companies now see open source tooling and agent frameworks as critical infrastructure, not optional layers. The same logic is visible in how platform companies are assembling their stacks. Vercel has argued publicly that the next generation of software will be built and operated not only by people but by autonomous agents that must be deployed, authenticated, and managed like any other actor in a system. Rather than build those layers in-house, it has been acquiring the open source teams behind them, taking on the Nuxt framework team in 2025 and publishing its AI SDK and v0 agent tooling under the same open-by-default approach it applies to Next.js. The pattern is worth noting for what it says about how this market consolidates. When an open project becomes a standard, the platform companies that depend on it have a strong incentive to bring the maintainers inside rather than compete with them. That suggests the teams behind widely adopted agent-era infrastructure are likely to attract strategic interest well before they reach conventional commercial scale.

Open source remains the fastest way for these tools to spread, because developers and agents can adopt and experiment with them instantly. The implication for investors is straightforward: the next generation of large open source companies will likely emerge from infrastructure that supports agent-driven software systems. The underlying model remains the same, open distribution establishes the standard, and companies monetize the production layer. What changes is the speed at which those standards emerge.

African Founders at an Inflection Point in the Open Source 3.0 Era

Open source has long allowed developers anywhere in the world to build software that reaches global audiences. For many African and African diaspora founders, this has been particularly important. Building in environments where computing resources, data infrastructure, and capital are more limited has often forced teams to rely heavily on open source tools and shared infrastructure. In practice, this pushes developers to design systems that are efficient, modular, and resilient from the start. Over time, it has produced a generation of founders who are comfortable building globally competitive open source products.

Because open source software spreads through developer adoption, some of these projects quickly gain global traction. Several African and African diaspora founders have already built products used internationally, demonstrating both the talent and the market emerging from these environments. That momentum is now pushing a new wave of founders to build the missing infrastructure itself, from data systems to AI platforms and developer tools, in recognition that the next generation of software will increasingly be driven by AI agents. In many ways, these builders are able to leapfrog older technology stacks. With less legacy infrastructure to maintain, they can design systems directly for modern AI workloads from the start. Global technology companies are beginning to recognize this shift as well. NVIDIA recently announced a ~$700Mn partnership with Cassava Technologies to deploy AI supercomputing infrastructure across Cassava’s data centers in Africa, signaling growing global investment in the region’s computing ecosystem. The companies highlighted below illustrate how African founders are beginning to build globally relevant infrastructure in the Open Source 3.0 era:

InstaDeep is another strong signal of how African-founded AI research is contributing to global technology ecosystems. The company, founded by Tunisian entrepreneur Karim Beguir, built several widely used AI research tools and models. One example is Jumanji, an open-source framework that helps researchers train reinforcement learning systems more easily and run large-scale experiments. Another is the Nucleotide Transformer, a family of large AI models designed to understand DNA sequences and help researchers study genetics and disease. The company also released related projects such as Agro-Nucleotide Transformer, which applies similar techniques to agricultural genomics, and SegmentNT, a model used for genome analysis and annotation. These systems show how AI infrastructure and open research tools developed by African-founded teams can contribute to global scientific progress. That work ultimately led to one of the largest exits in the African technology ecosystem when BioNTech acquired InstaDeep in a deal worth ~700Mn to strengthen its AI-driven drug discovery capabilities. 

Cerebrium provides another example of African diaspora talent building infrastructure for this new software stack. Africa’s linguistic diversity, with hundreds of languages and dialects, makes voice a practical interface rather than simply a convenience, especially since many of these languages remain underrepresented in existing AI systems. This reality compelled Michael Louis, the Cape Town-based founder of Cerebrium, to build infrastructure at that layer of the AI stack. Cerebrium provides serverless infrastructure that allows developers to deploy and scale AI models without managing GPUs or complex backend systems. This is particularly important for real-time inference and speech-to-speech models that power voice agents and other multimodal AI applications. As conversational interfaces become a primary way users interact with AI systems, the ability to run low-latency inference becomes a core infrastructure capability. Platforms such as LiveKit have already demonstrated how quickly real-time voice infrastructure can become foundational for AI products. Cerebrium is positioning itself to support the underlying model and inference layer powering these systems. In 2025 the company raised $8Mn in seed funding led by Gradient Ventures, Google’s AI-focused early-stage venture fund.

