Content About Stocktwits
No Image Available

‘The World on its Shoulders’ in AI. Nvidia’s Mega Week. ARD #147

Author
Michael Parekh AI: Reset to Zero
7 hours ago
10 MIN READ

‘The World on its Shoulders.’ It is all Nvidia today. Both the Burdens and the Opportunities.

The most valuable company in the world, $5 trillion and counting, closes out the earnings season this Wednesday carrying Wall Street’s whole AI thesis, a widening financing pipeline underneath the data center buildout, and now America’s open source AI answer to China. Three narratives around that one company in this AI Tech Wave, plus a darker Gadget AI on where Nvidia’s smallest computers just showed up without an invitation. My takes below, as discussed on the show.

(1) Nvidia’s Quarter Wednesday. Does it Reign On?

Nvidia reports Wednesday, August 26, after the close, and the WSJ headline says the quiet part plainly: Wall Street is counting on Nvidia to keep the AI party going. One UBS strategist put it best: “we joke internally that we’re all Nvidia analysts now.” Analysts project record sales of $92 billion, a bar that has risen from $78 billion just since the start of the year, and the company must outrun 95% annual earnings growth, to over $51.5 billion of quarterly net income, to beat. It has beaten estimates for 14 straight quarters, every quarter of the AI boom, including 210% net income growth last quarter against a 126% Street projection.

What has changed this cycle is Nvidia’s widening role as backstop. This month alone: a $500 billion AI-financing framework with six Wall Street firms, with Nvidia guaranteeing up to a quarter of some transactions. An up to $105 billion backstop for OpenAI’s new Ohio data center. Stakes in land-and-power developers, Lancium and Cloverleaf the latest, the third in a rapid-fire series The Information’s AI Agenda newsletter calls Jensen Huang’s ‘manic moves’, locking Nvidia’s hardware and software bundle in at the very start of the data center process. The FT frames the emerging model as an ‘asset-light pseudo-cloud’: selective credit to neoclouds and sovereigns, a share of the revenue those new data centers generate, and Jensen arguing Nvidia chips are now an asset class you can lend against, with CoreWeave renting A100s out to 2029, nine years after that generation launched. One more number that lands close to home: server makers are flagging a roughly 17% price hike on flagship Blackwell and Rubin systems next year, which The Information notes adds at least $5 billion to the cost of a gigawatt data center.

My take: much rides again on Nvidia’s quarter this week. The results will again come without meaningful input from the world’s second largest AI market, China. That remains elusive for now, and awaits a more positive political resolution, potentially after the second US-China meeting at the White House on September 24th. In the meantime, Nvidia continues to be the key provider of AI pickaxes and shovels to customers globally, for this and other quarters to come. Quite the responsibility on their shoulders.

Sources:

For longtime readers:

(2) The Perplexity Investment Possibility

The Information reports Nvidia is discussing an investment in Perplexity as part of an equity round valuing the startup at more than $30 billion, up more than 50% from a year ago. The business behind the round: annualized revenue past $750 million, from under $250 million at the start of the year, driven in part by Perplexity Computer, its agent that professionals use to automate computer tasks. Earlier talks contemplated a licensing-and-hire deal, the play Nvidia ran with Groq on inference chips and Enfabrica on networking, before morphing toward a traditional investment. Perplexity joined Nvidia’s Nemotron Coalition in March, meets with Nvidia teams multiple times a week, plans to run agentic queries on Nvidia’s Vera CPUs, and CEO Aravind Srinivas has floated a 2028 IPO timeframe.

My take: I have long written about Perplexity, and admired how Aravind has executed the multi-model opportunity, closed and open, well beyond the ‘wrapper’ label critics hung on it. Perplexity does not train its own frontier models; it routes every request to the best model for the task, acts as a pass-through of API revenues to the model makers, and innovates hard at the user experience layer, with distribution deals like Samsung’s Android phones. That multi-model direction is where the world is going, as we saw with Stripe’s OpenRouter acquisition. A natural fit with Nvidia’s role as the AI computing ‘Kingmaker’, picking and funding the winners up and down the stack.

Sources:

For longtime readers:

(3) Poolside, and Carrying the US Open Source Flag

Nvidia is spending $6 billion to license technology from Poolside and hire about 100 of its engineers, including the team behind Laguna S, one of the most popular open weight models in the West, per the WSJ. Plus, per The Information, a separate $1 billion equity investment at a $12 billion pre-money valuation. The goal: a Nemotron open model that rivals the best frontier models within the next year, competing head-on with China’s DeepSeek and Kimi K3, at ‘super-scale’, a trillion parameters and up. The backstory is a parable of the compute era: Poolside told shareholders it had six weeks at the end of last year to raise $2 billion for a 40,000-GPU cluster. It missed the window, and lost the cluster. The company that owns the chips became the partner without that problem.

