Content About Stocktwits
No Image Available

Shift Happens — to IBM, Meta & Microsoft this week| ARD #118

Author
Michael Parekh AI: Reset to Zero
13 hours ago
14 MIN READ

Today’s theme: shift happens — especially the unexpected shifts in this AI tech wave. Three events make the point — around IBM, Meta, and Microsoft — each with my Take, then my Overall Take. Three events for the AI Tech Wave.

(1) IBM’s Fortunes Shift Dramatically — the Biggest Stock Drop in Its History

MP TAKE: IBM saw its fortunes shift dramatically — the biggest stock drop in its history, over 25%, a quarter-trillion-dollar company down $50 billion in a day. The culprit is a shift in spending by enterprise customers — away from traditional, legacy hardware, software and services, and toward AI-related budgets. Their results were weaker than expected on revenues, but the more important signal was that budget shift, and the markets have been laser-focused on it. You’ll recall the huge concerns a few months ago over the ‘SaaS-pocalypse’ — whether AI will disrupt the way traditional software and hardware get sold. I wrote a lot about it; it’s a perennial question, and it’ll stay with us not for a few quarters but for years, because a lot of change is coming.

Here’s the thing: none of this is binary. Markets love to see it as this-versus-that, but the answer over time is typically both. We’re going to see hybrid systems — lots of AI built on top of legacy systems — and companies like IBM will have plenty of opportunity to participate. It’s just that, quarter to quarter, these are extraordinarily hard things to execute through, and the markets react on a dime. There’s an irony here worth savoring: IBM is the archetype of mainframe computing in our 50+ year tech history — and it’s now being hit by what I’d call the mainframe stage of AI, all these data centers taking huge amounts of compute and renting it to enterprises for billions. IBM is one of the first. There will be lots of other software and hardware names that get this same attention and volatility as the shift plays out. (It’s also, by the way, one more reason the biggest AI companies prefer to stay private longer — or, now that the capital needs are so vast, trade quarter-to-quarter volatility for access to immense amounts of capital via the trillion-dollar SpaceX/OpenAI/Anthropic IPOs.)

Sources, in narrative order: Bloomberg — ‘We Faltered’: IBM Falls Most Since at Least 1968 on Sales Miss. For longtime readers: ‘Concerns over an AI-led Software-apocalypse’ in AI-RTZ #988.

(2) Meta Shifts Its Strategy on Siting Its AI Data Centers

MP TAKE: Meta has been one of the top spenders on AI data-center infrastructure — $130, $140, now $145 billion this year — with Mark Zuckerberg laser-focused, going mano a mano with Elon Musk on the build-out. But this week they shifted their strategy on siting those data centers. And data centers, as I’ve written a lot, are getting extraordinarily political. New York State just put a one-year moratorium on AI data centers; there’s rising local resistance at the state and county level, for all sorts of reasons — people are afraid of them, worry they’ll take jobs, worry about existential risk — and it’s showing up sharply in this very political year.

So here’s what stood out: Meta put about $40 billion of data centers around Louisiana — and did it with a very sensitive, community-first approach, working closely with local constituencies and getting them, at least publicly, to say favorable things. That deft touch is the shift I wanted to highlight, because it’s something all the hyperscalers — Microsoft, Amazon, Google, and OpenAI and Anthropic too — are going to have to execute. Meta surprised people with the subtlety here, because otherwise they’re pedal-to-the-metal, consequences-be-damned on data centers. A noteworthy shift in strategy.

Sources, in narrative order: The Information — Meta’s Louisiana Purchase. The Verge — New York becomes the first state to enact a data-center moratorium. For longtime readers: ‘US AI Data-Center Build-Out Schedules Are Slipping’ in ARD #89; and ‘Power Plays Begin to Build Gigawatt AI Data Centers’ in AI-RTZ #481.

(3) Microsoft Accelerates Its Shift to Enterprise Customers — with a New Nadella Essay

MP TAKE: Microsoft continues to accelerate its shift to enterprise customers, and away from its core partners and technology providers, OpenAI and Anthropic. They have a leg in each camp — but, as I noted in a post a few days ago, Microsoft is picking a side, just like Palantir. Alex Karp made the vocal argument that enterprises should shirk away from trusting their data and business IP to the big frontier-model companies, who will ‘hoover up’ their domain knowledge and compete with them down the road. So Palantir — and now Microsoft — say: trust us instead. We’ll protect your data and IP; host it on our software and cloud, and we’ll envelop it in security. That’s talking their book, of course — but it’s an ‘Us vs Them’ strategy that’s been employed for thousands of years, and here it’s being done well.

