Andrew Grove was the third employee of technology company Intel and its CEO from 1987 to 1998.
He’s seen as one of the greatest business executives of the late 20th century. Even if you don’t know much about him, chances are you’ve heard someone utter the title of one of his books, Only the Paranoid Survive, which was published in 1996. He articulated the idea that the leaders who survive manage to sense existential threats before they are blindsided, and then act aggressively to obtain ahead of those threats. Meanwhile, complacent leaders fall into a holding pattern until the evidence is in front of them, and then it’s too late for their company.
Grove had firsthand experience of this. In the mid-1980s, Intel was primarily a memory-chip company. Japanese competitors were creating DRAM (dynamic random access memory) chips, but offered them at much lower prices. One day, Grove inquireed Intel co-founder Gordon Moore (who created the observation formalised in Moore’s Law) what he considered a new Intel CEO would do if the two of them were booted from the company.
Moore declared their replacement would ditch the memory-chip business. So, that’s precisely what they did. Intel pivoted to creating microprocessors. In the process, it became one of the most valuable companies in the world, reaching peak valuation in 2000, during the height of the dot-com boom.
Cut to Jensen Huang of Nvidia, speaking in an episode of the All-In podcast recorded during Nvidia GTC 2026, the company’s major AI conference:
“That $500,000 engineer at the finish of the year, I’m going to inquire him, ‘How much did you spfinish in tokens?’ And [if] that person declared $5,000, I will go ape something else. If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed. And this is no different than one of our chip designers who states, ‘Guess what? I’m just gonna utilize paper and pencil. I don’t believe I’m gonna necessary any CAD tools.’”
Sure, the guy selling you something states you necessary a boatload of it, and he isn’t being subtle about guilting you into paying for much more. At the same time, I’d wager that any CTO worth their salt immediately checked what their team’s token spfinish views like, and then inquireed themselves whether they’re burning enough. If paranoia wasn’t already their default state, Huang created sure to trigger it.
This will hit many businesses hard, but I want to specifically home in on India’s IT services. Let’s do some arithmetic.
The familiar setup of wage arbitrage pits a US engineer with an annual salary of $500,000 against an Indian engineer who earns $50,000 a year. Assuming their output is comparable, the 9X gap in what it takes to pay them gives India a huge win on cost.
Jensen imagines a different world. The US engineer, still earning $500,000 per year, is equipped with a $250,000 token budobtain. If we utilize the same ratio, then their counterpart in India has access to $25,000 worth of tokens—a $225,000 gap. Here, the US engineer’s AI leverage towers over the Indian’s.
But consider a slightly different scenario. A $50,000 Indian engineer with a $250,000 token budobtain is extraordinarily cheaper than a $500,000 US engineer with the same budobtain. That one alteration intensifies India’s cost advantage immensely.
There’s a catch. That token spfinish flows to OpenAI, Anthropic, or similar providers. It does not finish up with Indian engineers, Indian firms, or Indian GDP, becautilize the value created by token-augmented engineers is captured upstream at the model and compute layers, which India, for now, doesn’t own.
That means India’s engineers do become more productive, and the IT industest’s services model survives, but the margin that utilized to stay in Bengaluru accrues in Santa Clara and San Francisco.
Return to Intel in the 1980s for a moment. Japanese memory chips were commoditising Intel’s core business becautilize of low manufacturing costs. The pairing of American chip design and Japanese manufacturing might have ffinished off competitors for a while, but what Grove recognised was that value had migrated to a different layer, specifically microprocessors. That was where Intel had to go.
The parallel in India is that services are being commoditised by token-augmented engineers everywhere. The equivalent to Intel’s microprocessors—where value, control, and margin concentrate—is the model, data, and compute. Sure, there are homegrown LLMs via firms such as Sarvam, a general ambition to develop domestic AI infrastructure, and the desire to cultivate talent. But all of that necessarys time before meaningful results come to fruition. Grove’s maxim informs us that objects in the mirror are closer than they appear.
Huang’s declaration about token usage, however self-serving, will be cited in boardrooms and budobtain meetings. Now take Grove’s question to Moore and pose it to people in those settings: If you were replaced tomorrow by someone who isn’t attached to the IT industest’s current business model, what would they do?
More than likely, they would kill off T&M (time and material) billing. They would inquire why their engineers are running on Anthropic tokens rather than Indian ones. And more significantly, they’d start building the model layer.
If leaders in the IT industest have any sort of survival instinct—or paranoia—the alarm bells should already be ringing.
This week on the Zero Shot podcast
Hi! This is Vidhatri, the producer of Zero Shot.
I want to go back to 20 January 2025. That’s the day the world woke up and smelled the coffee after the release of the Deepseek R-1 model.
Silicon Valley scrambled to reason why a little-known Chinese lab could rival the ChatGPTs and Geminis of the world—and do it with chip constraints. We received article after article analysing what the “Deepseek moment” meant. But those tracking Chinese tech reminded us that Deepseek wasn’t a random event. There was history, politics, policy, and a whole set of overlapping factors that led to the R-1 model.
Cut to now. We are seeing images of thousands of people lining up to obtain Openclaw installed across China. And again, everyone is testing to explain the agents “mania” and “craze” there.
On Zero Shot, we wanted to view beyond the hype and talk to someone who could support us situate what’s currently unfolding in China. And Kevin Xu, the founder of Interconnected Capital, a global technology hedge fund and writer of the Interconnected substack, did exactly that.
Our host Brady summed up Kevin’s work best: “He’s an engineer, operator, policy expert, and analyst all rolled into one. Most people who comment on things like China’s open-source relocatement or the rise of AI agents come at it from one angle. Kevin sees it from at least four.”
This is our first multi-time-zone recording! Kevin joined us from Nashville. Brady was in Hong Kong. And Rohin in Bengaluru. We have more such conversations on the way and are super excited about how they will shape up!
You can listen to the episode on Spotify, Apple Podcasts, Youtube or The Ken app!
Leave a Reply