Today’s theme: fear and greed are driving risk mitigation around global AI data-center infrastructure investments. The US is now considering stakes in leading AI and tech companies — motivated by both potential gain and the need to contain company risks. Three Takes today, each with my Take — and my Overall Take. (It’s a busy week: WWDC today, the SpaceX IPO by week’s end — much more on both ahead.)
Axios had it — “Trump wants a slice of the AI boom” — with CNBC on the Trump/OpenAI discussions of a possible government stake and Bloomberg on Trump considering stakes in top AI labs. OpenAI figured out it has a transactional president — and Sam Altman pitched both sides of the aisle: the right fearing it misses the AI upside, the left fearing AI takes jobs. The charging into global AI capex frame is in AI-RTZ #667.
MP Take: A good deal for the companies in question if they can get it — but it warps the US capital market. And it’s more negative for the AI companies to come, who won’t have that benefit and will take different risk/reward actions. The second thing to keep in mind: this makes the US system more similar to the Chinese system, where the government takes ‘Golden Shares’ — controlled shares — in its tech and AI champions. It’s ironic that while competing geopolitically with the Chinese Communist Party, we’d be taking on more of their persona in our investment profile. Not likely to happen, but an important consideration across administrations.
The Information ran the piece — “Goldman Sachs, JP Morgan explore new ways to tame AI lending risks.” The tens of billions the Blackstones and others are putting into financing vehicles — entities being packaged and resold, 2008-CDO-style, to insurance companies and pension funds. The AI data centers’ growing global financing needs frame is in AI-RTZ #777.
MP Take: Continued financial engineering can extend the current wave of institutional financing interest in global AI infrastructure capex — equity and debt, private and/or public. Google’s recent $80 billion+ equity raise is a new catalyst for other big tech — the Microsofts, Metas and Amazons — to consider similar options. Beyond equity and debt, hybrid structures like convertibles can play a bigger role. A step in the right direction — but as we’ve seen in past cycles, you’ve got to watch it: sometimes the risks get ahead of the reports. We’re in the early stages.
The Information had it — “Google and Nvidia consider Intel as backup chip manufacturer.” And so is Elon — after TSMC (the ‘Fed of the global chip economy’, AI-RTZ #1108) capped growth at about 30% while everyone wants more. Plus Nvidia’s fresh South Korea memory deals with SK Hynix, Samsung and LG, wrapping his Computex/GTC swing. The Nvidia vs its top customers for AI GPUs frame is in AI-RTZ #554.
MP Take: Now that the template for these deals is set — and public-market interest continues — expect this to accelerate. And it’ll be global in nature. Intel in the US remains a prime beneficiary — with not just Google and Nvidia, but Elon and Apple of late. These are commitments of tens and hundreds of billions, up the risk/reward curve: if something goes askew on timing here versus there, it could throw off the precise calculations. We’re seeing it downstream too — not just in fabs for GPU/CPU chips, but in memory and beyond.
The common thread: the accelerating AI infrastructure ‘musical chairs’ game is continuing and accelerating — all while every participant logically asks, how do we reduce risk, or at least transfer it to other parties, and provide new incentives for stakeholders in this AI Tech Wave competition?
MP Take: That’s why OpenAI is proactively soliciting government stopgaps — and paths to regulatory capture vs newer AI startups. It’s why the administration is leaning in, given its proclivity to transactions. And it’s why big tech keeps leaning into increasingly ‘circular’ deals while the banks work to make everyone happy — my alma mater Goldman Sachs leading all three major mega-IPOs and the equity raises like Google’s. Given the diversity of vehicles and structures, there’s ‘something for everyone’ in terms of investment needs and priorities, by asset class and geography. Despite the higher risk/reward volatility, we are in the early phase of these risk-mitigation and profit-maximizing opportunities around global AI data-center and power infrastructure.
The New York Times has it — “Modern smartphones rolled out in 2007 — the year fertility rates began falling. Two new studies say it’s not a coincidence.” Global fertility is declining faster than demographers expected — and not just in developed countries (US 1.8-1.9, far below the 2.1 replacement rate), but China and India drifting toward 1.7-1.8, and now even parts of Africa, unexpectedly. Of 8 billion people, about 4-5 billion now have smartphones. The societal AI angst this tech wave frame is in ARD #78.
