The long-standing fear that software will take over every facet of industry is no longer an accurate assessment of where technology is headed. From being a mere add-on, artificial intelligence has assumed the role of operating layer, while in the process devaluing traditional SaaS products.
As AI models take on ever-increasing roles in analytics, workflow logic, and decision-making, many software platforms have been relegated to thin interfaces. For technology-focused venture capitalist, Rami Beracha, this shift raises a deeper question about control: will AI be consolidated into a small group of centralized providers or will it evolve toward decentralized, network-owned intelligence?
The Shift From SaaS to AI Models
Given his extensive background in SaaS, Web2, and the evolution of decentralized systems, Rami argues that the declining valuation multiples of legacy SaaS companies are more a structural change than a market cycle. As AI systems increasingly replace discrete software features with continuously learning models, economic power is becoming concentrated in training data, compute, and distribution.
While centralized AI platforms undoubtedly benefit from scaling, concentration itself raises potential problems. With the shift in control over data access, model updates, and pricing, dependent businesses become exposed to risks. For Rami, it’s no longer about AI “eating software,” but redefining who derives value in the stack.
Centralized AI and the Question of Control
To better understand this dynamic, it’s important to realize that corporate AI platforms thrive on proprietary data and closed architectures. While this approach accelerates short-term performance, it also reinforces gatekeeping.
The sources that shape today’s AI economy—particularly compute allocation and model governance—are unfortunately inaccessible to most stakeholders. This will inevitably hinder experimentation and limit who can benefit from advanced intelligence. As AI becomes ubiquitous across industries, centralized ownership will become more of a hindrance than an advantage.
Rami Beracha proposes an alternate path: decentralizing AI networks. By distributing model ownership, training incentives, and governance across participants, dependence on single providers will be reduced.
SingularityNET is an example of how open protocols allow models to collaborate, evolve, and be monetized without centralized control. Within its framework, intelligence is considered more of a shared infrastructure than a corporate product.
Changing the Power Curve Through Decentralization
For Rami, decentralizing AGI isn’t merely an ideological concern. It actually has significant implications with regard to who has bargaining power. With portable and interoperable models governed by transparent rules, developers gain considerable leverage. Enterprises, for their part, have reduced strategic risk because intelligence isn’t hamstrung by a single vendor’s roadmap.
Even end users can benefit from competitive dynamics, reaping rewards based on performance and trust rather than on scale alone. With Rami Beracha’s extensive experience investing in foundational technology shifts, he sees this as a reflection of earlier shifts in the internet itself, when open standards surpassed closed networks.
Moving Toward the Inevitable Realignment
Decentralized AGI offers a credible challenge to centralized dominance, redefining intelligence as a shared economic resource. As the transition from SaaS to AI progresses, decentralization will ultimately determine who builds, who profits, and who controls this resource.