For thirty years, the digital economy has swung between promises of decentralization and new waves of concentration.
Each cycle starts with the same hope: more access, more distribution, more autonomy. Then scale takes over. Infrastructure centralizes. Standards are captured. Data accumulates. A new center of gravity emerges.
AI is not escaping this pattern. It is accelerating it.
Today, the CTO depends on model providers. The CIO depends on hyperscalers and enterprise platforms. The startup CEO builds on APIs controlled by a handful of companies. The deep tech founder depends on compute, data, and capital intensity that naturally push toward concentration.
In theory, innovation spreads power. In practice, it often recentralizes it.
That is why, in the age of AI, deep tech, and quantum computing, open source should no longer be treated as a secondary technical preference.
It is becoming a strategic investment.
1. Every wave creates a new center
The history of technology keeps repeating the same movement.
Mainframes gave way to client-server. Client-server gave way to the web. The web gave way to cloud. Cloud opened the illusion of modularity — and then recentralized power again through infrastructure, platforms, and dependency.
The same happened in business: bundle, unbundle, rebundle.
What looks like liberation often becomes a new form of capture.
AI is now following the same path:
- concentration of models
- concentration of compute
- concentration of cloud dependency
- concentration of interfaces
- concentration of distribution
Every wave of technology promises decentralization. Most end up creating a new center of gravity.
2. The real issue is power over the stack
The real question is no longer who can innovate fastest. It is who gets to govern the systems everyone else depends on.
That power now sits across several layers at once:
- model access
- cloud infrastructure
- developer tooling
- data storage
- enterprise software
- safety and interoperability standards
- capital allocation
This is why the AI era feels different. It is not just another product cycle. It is a concentration cycle.
And that concentration is not only economic. It is also political.
War has returned openly across multiple regions. Supply chains are weaponized. Access can be blocked. Demand can collapse. Even market leadership means little if customers can no longer buy, host, access, or trust what you sell.
At the same time, the ultra-local cannot be the full answer either. Local communities matter, but they cannot alone sustain semiconductors, frontier research, secure infrastructure, or scientific coordination at scale.
The answer is not total autonomy.
The answer is resilient interdependence: systems open enough to cooperate, distributed enough to resist capture, and structured enough to survive shocks.
3. Some layers are too important to close
Not everything should be open. But some layers are too important to become fully enclosed.
That is where open source, open standards, and technical commons matter most.
Not as ideology. As strategic infrastructure.
The point is not to eliminate proprietary business models. It is to prevent the full stack of intelligence from becoming privately enclosed.
If all strategic layers become closed, power will concentrate by default. If some foundational layers remain open, autonomy remains possible.
4. Four signals leaders should pay attention to
Some decision-makers already point in this direction, even from very different worlds.
There is no law of computer science that says that AI must remain expensive and must remain large.
Arvind Krishna
Krishna’s line matters because it pushes against the idea that enterprise AI must inevitably become more centralized, more expensive, and more closed. IBM’s recent positioning has consistently emphasized integration, enterprise value, and more open foundations rather than blind dependence on one giant stack. (crn.com)
We’re not using this to lay off people. We’re using it to make societies more productive, more efficient, and to lift everybody up.
Nicolai Tangen
Tangen’s warning is essential for boards and investors: an AI transition that becomes too brutal, too extractive, or too narrowly focused on labor removal will create backlash. He has also warned that dependence on a few providers and the unequal distribution of AI infrastructure can deepen both social and geopolitical inequality. (reuters.com)
The biggest risk in AI is concentration of power, capabilities and wealth. Let’s fight this with open-source!
Clem Delangue
That is probably the clearest strategic summary of the open-source case. The point is not that all business should be open. It is that some layers must stay contestable, inspectable, and shareable if the future of intelligence is to remain plural. (linkedin.com)
The question is not whether A.I. weapons will be built; it is who will build them and for what purpose.
Alex Karp
This is why the rise of AI manifestos matters. Some technology leaders no longer want only to build products or comment on regulation. They are beginning to articulate broader political doctrines about who should control technology, what it should serve, and how states and companies should work together. Palantir’s published principles and Karp’s public worldview are one of the clearest examples.
These four signals do not say the same thing.
But together they point toward one conclusion: deploy AI, yes — but do not let the foundational layers become impossible to inspect, challenge, or replace.
5. What leaders should do now
For CTOs, CIOs, and startup CEOs, this is no longer abstract.
A few practical choices matter immediately.
Reduce avoidable dependency. Do not build your future entirely on one closed layer if that layer can later dictate cost, access, speed, and margins.
Support open components directly or indirectly. That can mean contributions, partnerships, sponsorship, procurement choices, architecture choices, or ecosystem participation.
Separate what should remain proprietary from what should remain universal. Your differentiation may be closed. Your foundations should not all be.
Think beyond cost and speed. Open foundations are not only about cheaper software. They are about resilience, bargaining power, auditability, portability, and strategic freedom.
Treat open source as governance, not only engineering. In a more fragmented world, the question is not only what works. It is what remains governable under stress.
Conclusion
Recent Nobel work in economics points in the same direction: progress does not distribute power by itself.
The 2024 economics prize recognized Daron Acemoglu, Simon Johnson, and James A. Robinson for showing how institutions shape prosperity — and how extractive institutions undermine durable progress. The 2025 prize recognized Joel Mokyr, Philippe Aghion, and Peter Howitt for explaining innovation-driven growth and the role of technological change and creative destruction in long-term development. (nobelprize.org)
Taken together, the lesson is clear.
Innovation can create growth. But institutions determine whether that growth becomes broadly productive or merely extractive.
That is why the AI era should not be read through the myth of decentralization alone. Every major technological wave claims to spread power. Most eventually create a new center of gravity.
The question for leaders is therefore no longer only how to innovate faster. It is how to keep innovation compatible with pluralism, resilience, and freedom of action.
Open source will not replace politics. It will not replace law. It will not replace industrial strategy. But it may be one of the few realistic counterweights to excessive concentration.
In the age of AI, deep tech, and quantum computing, investing in open foundations is no longer only a technical choice. It is a strategic choice about what kind of future remains possible.
What’s the last open-source component your organization deliberately invested in — not just used, but invested in? I’m curious.
Originally published on LinkedIn. Join the discussion →
