The Battle for AI Resources Accelerates. ARD #103

Today the theme is the battle for AI resources accelerating — whether it’s for AI talent, AI compute, or sovereign AI share and access. Three events I’d like to dig into for the AI Tech Wave — each with my Take first, then my Overall Take.

(1) Frontier AI Labs Accelerate the Talent Musical Chairs vs Incumbents Like Google

MP TAKE: This is part of the AI Talent race that went into high gear, driven by Meta last year. The concrete tell this week: in a matter of days Google DeepMind lost Noam Shazeer — co-author of the seminal 2017 “Attention Is All You Need” paper that gave us the Transformer, the ‘T’ in ChatGPT — to OpenAI, and Nobel laureate John Jumper (of AlphaFold) to Anthropic, while Nvidia acqui-hired the Essential AI team, including another Transformer co-author, Ashish Vaswani.

OpenAI and Anthropic are logically taking advantage of their frontier-lab ‘mythos’ (pun intended) and their imminent mega-AI IPOs later this year to top up with the best AI researchers. This is the 2026 version of the frenetic AI-talent musical chairs game that Meta and its founder/CEO Mark Zuckerberg kicked off. Beyond the financial rewards in an upcoming IPO, both companies likely also offer new top-level talent bigger allocations of AI compute for their favorite projects, with fewer friction points. Larger companies like Google, Meta and Microsoft have huge allocations locked up serving billions of mainstream customers each — so even at their scale, there’s relatively less to go around for green-field bets. And AI researchers are drawn to those green fields, to other world-class talent, and to the chance to build new AI applications from scratch. To widen the lens: Axios notes the deeper driver is timing — many in the field believe there’s a short window before models start improving themselves (’recursive self-improvement’), and everyone wants to be at the lab they think will win. Expect this AI-talent musical chairs to continue through this year at least.

Sources, in narrative order: Axios — AI lab musical chairs hits Google the hardest. For longtime readers: the AI Talent race for the Mag 7s spurred by Meta in AI-RTZ #769.

(2) SpaceX Provides Open-Source AI Compute to Reflection AI, a Core Nvidia Partner

MP TAKE: Elon Musk continues to tactically expand what I call his ‘EWS’ — Elon Web Services, modeled loosely after Amazon’s AWS — renting out the excess capacity of his AI-data-center infrastructure. After signing Anthropic and Google to a combined $70-billion-plus in AI-infrastructure deals — cancellable at short notice by all parties — Elon is extending the same opportunity to Reflection AI. This new AI startup is backed by core investor Nvidia, which keeps leaning into open-source AI opportunities in the US versus China.

This move is a win/win/win for all three parties — Elon, Jensen and Reflection — providing a path to everyone’s long-term objectives while making financial sense for all in the short term: Elon monetizes idle compute, Jensen advances his open-source-champion thesis, and Reflection gets the compute to compete (reportedly around $150 million a month, roughly a $6-billion deal over two-to-three years, as it trains its first US-based open-source model). Widen the lens, though: these are short-notice, cancellable arrangements, so the ‘win’ holds only as long as the incentives stay aligned. Expect more of the same.

Sources, in narrative order: Axios — Open-source AI gets more compute from SpaceX/xAI. Reuters — SpaceX lands Google AI-compute deal after the Anthropic pact ahead of its IPO (combined $70B-plus). For longtime readers, in narrative order: Nvidia’s accelerating role as AI computing kingmaker in AI-RTZ #1036; and how Nvidia and Apple can be the global US open-source AI champions vs China in AI-RTZ #1089.

(3) AI Now Sits at the Same Table as Heads of State

MP TAKE: This is a notable development — tech heads sitting at the same table as heads of state. The scene at this week’s G7 in the French Alps was striking: President Trump flanked by OpenAI’s Sam Altman and Google DeepMind’s Demis Hassabis; host Emmanuel Macron flanked by Anthropic’s Dario Amodei and Salesforce’s Marc Benioff; AI CEOs posing for bilateral photos in front of national flags, in the chair usually reserved for a president, prime minister or chancellor.

We have not seen this version of events in prior tech waves — not even the internet one in the nineties. It highlights the broad global recognition of the bigger imperatives around the AI Tech Wave, with far more unknowables than knowables. And no head of state wants to be caught without a positive, proactive, and responsive AI story. The AI leaders themselves struck a careful note — Altman warning the room “not to cede your responsibilities to AI labs like mine,” and Amodei urging democracies to “resist the temptation to splinter.” Think of the Anthropic-versus-Washington clash as a small test run of this dynamic. Again, a trend likely here to stay as the global AI musical-chairs game continues.

