Nvidia’s strongest quarter on record landed Tuesday evening without the founder who built it, as interim chief executive Colette Cress guided the company through its first earnings call since Jensen Huang’s resignation five days prior.
Third-quarter revenue reached $115.8 billion, clearing the $109 billion analyst consensus by a significant margin, while data center sales of $107.6 billion beat expectations by roughly $6 billion.
Adjusted gross margin came in at nearly 75%, R&D spending of $7.45 billion came in below the $7.7 billion estimate, and free cash flow hit approximately $43.8 billion, with shares climbing around 3% in after-hours trading.
Behind those headline figures lies a financing structure that Cress herself described with unusual candor on the call: “We get paid twice, once on the hardware sale. And again through the share of rental revenue.”
Nvidia built a second tier of customers beyond big tech by investing in smaller cloud operators, known as Neo Clouds, companies like Coreweave and Lambda, ensuring they received scarce GPUs and in some cases promising to rent back capacity if no other buyer emerged.
The strategy has kept demand elevated but has also stretched receivables sharply, with money owed by customers jumping to $63 billion in July from roughly $40 billion just three months earlier, as Nvidia extended payment terms ranging from 90 days to a full year.
While the numbers dominated trading screens, the governance drama surrounding Huang continued to reverberate through Washington and Silicon Valley simultaneously, with the two constituencies drawing starkly different conclusions.
Huang, who resigned as CEO five days before the earnings release, published a letter on X in which he apologized for what he called “serious mistakes in his personal life,” adding that if the board determined his return served the best interests of shareholders, employees, the company, and America, he would be honored to come back.
The letter’s opening lines carried a deliberate resonance, with Huang writing that he came to the United States 53 years ago and that “becoming an American citizen is one of the greatest gifts and proudest moments of my life,” before declaring “I love our country” just lines later.
The loyalty signaling appears calculated toward Washington rather than solely toward Nvidia’s board, given that U.S. intelligence agencies believe a woman Huang had an affair with was passing information to Beijing.
The woman worked at Megaspeed, a Singapore-based company accused of purchasing Nvidia chips on behalf of Chinese buyers seeking to circumvent American export restrictions, and Treasury Secretary Scott Besant has publicly called for a criminal investigation.
Several agencies had pushed to blacklist Megaspeed for months, but the Commerce Department never acted, a bureaucratic failure that has since fueled an inter-agency dispute over accountability.
Nvidia’s board launched an independent investigation and stated it had found no reason to believe company secrets or national security had been compromised, while President Trump confirmed the FBI is also examining the matter.
Support for Huang’s return has been vocal inside the company, with an internal thread drawing more than 1,600 comments, most of them backing him, and Elon Musk publicly stating he would sign a petition calling for Huang to be reinstated.
Semaphore editor in chief Ben Smith, speaking from the publication’s Silicon Valley and the World Conference, noted the gulf between the two power centers, observing that in Silicon Valley the question is not whether Huang returns but when, while in Washington the debate centers on potential criminal liability.
Away from the Huang saga, Microsoft (NASDAQ: MSFT) AI chief Mustafa Suleiman used Tuesday’s Microsoft Ignite event to unveil six new artificial intelligence models targeting coding and cybersecurity, with the company positioning cost efficiency as the defining competitive advantage.
Suleiman said the new models deliver 98% of the performance of GPT 6.1 Seoul at 40% of the cost, while Microsoft’s new cybersecurity-focused models match the performance of 6 Astra at just 2% of the cost, enabling what he described as continuously running background agents monitoring enterprise systems in real time.
Microsoft launched a superintelligence team 12 months ago and renegotiated its contract with OpenAI, and Suleiman said the company faces no compute constraints, citing “gigawatts of compute across the Microsoft fleet,” with a general-purpose frontier model also in development.
Suleiman addressed the concept of “handoverism,” a term investor Bill Gurley described as the belief held by some in the AI industry that technology will eventually become powerful enough that humanity surrenders decision-making to it, stating flatly: “I’m strongly against the idea of handoverism. I’m a humanist.”
The Microsoft AI chief also pushed back against Anthropic’s approach of treating AI systems as potentially conscious entities deserving welfare protections, arguing the industry cannot afford to blur that line as models grow more capable, particularly given what he called a “total watershed moment” in recursive self-improvement capabilities over the past three months.