Panel
The AI Gold Rush: Models Are Shovels, Data Is the Gold | RAISE Summit 2026
- Open source models are projected to reach intelligence levels sufficient for most knowledge work within a timeframe reducing the need for model differentiation, shifting value toward data access and enabling decentralized markets where companies build products by combining commoditized intelligence with specific data.
- Data is anticipated to evolve into a "new gold" with value shifting across transportation, banking, and human-agent interface technologies over the next five years, while more organizations will build value atop shared open-weight model infrastructure similar to the Linux ecosystem.
- Over the next five years, business value is expected to migrate from general-purpose web-scraped models to specific, fine-tuned implementations like custom post-training and RAG, accompanied by the maturity of privacy-enhanced federated learning to enable collaboration without base data disclosure.
- The market is predicted to see a proliferation of valuable companies rather than extreme centralization, with a trillion-dollar to multi-trillion-dollar economy distributed across chips, security, and search, where winners prioritize ferocity, data network effects, and unique physical resources.
- Distributed, disaggregated models living globally at the edge are expected to become a standard AI infrastructure feature, with an "agent economy" projected to grow 1,000 times over the next few years to allow individuals and companies to monetize data accessed by agents.
- Enterprises are expected to face a lag in unlocking proprietary data value due to unstructured, unlabeled data and complex regulations, particularly in Europe, until governance, security, and assurance mechanisms are sufficiently built out by the ecosystem.
- Data governance, security, and sovereignty are expected to remain critical pillars for at least the next five years with a heightened cost of failure, prompting businesses to seek on-prem or air-gapped solutions to prevent internal data from training public models.
- Real AI production systems are expected to replace the 2023 "sandbox" phase, with a critical focus on security, while thoughtful customers begin architecting reference blueprints for agentic enterprises to traverse the innovation-security chasm.
- A "headless world" is expected to emerge as traditional B2B software transforms, changing interactions for consumers and employees, while companies will eventually need systems granting internal agents full access to all company information to maximize speed.
- The future of value distribution remains unclear with scenarios ranging from the breaking up of centralization to the accrual of power by labs, while the number of companies capable of being "big" is vast, with even tiny niches potentially growing into large entities.
- New approaches to data relationship negotiation are expected to emerge to allow entities with specific or hard-to-access data, such as that in arcane languages, to share it on fair terms, while DDN has already demonstrated technical feasibility for sovereign data operations at the national level.