And Satya Nadella is doing it deftly and with determination. He penned an essay titled ‘The Reverse Information Paradox’ — the acronym, not subtly, is RIP: in other words. If you don’t do what I’m suggesting, your data and business knowledge get taken away by the frontier labs, and then, well, rest in peace. It’s a clever, well-written, very accessible essay — worth a read — and it’s falling on the ears of enterprise CEOs, acting as a catalyst for exactly the shift that’s hitting IBM. The tell: Nadella still relies on — and will rely on — OpenAI’s IP and models for the next five years. But at the same time, he is moving to where the money is. That’s an important shift to note.

Sources, in narrative order: Satya Nadella (X) — ‘The Reverse Information Paradox’. For longtime readers: ‘Microsoft Picks a Side vs the Frontier AI Models’ in AI-RTZ #1127; and ‘’Us vs Them’ Comes to Trillion-Dollar Enterprise AI’ in ARD #111.

MP OVERALL TAKE

All three events underline how fast the big tech companies are adjusting their strategies — doing whatever’s needed not to be buffeted by the AI winds blowing through the wave. In one instance — IBM — they’re on the receiving end of very strong winds. In the others — Meta and Microsoft — they’re adjusting to them: to the forces stiffening against data centers, and to the question of whether enterprises trust or distrust the frontier-lab models, trying to navigate a path that serves their interests.

That’s the overall shape of it, and the set of shifts I think will keep happening — quarter over quarter, and over the rest of this decade. In a wave moving this fast, strategy is no longer a five-year plan; it’s a quarterly reflex — and the companies that adjust fastest, on offense and defense, are the ones that keep the lead.

Gadget AI — AI Leads to ‘AI-Defined Vehicles’ (ADVs): Nvidia, Tesla, Google & Many in China

MP Take: Let’s talk self-driving cars again. As excited as I’ve been about them for over two decades, they’ll still take at least half a decade or longer to reach true Level 5 — no wheel, every type of weather. Companies like Tesla show off Level 2++ /ADAS (Advanced Driver-Assistance System)— hands on the wheel, the car more or less driving itself on clear-weather, mapped, defined roads. That gap between where we are and where we’re going is still an enormous, multi-trillion-dollar opportunity, in Tesla’s case and beyond.

But there are other companies making far more strategic investments in hardware, software and open-source AI — and getting less appreciation. Chief among them: Nvidia. My Gadget AI item today is a detailed interview — The Verge’s Nilay Patel on Decoder, talking with Xinzhou Wu, who heads Nvidia’s ‘AI-Defined Vehicles’ (ADV) unit. Nvidia has a very defined hardware roadmap — Orin chips moving to Thor chips, with Alpamayo software — and there’s a lot of rich technical detail here you don’t usually get, right down to the question of ‘do we need Lidar?’ Elon at Tesla has famously said no YEARS ago— cameras only. But the reality is that Lidar is used by everybody except Tesla, and Nvidia’s systems are now being experimented with and deployed by over 80% of auto companies worldwide, including top Chinese automakers like BYD.

So Nvidia remains, to my mind, the most under-appreciated player in self-driving. Everyone keeps focusing on Tesla, Google’s Waymo, and Uber — all important — plus a lot of Chinese companies most US investors don’t track. This interview is worth a read and/or a listen.

Sources, in narrative order: Decoder / The Verge — Nvidia’s head of AI Automotive Xinzhou Wu on ADVs, Lidar and the road to L5 · (YouTube). For longtime readers: ‘Nvidia’s Dense Announcements’ in AI-RTZ #1031; ‘Latest on Google Waymo’ in AI-RTZ #985; ‘Tesla’s Robotaxi & Robots’ Ride’ in AI-RTZ #510; and ‘Bumpier Roads for Self-Driving Cars’ in AI-RTZ.

Questions

Q1 — What was Michael’s biggest surprise from the discussion?

Not a huge surprise, but a telling one: even within Nvidia — the company that makes the chips that make AI go — there’s a compute shortage. Even people in the AI division have to fight for their share, and Jensen himself often has to allocate it. It underlines that we’re in a supply-constrained environment — ‘RAMageddon’ — that will continue for the next two or three years. It’s an endemic reality across every tech company, whether they’re using the compute or, like Nvidia, making and selling it.

Q2 — What’s the biggest self-driving issue that riles Michael the most?

