The Four Corners & Key Pieces of 2026 AI Chess Board Clearer. ARD #93

Michael Parekh AI: Reset to Zero

9 hours ago

14 MIN READ

Today’s theme: the 2026 AI roadmap for the second half is now far clearer — it’s coming into focus. Why? We now have three mega-AI IPOs all confidentially filed — each a trillion dollars and more — and four major developer conferences over the last week and a half, capped by Apple’s WWDC 2026. Of course, Google I/O 2026, Nvidia’s GTC Taiwan 2026 and Microsoft’s Build Conference in recent days. Those four key companies — Nvidia, Microsoft, Google and Apple — define the four corners of the AI chess board, covering all the global AI user hardware/software surfaces across consumers and businesses. We now understand better, all their North Stars — and in which direction and at what speed they’re all sailing. Three Takes today, each with my Take — and my Overall Take.

(1) OpenAI Officially Files for IPO — Confidentially, of Course

The WSJ phrased in succintly — “OpenAI Files to Go Public” — with CNBC noting OpenAI has “not decided on timing yet.” Just like SpaceX and Anthropic last week — completing the trillion-dollar trio. Sam Altman flagged a few things still being sorted; timing is “some time this fall,” September to December. Anthropic and OpenAI are racing hardest, wood behind the arrow of the enterprise and AI coding market with the best product-market fit. And Elon comes on strong with the $1.75 trillion SpaceX IPO this week — having baked in a $60 billion acquisition for Cursor, the largest private-tech purchase on record. Anthropic makes three is in AI-RTZ #1105, Elon’s whitewashed fence in AI-RTZ #1110, and SpaceX/xAI’s boundless AI ambitions in AI-RTZ #1066.

MP Take: OpenAI is the best-known AI brand with ChatGPT — the fastest-growing consumer software product to date. Despite a plethora of unique challenges, it’ll give its public debut its best-laid plan — fusing ChatGPT and Codex, wrapping in super-app features, before the final IPO outing. Not as laser-focused on the enterprise opportunity as sibling rival Anthropic — still the Coke vs Pepsi, even though Pepsi’s been running ahead of Coke of late. And SpaceX/xAI — the one with the relatively least risk/reward metrics — is up first with the largest offering this week.

(2) Microsoft’s Build + Nvidia’s GTC — the Bookends of the Global AI Market

The Information ran the piece — “Microsoft’s AI Independence Day.” Nvidia supplies ~90% of AI GPUs (and now CPUs and memory), reaffirming its Vera Rubin roadmap in full production at GTC Taiwan, plus leading open-source bets like its Nemotron Ultra 3 model. Microsoft, post its iconic $13 billion+ (now over $200 billion) OpenAI stake, finally set a clearer independent AI strategy under aqui-hire Mustafa Suleyman — its own large and small models, with Copilot across every AI PC (Qualcomm, Intel, AMD, and now Nvidia’s DGX Spark). Microsoft & Nvidia chart their AI strategies is in ARD #87, and Nvidia and Microsoft focus on AI together in AI-RTZ #1104.

MP Take: Microsoft has five more years to leverage its access to OpenAI IP and models — while investing hundreds of billions turbo-charging Azure, now partnered with both OpenAI and Anthropic, and building its own ‘good enough’ models, large and small. An aggressive strategy to roll out Copilot across every Windows computer in the world, now with Nvidia AI chips— one of the biggest hardware surfaces, next to Google and Apple — going up against Amazon AWS above and Google Cloud behind. They’ve reset their AI stage and playbook post OpenAi. Now the new game is on again.

(3) Apple & Google Bring Up Comprehensive Consumer AI — Cloud to Edge

Bloomberg had the backgrund read — “Inside Apple’s secret meeting that led it to take AI seriously.” Apple’s WWDC confirmed the Google Gemini partnership plus its own large and small foundation models — an AI Siri that’s utilitarian, designed for mainstream users with high attention to privacy, trust and safety, monetized via hardware + services across 2.5 billion+ Apple devices. Google’s I/O: Gemini across 3.5 billion+ users and half a dozen billion-user services (Gmail, YouTube, Docs). Apple & Google roll out AI strategies is in ARD #67, why Google is still in consumer AI pole position in AI-RTZ #1096, Apple’s AI bounce picks up steam in AI-RTZ #978, and the race for consumer AI agents in ARD #70.

MP Take: Apple and Google have the most interesting ‘Frenemies’ opportunity around AI — with both their individual and joint strategies. With immediate impact on billions of mainstream users about to experience AI search, chat and reasoning agents for the first time — across products and services that are deep everyday habits. Both sparing no expense. Google, in my mind, remains the most pole-positioned company on the consumer AI side, especially on cloud-based algorithmic software — with Apple a close partner on search and consumer agents. The consumer AI game is indeed afoot. Two of them, Nvidia and Apple, can lead the US charge on Open Source AI vs China in particular this AI Tech Wave.

MP OVERALL TAKE

These four companies — Nvidia, Microsoft, Google and Apple — touch the cumulative hardware and software surfaces for over 5 billion people doing anything with computing and AI, from search-and-ask to do-functions with AI agents across phones, devices and increasingly wearables.

