Weekend Warp
Top Big AI and Tech Developments every week that you must know
This has probably been the most defining week in AI since the one with 30 November 2022. That was the day ChatGPT launched, setting off an extraordinary boom that has since transformed technology, business and the world of work.
Almost exactly four years later, the boom has suffered its biggest personal shock, triggered not by a technological breakthrough, a competitor or a market crash, but by the indiscretions of one man.
And not just any man, but arguably the most important man in AI. Jensen Huang.
Meanwhile, Google is smelling an opportunity, and SoftBank is discovering just how expensive the AI revolution can be. Interestingly, all three stories are about hardware, infrastructure and the enormous amounts of money powering AI. The models and chatbots that started this revolution seem, for once, to have receded into the background.
The Dethroning of the King
Jensen Huang, the charismatic leather-jacketed founder and CEO of Nvidia, resigned on Thursday after revelations about extramarital relationships, including one with a Chinese businesswoman suspected by American authorities of links to Chinese intelligence.
The allegations have raised uncomfortable questions about national security, semiconductor exports to China and corporate governance. Nvidia appointed its longstanding CFO, Colette Kress, as interim CEO, bringing an abrupt end to Huang’s remarkable 33-year leadership of the company.
The financial consequences have been extraordinary. Nvidia shares plunged as much as 14% during Thursday’s trading, briefly wiping out approximately $600 billion in market value, before recovering to close 4.3% lower.
Over the week, the shares lost roughly 11%, representing close to half a trillion dollars in shareholder value. From its October peak, Nvidia has lost around a third of its market value, although that larger decline cannot be attributed entirely to this scandal. These are astonishing numbers, even for an industry accustomed to dealing in trillions.
But the real significance goes beyond the stock market. Huang was not merely selling chips. He was helping build the entire AI economy, linking chip manufacturers, cloud providers, model developers, governments and investors into an increasingly interconnected ecosystem. Nvidia was, in many ways, the hardware equivalent of what OpenAI became in software. It is difficult to imagine the AI revolution of the last four years without either company. It was the ‘Bank of Nvidia’
I find two things particularly interesting.
First, the world’s most consequential AI company appears to have had a very human single point of failure.
Second, Nvidia’s dominance may now face its most serious challenge, not because its technology has suddenly deteriorated, but because competitors have been handed an unexpected opening. The technology is intact, but leadership, confidence and potentially even geopolitics have changed overnight.
The larger question is whether this is merely the fall of an extraordinary individual or the beginning of a more fundamental rebalancing of the AI industry.
Read more here.
Google’s unexpected windfall
As if on cue, Google announced two new variants of its custom AI chips this week, the TPU 8tx and TPU 8i Max. The former is designed for secure computing in regulated industries, while the latter targets demanding AI workloads requiring enormous memory and processing power.
Google claims significant performance improvements over earlier generations, and the company is increasingly making its chips available beyond its own infrastructure, including to external cloud customers.
The timing could hardly have been more exquisite.
Google is one of the few companies capable of competing with Nvidia across multiple layers of the AI stack, from chips and cloud infrastructure to Gemini models and consumer applications. While these chips were clearly developed long before Nvidia’s present troubles, Google’s opportunity to sell an alternative to nervous customers has suddenly become much more compelling. AMD and Broadcom have also benefited from the market’s reassessment of Nvidia’s dominance.
For enterprises, this is potentially excellent news. Competition in computing infrastructure could ultimately mean greater choice, lower costs and less dependence on one supplier.
For India, with its ambitions around sovereign AI and domestic computing capacity, this diversification is especially relevant. The AI race is no longer simply about who builds the smartest model. Increasingly, it is about who owns and controls the machinery on which those models run.
Read more here.
SoftBank and the price of the AI dream
And then there was SoftBank.
Masayoshi Son’s investment giant reported a quarterly net loss of approximately $7 billion, including a $3.7 billion charge relating to advance purchases of power-generation equipment for American data centres after project requirements changed. Financing costs more than doubled, even as SoftBank continued borrowing and raising enormous amounts of money for its AI investments.
There is a larger lesson here.
AI may be digital, but building it is an intensely physical and capital-hungry business. It requires chips, data centres, electricity, cooling, land and vast amounts of financing, often committed years before revenues materialise. SoftBank’s predicament illustrates the risks of building that infrastructure faster than demand and engineering specifications can settle. The opportunity may be extraordinary, but so is the bill.
Read more here.
Four years ago, ChatGPT made us marvel at what software could do. This week reminded us that the future of AI increasingly depends on who controls the hardware, who finances it and, perhaps most unpredictably, who runs the companies building it.
The machines may be getting smarter.
Their human masters remain reassuringly, and sometimes catastrophically, human.
— Jaspreet Bindra
See you next week!
Read more from Dharma of AI
The Fallible Human in the AI Machine
A different version of this essay was first published in The Economic Times-9 Nov, 2026. Read here.






Discussion about this post