My take: this is another Nvidia step in investing wide and deep in US open weight and open source AI technologies, alongside Thinking Machines, Reflection AI, Mistral and Perplexity in the Nemotron Coalition. It counters China’s open-weight momentum, and more importantly offers ever-accelerating alternatives to closed source frontier models, even though those companies, OpenAI and Anthropic, are among Nvidia’s very best customers. Nvidia is now even one of the biggest users of its own chips, renting tens of billions of dollars of its own servers back through clouds like Lambda and Amazon. A heavy task, proactively addressed as a long-term opportunity. And just me speculating: I would not be surprised if Meta’s open-source model plans, or an open version of Elon’s Grok, found their way into this coalition too.

Sources:

For longtime readers:

My overall take: all three events highlight the critical role of Nvidia in this AI wave, still in its early innings. No let up in sight, both in the execution of its core product strategies and in its widening array of investments up and down the AI tech stack. A fanatical focus on execution by founder CEO Jensen Huang and his team, one of the best I have had the opportunity to analyze up close in over three decades, going back to heading Goldman Sachs’s internet research effort. One company as the arms supplier, the bank, the kingmaker, and now a flag-bearer of the wave. Atlas holding the world was a punishment. Nvidia volunteered.

Gadget AI: Nvidia’s Tiny AI Mobility Computers, in Russia’s AI Drones

A sad piece of news via the New York Times. Ukrainian forensic examiners found Nvidia Jetson Orin minicomputers, inexpensive consumer-grade modules built for students, robotics developers and startups, in the wreckage of Russian drones tested over Zaporizhzhia between May and July. These were fully autonomous, self-targeting systems: earlier AI-guided drones kept a human in the loop to confirm targets, with AI handling only the last jam-proof stretch; these left the strike decision entirely to the computer. One killed three people on July 6, on a flight path toward a gas station. A Jetson has also turned up on a new Russian missile being tested against Ukraine. Nvidia does not sell the devices in Russia and says they were not designed for military use; they are widely available on resale markets, virtually impossible to track, and the circuit board in the wreckage was stamped Made in China.

My take: this development, tragic as its human consequences are, is as inevitable as militaries discovering that biplanes could drop grenades from above, to the surprise of the incumbent militaries of World War One. Technology has always been dual-use, from fire and the bow and arrow on. AI in self-targeting drones is the next step in that long, grim direction, and Nvidia’s general purpose AI technologies, available globally, are bound to be used without approval in these escalations. There will be measures and countermeasures, and every military in the world is taking deep note.

Sources:

For longtime readers:

Q&A

Q1: What is MP’s take on the use of these AI capabilities in military applications?

ANSWER: Unfortunate in their human consequences, but inevitable in their use. Every general purpose technology of consequence has been pulled into warfare, from the biplane to machine guns, aircraft, tanks, and battleships surpassed by aircraft carriers. And counter-technologies will be developed rapidly against these developments, the way jamming bred jam-proof guidance. Technology has always evolved in warfare, on both sides of the ledger.

Q2: What are the wished-for safeguards?

ANSWER: Governments have to get together to determine the rules and boundaries of these technologies. And crucially, do it in peacetime, so the frameworks are ready in the worst of war times rather than improvised mid-conflict. It has been done for existential technologies before: nuclear weapons got test bans and non-proliferation regimes, chemical and biological weapons got conventions. None of them perfect, all of them better than nothing, and the world muddled through systematically. With AI it is the beginning of the beginning, and I am optimistic on a net basis that societies will figure out how to do more good than bad with these tools.

Nvidia carries a remarkable share of this AI Tech Wave on its shoulders, by its own choice. Wednesday brings the next weigh-in. Stay tuned.

Full Source Reading

Nvidia’s quarter, and the financing machine

Perplexity

Poolside and open source

Gadget AI: Jetson in the drones

Clips from today

Nvidia’s $6B Open Source Bet

Nvidia Chips in Russian AI Drones

AI Warfare’s Biplane Moment

Every New Tech Feels Existential

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here.)





Want the latest?

Sign up for Michael Parekh's Newsletter below:


Subscribe Here

More like this

BROWSE ALL ARTICLES