Sources, in narrative order: Axios — New global order: AI CEOs as heads of nation-states. For longtime readers: the alluring global ‘Sovereign AI’ market in AI-RTZ #732; and Mistral, a strong AI wind from France, in AI-RTZ.

MP OVERALL TAKE

All three events above highlight the ongoing, accelerating competition for AI resources — from AI Talent to AI Compute to AI Share and Access among countries. And we haven’t even discussed the other AI resources in global short supply: power, memory, and the broader chip-manufacturing capacity and supply chains. All of it makes for a far more intense competition for supply-constrained resources than tech waves past — with a lot more zeroes attached, and real implications for whoever ends up holding the shorter straw.

To widen the lens — scarcity also forces cooperation. The talent, compute and sovereign deals above are as much about who you partner with as who you beat. It’s frenemies accelerated: everyone needs to cooperate with each other even as they compete tooth and nail on talent, compute and sovereign deals. For now, a trend with more tailwinds than headwinds — and one we’ll keep tracking.

Gadget AI — Meta’s Updated AI Smart Glasses, Now Meta-Branded at $299

MP Take: These are just the beginning salvo of AI smart-glass choices — not only from Meta, but Google later this year, and likely Amazon. Apple is in the mix too, though potentially slated for next year. The questions are going to be cost, style and fit, AI and tech features, and of course where on the privacy/trust spectrum each vendor’s offering falls. For now, my ranking from least to highest potential public trust remains Meta, Google, Amazon and Apple. OpenAI is potentially in the mix (with Jony Ive), with no officially announced products yet — and smaller players like Snap are trying to get into the mainstream distribution game. A milestone, with 7-million-plus units already sold by Meta and others, but still early.

Sources, in narrative order: Meta — Introducing Meta Glasses. Bloomberg — Meta debuts glasses under its own brand at a lower $299 price. The Verge — Meta launches cheaper AI smart glasses without Ray-Ban. Wired — Meta’s very own smart glasses go on sale today for $299. Also: Wired — hands-on with Google’s upcoming Android XR / Gentle Monster glasses. For longtime readers: Meta leans in on AI smart glasses in AI-RTZ #849.

Questions

Q1 — What’s MP’s most favorite use for AI smart glasses today?

The open-ended Bluetooth headphones — being able to listen to audio without wearing in-ear buds, which is genuinely nice when you’re walking around. That said, I always end up defaulting back to my Apple AirPods and ecosystem. I still haven’t found my killer use case for smart glasses — they’re absolutely interesting to try, but I’m not quite there.

Q2 — How much does MP use the cameras on the glasses?

Not very much — to me it’s a novelty. The privacy issue keeps coming up. I don’t feel it’s appropriate taking pictures and videos with the glasses even with the indicator light, and I end up reverting to my smartphone. With a phone, it’s clear to everyone you’re taking a picture or video — so you have the implicit social permission. AI smart glasses haven’t yet earned that societal permission — even with 7-million-plus units sold by Meta and others. That’s something that will need to be resolved over the next year or two as millions of people try these as a gadget.

Full Source Reading —

For the broader context, see the canonical sources for ARD 103 — in today’s narrative order:

Event 1 — Frontier AI Labs Accelerate the Talent Musical Chairs

• Axios — AI lab musical chairs hits Google the hardest

• AI-RTZ #769 — The AI Talent race for the Mag 7s spurred by Meta

Event 2 — SpaceX Open-Source Compute to Reflection AI

• Axios — Open-source AI gets more compute from SpaceX/xAI

• Reuters — SpaceX lands Google AI-compute deal after the Anthropic pact ahead of its IPO

• AI-RTZ #1036 — Nvidia’s accelerating role as AI computing kingmaker

• AI-RTZ #1089 — How Nvidia and Apple can be the global US open-source AI champions vs China

Event 3 — AI at the Same Table as Heads of State

• Axios — New global order: AI CEOs as heads of nation-states

• AI-RTZ #732 — The alluring global ‘Sovereign AI’ market

• AI-RTZ — Mistral, a strong AI wind from France

Gadget AI — Meta’s $299 AI Smart Glasses

• Meta — Introducing Meta Glasses

• Bloomberg — Meta debuts glasses under its own brand at a lower $299 price

• The Verge — Meta launches cheaper AI smart glasses without Ray-Ban

• Wired — Meta’s very own smart glasses go on sale today for $299

• Wired — Hands-on with Google’s upcoming Android XR / Gentle Monster glasses

• AI-RTZ #849 — Meta leans in on AI smart glasses

Shorts Clips from today

Clip 1 — My Favorite Use for AI Smart Glasses

Watch on YouTube Shorts

MP’s favorite use for AI smart glasses today isn’t the camera or the AI assistant — it’s the open-ear Bluetooth headphones. Audio without wearing in-ear buds is genuinely nice when walking around.

