Open-source AI without Jensen
A decapitation. It’s time to rebuild.
The resignation of Nvidia CEO Jensen Huang has created a significant challenge for the development of open model ecosystems in the United States. Although Nvidia still has financial interests in supporting open models, the process is far more complex than merely engineering a solution. Nvidia's business operations heavily rely on Huang as an operator, so it is widely believed that there will be further departures due to chip-smuggling investigations or other factors.
The company's primary goal is to stabilize the situation, deliver to customers within the next 24 months, and prevent implosion. Open models were one of Huang's most ambitious bets, but funding for this initiative is likely to diminish. Interconnects AI, a publication focused on AI, emphasizes the crucial years ahead for establishing use-cases, revenue flow, and motives to continue training the next generation of models.
One of the positive aspects of open models is the "ownership of intelligence" concept and the enterprise post-training revenue stream. However, the true long-term benefit of open models as truly commoditized intelligence is still far from realization, given the scarcity of compute resources. In the near term, open-weight models struggle to compete with inference startups that secure enough chips, as evidenced by DeepSeek's DeepInfer release.
The closed labs with better economies of scale and integration from model to datacenter and chips are likely to emerge as winners. The article highlights the need for a truly frontier open model player in the US, as Huang's leadership and protection provided essential support to the ecosystem during its formative years.
Written by urgent.news from Interconnects's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
