Conference Presentation, Panel
The Building Blocks of AI: Inside the Open Stack | ZML, LiveKit, ClickHouse & Neo4j | RAISE 2026
RAISE SummitSteeve Morin, Russell d'Sa, Alexey Milovidov, Philip Rathle, Abel Samot, Alexey Kurtenkov
- ClickHouse is identified as the sole viable database solution for companies managing quadrillions of records and hundreds of petabytes of data, while standard SQL structures established in 1987 are deemed ill-suited for modern connected networks.
- Neo4j has onboarded more than 700 AI startups, leveraging knowledge graphs as a significant growth driver to support this expansion.
- LiveKit Cloud, launched in late 2022, operates a global network of data centers to deliver streaming with latency of hundreds of milliseconds or less while reimagining the entire infrastructure stack for a stateful, real-time world.
- Future human-computer interaction is expected to evolve into direct verbal communication and physical movement, such as via humanoid robots, as companies build end-to-end stacks to support this shift.
- ZML currently supports eight chip architectures and projects reaching 12, potentially 15, by the end of the current year.
- Enterprises will require operational control for AI to become load-bearing within a future defined by the composability of software and hardware.
- Token prices are forecasted to rise due to GPU constraints, reversing prior predictions that values would decrease to zero.
- A missing "harness layer" or "application infrastructure layer" is noted as critical to maximize AI model capabilities, rendering Jensen Huang's five-layer model incomplete.
- Data is characterized as the non-commoditized "fuel" and the primary long-term differentiator, whereas application code becomes trivially copyable, making brand and network effects the main value capture mechanisms.
- While brand loyalty is difficult to commoditize, consumer populations exhibit low switching costs when open source options emerge, making open source the most effective method for capturing attention.
- AI will lower barriers to open source contribution, enable maintainers to improve code quality, and assist in clearing backlogs, while simultaneously reducing the viability of proprietary approaches as agents can re-implement proprietary APIs using open source models.
- The pragmatic defense for open source companies relies on the community and brand value, which cannot be replicated even if code is forked, alongside the maintenance of proprietary "super hard environments" to protect infrastructure from copying.
- Europe is undergoing a hardware and software renaissance with accessible talent and fungible capital, allowing local founders to secure funding without relocating to the US.
- France and the US are progressing toward sovereign AI implementations, with LiveKit already powering US 911 services and the French government's sovereign Zoom version.
- Organizational sovereignty focuses on ensuring data remains under customer control and preventing disconnection, while the consensus notes that attempting to build independent silicon supply chains across many nations may prove impractical.