MP Take: It’s the one feature I actually reach for — though I still end up defaulting back to my Apple AirPods and ecosystem. I haven’t found my killer use case for smart glasses yet, but the open-ear audio is the closest.

Clip 2 — AI Talent War: Google’s Brain Drain

Watch on YouTube Shorts

In a matter of days, Google DeepMind lost Noam Shazeer — co-author of the 2017 “Attention Is All You Need” paper that gave us the Transformer — to OpenAI, and Nobel laureate John Jumper (of AlphaFold) to Anthropic. Nvidia also acqui-hired the Essential AI team.

MP Take: OpenAI and Anthropic are using their frontier-lab mythos and imminent mega-AI IPOs to top up with the best researchers. They’re drawn not just by the money but by green fields, world-class peers, and bigger compute allocations with less friction than at the billion-user incumbents.

Clip 3 — Meta’s $299 AI Smart Glasses

Watch on YouTube Shorts

Meta launched cheaper, own-branded AI smart glasses at $299 — no Ray-Ban — with Google’s range coming later this year and Amazon and Apple in the mix. The features roughly match the Gen 2 Meta glasses, minus the display.

MP Take: These are just the beginning salvo of AI smart-glass choices. The questions will be cost, style and fit, AI features — and where on the privacy/trust spectrum each vendor falls. My ranking from least to highest potential public trust: Meta, Google, Amazon, Apple. A milestone, with 7M-plus units sold, but still early.

Clip 4 — AI Talent War: The Pre-IPO Talent Grab

Watch on YouTube Shorts

OpenAI and Anthropic — both poised for trillion-dollar-plus mega-AI IPOs this year — are taking advantage of the moment to grab top AI talent from incumbents, the second phase of the race Meta kicked off in 2025 with its $14B Scale AI acqui-hire.

MP Take: Beyond the financial rewards ahead of an IPO, these labs offer bigger compute allocations and fewer friction points for researchers’ favorite green-field projects. The deeper driver, per Axios, is timing — many believe there’s a short window before models start improving themselves, and everyone wants to be at the lab they think will win.

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 #1126.

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 #103.

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 — Frontier AI Labs Accelerate the Talent Musical Chairs:

• Axios — AI lab musical chairs hits Google the hardest

• AI-RTZ #769 — The AI Talent race for the Mag 7s spurred by Meta

Take 2 — SpaceX Open-Source Compute to Reflection AI:

• Axios — Open-source AI gets more compute from SpaceX/xAI

• Reuters — SpaceX lands Google AI-compute deal after the Anthropic pact ahead of its IPO

• AI-RTZ #1036 — Nvidia’s accelerating role as AI computing kingmaker

• AI-RTZ #1089 — How Nvidia and Apple can be the global US open-source AI champions vs China

Take 3 — AI at the Same Table as Heads of State:

• Axios — New global order: AI CEOs as heads of nation-states

• AI-RTZ #732 — The alluring global ‘Sovereign AI’ market

• AI-RTZ — Mistral, a strong AI wind from France

Gadget AI — Meta’s $299 AI Smart Glasses:

• Meta — Introducing Meta Glasses

• Bloomberg — Meta debuts glasses under its own brand at a lower $299 price

• The Verge — Meta launches cheaper AI smart glasses without Ray-Ban

• Wired — Meta’s very own smart glasses go on sale today for $299

• Wired — Hands-on with Google’s upcoming Android XR / Gentle Monster glasses

• AI-RTZ #849 — Meta leans in on AI smart glasses

Q1 + Q2 — MP’s favorite use + camera use on AI smart glasses:

• (no external sources — MP’s own analyst view)

Companion text:

• AI-RTZ #1126 — Meta’s $900 million acquihire in India for WhatsApp

AI Ramblings Daily on AI-RTZ is here to think through AI and reset. Together.

Today’s AI-RTZ #1126 — Meta’s $900 million acquihire in India for WhatsApp — Meta is paying around $900 million to acqui-hire an Indian entrepreneur to head up WhatsApp, the app that serves over three billion people worldwide — with broad implications for Meta’s AI distribution. Recommended as today’s reading post.

Tomorrow — ARD 104 on AI-RTZ 1127.

Thanks for joining us today, 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.