When my co-founders and I started building our platform, the prevailing industry consensus was that the future of AI belonged exclusively to a tiny handful of elite, hyper-capitalized technology labs. The dominant narrative insisted that massive centralized scale and closed proprietary control were the only viable paths to frontier capabilities.
Today, that multi-billion-dollar bet on proprietary infrastructure is facing a massive market disruption. From my perspective as a founder, the era of treating AI as a rented utility is rapidly drawing to a close, replaced by an urgent global demand for open-source independence and data sovereignty.
CEO and co-founder of Featherless.ai.
The first major driving force behind this change is the reality of corporate accounting. Being fully dependent on a cloud provider for the core infrastructure of cognitive capabilities has transformed from a convenient beginning into a huge liability in strategic and security terms. It simply does not make sense to remain in a permanent closed-door monopoly.
As shown by recent research carried out by a scholar at UC Berkeley, moving an enterprise project from a proprietary API to open-source cuts the cost of computation from $3,000 down to $31. Recent reports say that this economic revolution takes place all over the world due to the fact that the quality difference between open and closed systems has become non-existent.
Independent LMSYS leaderboard shows the leaders of open source solutions to be just two per cent behind the best proprietary systems, such as Claude Opus.
Sovereign AI
Beyond the obvious economic advantages, the call for open architecture has become highly geopolitical. Both within the government sector and the business sector alike, institutions are now realizing that there are real risks involved in being tied to a foreign company's products. We are starting to see how such friction points manifest themselves through governmental actions.
In Germany, the increasing conflicts between Bavaria and Microsoft regarding data protection issues and regulations illustrate precisely why businesses cannot afford to be tied into closed systems and rely on overseas hyperscalers.
This is why I am such an avid supporter of Sovereign AI, where one is able to have complete and absolute control of their own data, their own models, and their own jurisdictions instead of continually leasing it from a third party.
This fast deconstruction will result in a great wave of stranded investments due to the immense amounts of capital being channeled into centralized and monolithic data centers.
The trend of technology is no longer towards a reliance on hyper-scale. With open source becoming so much more efficient and smaller in its footprint, there is no longer any need to do everything within a few large-scale server farms.
Engineering advancements mean that highly specialized architectures can do all of the heavy lifting locally or in a distributed network of various hardware configurations. There will be no competition between monolithic centers meant for renting proprietary compute cycles and sovereign, local networks run by companies themselves.
Open source
It is no secret that I believe that attempts to tame AI by limiting access to closed models will always boomerang against the interests of the proprietary software developer. Once there is a looming possibility that they may lose access or encounter political restrictions on export, they simply get pushed into going for open models that they will own and control themselves in their jurisdiction.
That is the whole idea of the open source philosophy. The results of scientific work and technical progress cannot be confined to just a few private labs of major corporations. With an open model, everyone can adjust its parameters for their particular use cases.
Applications of open model architecture clearly demonstrate just how much open-model architectures can achieve in terms of adaptation to a particular localized purpose. Closed-model architectures will surely remain temporarily ahead in individual benchmarks, especially in terms of the most complex and experimental functions. However, most companies do not require a Porsche on each and every trip.
What they require is a highly advanced model that is completely optimized for a specific task, which they can always have access to without being dependent on the decisions of a couple of tech leaders.
Whereas AI has the potential to revolutionize all aspects of our collective existence as well as business operations in the world, such a process cannot be controlled by a handful of providers in any way that would be safe. The general market simply wouldn’t tolerate it. It would be too risky for any contemporary company to tie its strategic development path to a single provider.
That is exactly the reason why technology leaders are pouring resources into infrastructure that is neutral and open. We can’t just invest in software and hardware; we have to invest in sovereignty.
We've featured the best AI tool.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit