At the All-In Summit 2026, NVIDIA founder and CEO Jensen Huang delivered a forceful rebuke of the so-called "AI doomsday" narrative, characterizing the widespread warnings as scientifically unfounded and irresponsible. He systematically dismantled a series of AI predictions from recent years that failed to materialize, directly accusing certain tech companies and researchers of stoking panic and misleading the public.

His primary target was a recent blog post by Anthropic CEO Dario Amodei, which had sparked significant debate. The post, endorsed by several frontier labs, issued warnings about AI safety risks and even quantified probabilities of "civilization-scale extinction." Huang pushed back sharply, stating, "This is fabricated. These people are highly educated, referred to as researchers, working in labs. The combination of these elevated terms with this kind of prediction is terrifying, and it should not be done. It's irresponsible."

Midway through the summit, former President Donald Trump made an unexpected phone call to Huang, which was broadcast to the audience. Trump declared, "The whole thing is a scam," voicing strong support for data center construction and calling AI "the oil for the next 20 to 25 years." The combined voices of the two leaders injected a new dynamic into the ongoing debate over AI regulation and safety.

A Track Record of Failed Predictions: Huang Revisits the Doomsday Claims

Huang addressed a list of the most widely circulated AI panic predictions, comparing each one against what actually unfolded. He noted that some had predicted radiologists would be completely replaced by AI within five years, but the opposite has occurred, with demand for radiologists higher than ever. He also cited predictions that 90% of code would be generated by AI within 6-12 months, that 50% of junior roles would evaporate in 6-9 months, and that models like GPT-2 and Llama 3 were too dangerous to release, along with forecasts of a "white-collar jobs apocalypse." All of these, he pointed out, failed to come true.

Huang's conclusion was blunt. He asserted that these predictions are not based on science or research, and that all evidence grounded in both points in the opposite direction, attributing the panic to fear of the unknown. He acknowledged that the safety-related portions of Amodei's blog merit serious discussion, but argued the "scientific predictions" lacked evidentiary support, stating it is irresponsible when a scientist makes claims that are not scientific.

RSI Is Not Black Magic: The Engineering Logic of Recursive Self-Improvement

Addressing the recent market buzz around recursive self-improvement (RSI), Huang offered his technical perspective. He described RSI not as an uncontrollable phenomenon but as a combination of systematic ideas, including contextual learning, skill acquisition, reflection mechanisms, reinforcement learning, and synthetic data generation. These, he stated, are all entirely reasonable concepts for enabling AI to continually improve at problem-solving.

He explained the use of methods like Low-Rank Adaptation (LoRA) as a way to augment models without retraining their base weights, and described the longer-term process of periodically retraining foundational models with accumulated experience as fully controllable from an engineering standpoint. Huang argued that the term "RSI" is being weaponized by some to create an impression of a technology spiraling out of control. He offered a simple rebuttal, noting that while RSI can be conducted within a company, any product release requires evaluation, testing, and assurance against regression, with basic engineering control processes in place.

Live Call from Trump: Data Centers Are 'Oil for the Next 20 Years'

During the summit's strategic breakdown of NVIDIA, a surprise call came in from Trump, which was patched through the venue's sound system. He told the thousands in attendance that "the whole thing is a scam," and praised data centers as being great for wealth creation in the country. "It's the oil for the next 20, 25 years - much bigger than the internet," he said, criticizing voices against data center development as serving those who do not wish America success. He also asserted that robots and AI will not take over the world. Trump mentioned a current influx of roughly $20 trillion in investment into the US, a figure he contrasted with less than $1 trillion under the previous administration.

Following the call, Huang expressed gratitude for Trump's ability to "see through the scam," noting that the president's stance required courage given polls showing substantial opposition to AI and data centers.

NVIDIA's Strategic Logic: Reach for the Sky, Stay Grounded

Discussing capital allocation and ecosystem strategy, Huang outlined a core principle for NVIDIA: "go as high as we can, but stay as low as we can." He explained that the company's goal is to enable the entire ecosystem's success rather than carve out a slice for itself, citing contributions like cuDNN and Megatron Core as foundational elements that have allowed an entire industry to flourish.

Following the acquisition of Hugging Face, NVIDIA's open-source ambitions have become clear. Huang revealed the company maintains cutting-edge open-source models in five verticals, including autonomous driving and biopharmaceuticals, mentioning the "world's first thinking autonomous vehicle" and protein synthesis models. He emphasized that NVIDIA enters these domains out of unmet customer need, not a desire to disrupt competitors. In terms of financing, he noted partnerships with entities like Blackstone and Goldman Sachs to support AI infrastructure investments and the expansion of regional cloud provider networks globally.

'Super Intelligence Has Arrived in Certain Domains': Huang's Verdict

Closing the summit, Huang delivered his assessment of the current stage of AI development. When asked if we had reached the AGI moment, he responded in the affirmative. He offered his own definitional framework, pointing to autonomous vehicles as a narrow domain where AI already exhibits superintelligence, with accident rates at a tenth of human levels, and cited similar progress in protein synthesis. His final call was for reducing the drama and ensuring America advances as a unified front, which he deemed the key to collective success.