Nvidia CEO Jensen Huang: AI Has Reached a Critical Turning Point
Nvidia CEO Jensen Huang has recently stated on multiple occasions that artificial intelligence technology has reached a critical turning point this year, transitioning from a phase long questioned for its return on investment (ROI) to one that is "actually useful" and "extremely profitable." In a conversation with LangChain founder Harrison Chase in July, he remarked, "Now, AI has finally become useful; when AI truly works, every company in the world will benefit," emphasizing that this breakthrough has concentrated in the past six months, driven by the maturity of agent systems capable of autonomously calling tools, managing memory, and iterating tasks.
In a subsequent closed-door event held in Taiwan, Huang further stated that AI "has now become extremely profitable" and suggested that those who previously doubted AI's ROI now seem "unreasonable" in their stance. During Nvidia's Q1 FY2027 earnings call, he described the current growth trend in AI demand as "parabolic," attributing this acceleration to the proliferation of agent AI systems with reasoning, planning, and autonomous execution capabilities.
Supporting these statements are Nvidia's record-breaking financial data—Q1 FY2027 (ending April 26, 2026) revenue reached $81.6 billion, an 85% year-on-year increase, accelerating from 73% in the previous quarter. Data center revenue grew 92% year-on-year to $75.2 billion, and adjusted earnings per share surged 140% year-on-year to $1.87. Management stated that cumulative revenue visibility through 2027 has reached $1 trillion. By Q2, total revenue further climbed to $96.2 billion, a 106% year-on-year increase, with data center revenue rising 117% year-on-year to $89 billion, gross margin improving from 72.4% a year earlier to 75%, and net profit soaring 126% year-on-year to $59.68 billion, with diluted earnings per share rising from $1.08 a year earlier to $2.46.
Huang also revealed in his conversation with Harrison Chase that Nvidia's internal Nemotron 3 Ultra model scored about 86% in relevant benchmark tests, close to Anthropic's Claude Opus (87%), but at a cost only a fraction of Opus, outperforming open-source models like DeepSeek and MiniMax, which scored 82% to 83%. He opposed the complete outsourcing of AI intelligence by companies, advocating for the use of open-source frameworks to build proprietary "super agents," likening them to a company's "crown jewel." Nvidia and LangChain have jointly launched a platform called NemoCore, providing models, fine-tuning frameworks, blueprints, and secure operating environments for enterprise-level agent deployment.
Nvidia's next-generation Vera Rubin GPU architecture has received purchase orders from several cloud service and hyperscale customers, including CoreWeave, Alphabet, Microsoft, Oracle, and Nebius Group. The company's CFO stated, "We have received purchase orders from every major hyperscale cloud vendor, AI cloud service provider, and system integrator," and expects Vera Rubin to set a record for the fastest product ramp-up in Nvidia's history. Nvidia currently has a market capitalization of approximately $5.4 trillion, making it the highest-valued company globally, with analysts generally projecting a target stock price implying about a 54% upside. Huang reiterated at the Goldman Sachs Communacopia Technology Conference that global AI infrastructure investment is expected to reach $3 trillion to $4 trillion by the end of this century.
Market reactions and capital flows indicate that Huang's statements come amid the ongoing fermentation of the "AI bubble" theory—Nvidia's stock price has risen over 1017% in the past few years, and some investors are beginning to worry about the sustainability of AI capital expenditures. Huang's public emphasis on AI being "truly useful" and "extremely profitable" objectively helps alleviate market concerns about AI investment returns and stabilizes valuation expectations for Nvidia and the entire AI industry chain. Beneficiaries include investors holding Nvidia and related AI infrastructure stocks, as well as cloud computing and hyperscale data center customers that have secured orders for next-generation products like Vera Rubin. For cautious investors still on the sidelines, worried about the potential bubble in AI capital expenditures, Huang's statements serve as an official endorsement from the upstream chip supplier for the ongoing expansion of the capital expenditure cycle.
Source: Public Information
ABAB AI Insight
Huang has historically played the role of a "cycle setter" in the AI industry—back in February 2026, during Nvidia's earnings report, he stated that AI had reached a "turning point," and in March, he elaborated on revenue growth expectations surrounding the concepts of agents and "AI factories." In July, he asserted that "AI has finally become useful," and by September, he upgraded this to "extremely profitable." This series of progressively escalating statements is not isolated but rather a habitual communication strategy he employs to pre-set the tone for Nvidia's earnings and product release cycles, with each similar statement historically accompanied by corresponding validation from Nvidia's actual financial data.
A clear cycle can be observed in the flow of funds—Nvidia continuously reinvests the massive profits gained from its chip sales (with a quarterly net profit reaching $59.68 billion) into the development of the next-generation Vera Rubin architecture, while cloud computing and hyperscale data center customers (CoreWeave, Microsoft, Alphabet, Oracle, etc.) consistently allocate their capital expenditure budgets towards Nvidia chip purchases, forming a self-reinforcing loop of "chip sales profits funding R&D, downstream customer capital expenditures funding chip sales." Meanwhile, Nvidia's collaboration with LangChain to launch the NemoCore platform extends resources from merely selling hardware to helping enterprises build their own "proprietary super agents" through software tools, marking a shift in the capital path from hardware sales to software ecosystem penetration, aimed at ensuring that customers' AI investments are more embedded within the Nvidia ecosystem rather than flowing to singular leading model suppliers.
A comparable historical case is Cisco's role during the internet infrastructure construction cycle in the 1990s—at that time, Cisco also acted as the "shovel seller," providing confidence to the entire industry with continuously growing hardware orders and revenue data amid market skepticism about the sustainability of internet business models, until the internet bubble eventually burst. Currently, Nvidia's position in the AI infrastructure cycle is highly similar—being at the upstream of the computing supply chain with the strongest bargaining power, its financial data and management statements largely determine the emotional tone of financing and valuations across the entire AI industry.
Essentially, this represents a concentration of pricing power—in the current AI infrastructure investment cycle, Nvidia, as the upstream of the chip supply chain, not only controls the capacity allocation for AI infrastructure construction due to its de facto monopoly on advanced computing hardware but also significantly defines the narrative of whether the industry should continue investing in AI. This concentration occurs because the dependency of AI model training and inference on Nvidia's high-end GPUs is extremely high, with limited alternatives available, making Nvidia's revenue and profit data almost the core indicators for measuring the overall health of the AI industry. Once its management publicly releases optimistic signals, it directly influences downstream cloud vendors' capital expenditure decisions and the secondary market's valuation expectations for the entire AI industry chain, thus transforming chip suppliers from mere hardware vendors into actual cycle setters for the industry's narrative.
ABAB News · Cognitive Law
- The shovel sellers profit first, and the market dares to believe that the gold miners can make money.
- Whoever controls the supply of computing power controls the interpretive rights of the industry narrative.
- The best antidote to bubble theory is several consecutive quarters of real profits.