The almost-religious Elon driven division over whether you believe in Lidar or not. It made sense when Elon started arguing this over a decade ago, when Lidar was a very expensive sensor. But since then it’s come down in price by multiples, and much of the technology has shrunk to the chip level — especially in how it’s deployed in China. Lidar chips alongside cameras are must-haves, not nice-to-haves, for the safety and redundancy they add. In fact, New Jersey is on the threshold of requiring Lidar for self-driving cars and all robotaxis; if it passes, that’s a major question mark for Tesla — and neighboring New York is considering similar legislation. Once one of these dominoes drops, Lidar may end up a US requirement — and even Elon, with his no-Lidar belief, may have to change his tune, especially to go beyond Level 3 to Levels 4 and 5, where you need the higher sensor stack for safety and redundancy.

Wrapping up

Today’s AI-RTZ #1147 — Apple’s Unexpected Strengths in Its AI-Chip Roadmap Ahead — is on Apple’s underappreciated strategy for AI chips built on Apple Silicon. With John Ternus becoming CEO in September (the former hardware chief) and Johnny Srouji — the architect of Apple’s Silicon strategy — now the number two, both are accelerating Apple’s server and local chips right in the teeth of all the AI shifts coming. Noteworthy — take a look.

Tomorrow — ARD 119 on AI-RTZ 1148.

Thanks for joining us today, AI Curious Folk. Stay tuned.

— MP

Full Source Reading —

For the broader context, see the canonical sources for ARD 118 — in today’s narrative order:

Event 1 — IBM’s Historic Drop / Enterprise Spend Shift

Event 2 — Meta’s AI-Data-Center Siting Shift

Event 3 — Microsoft / Nadella ‘Reverse Information Paradox’

Gadget AI — ‘AI-Defined Vehicles’ / Nvidia

Clips from today

Clip 1 — Meta’s Subtle Shift in AI Data Centers

Watch on YouTube Shorts

On its ~$145B/year build-out, Meta put ~$40 billion of data centers around Louisiana with a notably local, community-first approach — a shift from pedal-to-the-metal.

MP Take: Multi-gigawatt data centers at $50B-plus per gigawatt take years of local permitting, and resistance is rising — New York just enacted a one-year moratorium. So it’s notable Zuckerberg is using a defter local touch. You can’t rent capacity you can’t get permitted and powered.

Clip 2 — Tesla’s Lidar Dilemma: US Laws Loom

Watch on YouTube Shorts

New Jersey is on the threshold of requiring Lidar for self-driving cars and robotaxis, and New York is considering the same — a potential problem for Tesla’s camera-only bet.

MP Take: The Lidar debate is almost religious. It made sense when Lidar was expensive a decade ago, but it’s dropped by multiples and shrunk to the chip level. If one state domino drops, Lidar may become a US requirement — and even Tesla may have to change its mind beyond Level 3.

Clip 3 — AI’s Impact on Software & Hardware Sales

Watch on YouTube Shorts

IBM just had the biggest stock drop in its history — down over 25%, shedding ~$50 billion in a day — as enterprise budgets shift from legacy IT to AI.

MP Take: IBM is the sharpest data point yet to enterprises tilting their spend toward AI. It echoes the SaaS-pocalypse selloff — likely overdone on the way down (not stock advice). None of this is binary; the answer over time is hybrid, AI built on top of legacy. IBM is the first of many.

Clip 4 — Nvidia’s AI-Defined Vehicles

Watch on YouTube Shorts

Nvidia’s systems are now deployed by 80-plus automakers worldwide, including China’s BYD — an open-source hardware/software platform (Orin-to-Thor chips, Alpamayo software) for ‘AI-Defined Vehicles.’

MP Take: Nvidia is the most under-appreciated player in self-driving — less hyped than Tesla, Waymo or Uber, but building a global platform for automakers everywhere, especially China. A multi-trillion-dollar opportunity — but it’ll still take at least five more years to reach Level 4 and beyond.

About AI Ramblings Daily (ARD), and AI-RTZ

Both are daily. Both are free. Both are about AI. But they’re different mediums carrying different messages.

AI-RTZ is the morning text — a deeper written take on one idea, published by at least 5 AM EST. Today: post #1147.

AI Ramblings Daily is the afternoon video + podcast — my ad hoc takes and perspective on the day’s AI issues & news flow, around 20 minutes, with short 1-2 minute clips for quick topic views. Today: episode #118.

Subscribe to either or both on michaelparekh.substack.com. They run as separate Sections you can opt into or out of.

Links used in today’s show (already embedded inline above; listed here for reference)

Take 1 — IBM’s Historic Drop / Enterprise Spend Shift:

Take 2 — Meta’s AI-Data-Center Siting Shift:

Take 3 — Microsoft / Nadella ‘Reverse Information Paradox’:

Gadget AI — ‘AI-Defined Vehicles’ / Nvidia:

Companion text:

(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.





Want the latest?

Sign up for Michael Parekh's Newsletter below:


Subscribe Here

More like this

BROWSE ALL ARTICLES