The broader voice AI ecosystem is beginning to draw global attention as well. Meta’s acquisition of the Egyptian-founded voice AI startup PlayAI, which had raised ~$21Mn in funding before the deal, marked the first time Meta acquired an African-founded company and underscored how voice AI is becoming a strategic priority for global technology firms. Although PlayAI itself was not open source, the team joined Meta’s AI division working on systems connected to open frameworks such as Llama, highlighting how advances in voice infrastructure are increasingly tied to the open AI ecosystem. Within this context, Cerebrium is building the infrastructure layer needed to train, deploy, and run the speech-to-speech models powering the next generation of voice agents.

BetterAuth illustrates another pattern that often drives open source innovation: developers building tools to solve problems they encounter firsthand. Its founder, Bereket Engida, an Ethiopian self-taught programmer, initially experimented with several authentication libraries while building his own applications but found many of them difficult to integrate, expensive to run, or overly complex for smaller teams. Instead of continuing to work around those limitations, he decided to build a simpler alternative that developers could adopt quickly and run more affordably. The result was BetterAuth, an open source authentication framework that allows applications to integrate identity and authorization while maintaining control over the underlying infrastructure. As adoption grew among developers looking for a more straightforward solution, the company raised a $5Mn seed round backed by Peak XV, P1 Ventures.

Stakpak reflects the same instinct one layer further down the stack, at the point where software stops being written and starts being run. Founded by Egyptian entrepreneur George Fahmy, the company started from a friction most engineering teams recognise: provisioning, configuring, and repairing cloud environments remains among the slowest and most manual parts of shipping software, even as the code above it is increasingly generated by machines. Stakpak built an open source autonomous DevOps agent that writes infrastructure-as-code across tools such as Terraform and OpenTofu and then operates that infrastructure directly. If authentication determines what an agent is permitted to do, infrastructure automation determines what it can actually execute. The company is backed by P1 Ventures, Digital Currency Group and 500 Global.

Strix, founded by a young Egyptian computer scientist, reflects how security must evolve in the open source 3.0 era. Ahmed Allam recognized that traditional security testing could not keep pace with the speed at which modern software is written and deployed. AI tools now allow developers to generate and ship code far more quickly, which means security testing must also operate continuously. Strix addresses this challenge by building AI agents that automatically perform penetration testing and vulnerability discovery directly inside development pipelines. As development cycles shorten and software changes more frequently, security tools must evolve to operate at the same speed. Strix is backed by 1984 Ventures and P1 Ventures.

Conclusion

Open source is entering a new phase. Earlier waves of open source spread through developer adoption and later became the backbone of large commercial platforms. That pattern still holds, but the pace at which it happens is accelerating. Software is increasingly written, tested, and operated by AI systems and autonomous agents that interact with infrastructure at machine speed. In this Open Source 3.0 era, the platforms that succeed will be those designed to work seamlessly with agents, systems that can handle automated workloads, continuous integration, and real-time execution at scale.

African founders have already played a role in earlier phases of this ecosystem. Amr Awadallah, an Egyptian and the co-founder of Cloudera, helped build one of the earliest large commercial companies around the open source Hadoop ecosystem, demonstrating that companies built on open infrastructure can compete globally. Today, a new generation of African and African diaspora founders is building companies for this next phase of the stack. The examples highlighted in this thesis, from Instadeep and Cerebrium to BetterAuth, Stakpak and Strix, show that these builders are already creating infrastructure aligned with the realities of Open Source 3.0. Their early traction suggests that African founders are not just participants in this transition, but increasingly part of the group shaping the next generation of open infrastructure.