MP Take: The causality data is getting early academic attention on the way to credibility — but actually connecting these dots is a difficult long-term game given the wide range of economic and social factors. I’m loath to say it’s all the phones. But these narratives shape societal and regulatory attitudes toward technology — and AI in particular — especially at a time of record-low early AI approval in the US. Demographics is destiny, and these are curves very tough to change (China’s one-child policy still echoes decades later). Maybe technology is another factor we haven’t considered. The smartphone is the ultimate gadget the world, rich or poor, spends hours a day with — if it affects how we think about social relationships in the real world, we need to be thoughtful about it.
Mainstream consumer AI apps aren’t here yet at scale — we haven’t seen the consumer equivalent of what enterprise got with AI coding and cybersecurity. Meanwhile people are inundated with job-loss and existential fears, often propagated by the AI founders themselves (looking at you, Dario Amodei). I disagree with the AI-job-loss posture — Jensen Huang is more right that AI will create more jobs than expected. Add the over-anthropomorphizing of AI as sentient when it’s just computer technology. Don’t expect the antipathy to abate until the mega-AI IPOs are done.
Google and Apple in particular — moving from the age of chatbots to the age of agents: ‘search and ask’ becomes ‘search, ask and do,’ with their vertically integrated global platforms. It’ll take two or three years to see at scale. And note the new reality for startups: you have to take whatever cool thing you have to billions of users in two or three years — a very different environment than my 30-year career, when startups had five to ten. Think Instagram and WhatsApp without Facebook, or YouTube without Google — that big-company-plus-startup synergy is what made them. We’ll see all of that in spades over the next three to five years.
For the full context, see the canonical sources:
On the consumer side we haven’t seen the mainstream, at-scale AI applications that wow everybody — like enterprise has with AI coding and cybersecurity. Instead people have been inundated with job-loss and existential fears, often propagated by the AI founders themselves.
MP Take: I disagree with the AI-job-loss posture. Jensen Huang is more right that AI will create more jobs than expected. It’s hotly debated — but the founders’ fear-selling is a big part of why mainstream consumers feel antipathy toward AI.
In today’s world an AI startup has to take whatever cool thing it has to billions of users in two or three years — a very different environment than most of my 30-year career, when startups had five to ten years to scale. Think Instagram and WhatsApp without Facebook, or YouTube without Google.
MP Take: That big-company-plus-startup synergy is what made those products what they became. We’re going to see all of that in spades over the next three to five years — no more waiting ten or fifteen.
Two new studies tie the 2007 smartphone rollout to falling global fertility — declining faster than demographers expected, and not just in developed countries. The US is at 1.8-1.9, China and India drifting toward 1.7-1.8, and now even parts of Africa unexpectedly. About 4-5 billion of 8 billion people have smartphones.
MP Take: I’m loath to say it’s all the phones — there are many factors. But the causality data is getting early academic attention, and demographics is destiny: these curves are very tough to change. The ultimate gadget the world spends hours a day with. Worth being thoughtful about.
Google and Nvidia are now viewing Intel as a backup chip manufacturer — and so is Elon, after TSMC (the Fed of the AI Tech Wave) capped growth at about 30% while everyone wants more. Plus Nvidia’s fresh South Korea memory deals with SK Hynix, Samsung and LG. These are commitments of tens and hundreds of billions.
MP Take: Now that the template is set and public-market interest continues, expect this to accelerate — and go global. Intel in the US remains a prime beneficiary. But it’s up the risk/reward curve: if timing goes askew here versus there, it could throw off the precise calculations.
AI Ramblings Daily on AI-RTZ is here to think through AI and reset. Together.
Sunday’s Bigger Picture — AI-RTZ #1110 — Google and Anthropic Help Paint Elon’s SpaceX/xAI IPO Fence — a lot of nuance there ahead of this week’s SpaceX IPO.
Tomorrow — ARD 93 on AI-RTZ #1112.
Thanks for joining us, AI Curious Folk. Stay tuned.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here.)
Subscribe to AI: Reset to Zero for daily AI Ramblings + Sunday Bigger Picture posts.