MP Take: So assume at least 10 AI agents per user — that means over 50 billion machine-to-machine AI-token (compute) surfaces, agents and people all working with and against each other to do things. A whole new market that requires a whole new internet — rebuilding computers and smartphones from the bottom up, as Jensen keeps saying. We’re talking trillions of dollars not just to put up data centers, but to rebuild all the components — every company has to rethink operating systems we’ve designed for 50 years for humans, now for agents. Right now we’re just bolting it on. That’s the core opportunity and challenge — with billions and trillions of dollars of new startups ahead. These four companies define the global outline of the AI market; Amazon, Meta, Elon’s SpaceX/xAI, OpenAI and Anthropic are the important other dots that connect the picture. A much clearer roadmap than we had a couple of weeks ago.

Gadget AI — Apple’s AI Requires 12GB+ of On-Device RAM, smartphones & Macs

9to5Mac has it — “iOS 27’s most powerful on-device AI requires iPhone 17 Pro or iPhone Air.” Apple famously makes its newest software work for free as far back as possible on older devices — but the most powerful AI features need 12GB of RAM, gating them to the iPhone 17 Pro and iPhone Air. The base iPhone 17 starts at 8GB; even MP’s favorite value MacBook Neo at 8GB won’t run the latest Apple Intelligence. Only M3 chip Macs and M4 chip iPads make the cut. Apple still best positioned on global supply chain here. Memory chip supply constraints and price hikes is in AI-RTZ #951.

MP Take: Memory continues to be the biggest local constraint, and we saw it directly at WWDC — Apple had the fewest products able to run its most powerful AI. SK Hynix, Micron and Samsung have retasked capacity to higher-margin data centers, so prices are up 10-30% and rising. Nvidia and Apple have locked up supply at the best prices at scale for the next two or three years. So if you’ll need a computer in the next year, buy now rather than wait — and get the higher-memory config; you’ll need it for the AI features. The same RAM challenge hits Windows and Android too.

Questions

Q1 — What is MP’s favorite part of the WWDC 2026 announcements?

Apple’s new Shortcuts app for ‘vibe tasks’ (TechCrunch). I’m a bit of a geek — I’ve used Apple Shortcuts for years; they were agents before agents were cool. Now they’ll be AI-Siri-driven: you’ll be able to vibe-talk and vibe-create shortcuts (and Safari extensions). Just say “send this message to my wife every day” and it automates it — no deep instructions needed. There’s a beta out; I’ll try it on my extra devices and report back.

Q2 — What is the one new Apple AI feature MP would NOT use out of the gate?

Apple Intelligence’s wholesale password management (Gizmodo). In iOS 27 it can scan all your passwords for ones compromised by internet leaks, then — with your input — update them all with far stronger ones managed in your passkeys. I’d trust Apple most here versus Google and everyone else, given their hardware-software integrity. But there’s a nightmare scenario: if it’s hijacked by bad guys, all my passwords change one morning, and I’m left contacting dozens of companies to undo it. I’m sure they’ve red-teamed it — but that’s the one feature I want to understand fully before using.

Source Reading — For the Full Context

For the full context, see the canonical sources:

Shorts Clips from today

Clip 1 — Elon’s $60B Cursor Coding Buy

Watch on YouTube Shorts

One of Elon’s boundless AI ambitions is AI coding — and he’s baked in a $60 billion acquisition for Cursor, the one independent AI coding company, the largest purchase price of any private tech company on record. Contingent on the SpaceX IPO succeeding (otherwise a $10B walkaway fee). In exchange, Cursor gets excess capacity from Colossus — the same data centers Elon rented to Anthropic ($40B) and Google ($30B) to paint his Tom Sawyer IPO fence.

MP Take: These are the times of big mega numbers with T’s instead of B’s — I remember when they were M’s. SpaceX/xAI is up first this week with the largest offering, the one with the relatively least risk/reward metrics of the trio.

Clip 2 — Apple AI Changes Your Passwords

Watch on YouTube Shorts

Apple Intelligence in iOS 27 this fall can manage all your passwords — Sherlocking the 1Password-style apps — scanning for which have been compromised by internet leaks, then with your input updating them all with far stronger ones managed in your passkeys.

MP Take: I’d trust Apple most here versus Google and everyone else — they have the hardware-software integrity to make passwords more secure than any third party. But it’s the one feature I won’t use out of the gate. The nightmare scenario: bad guys hijack it and change all my passwords one morning, leaving me to contact dozens of companies to undo it. I hope they’ve red-teamed it — that’s what I want to understand first.

Clip 3 — AI Agents: A New Computing Era

Watch on YouTube Shorts

Four companies — Nvidia, Microsoft, Google and Apple — touch the hardware and software surfaces for over 5 billion people. Assume at least 10 AI agents per user, and that’s over 50 billion machine-to-machine AI surfaces — agents and people all working with and against each other to do things. A whole new market.

MP Take: It requires a whole new internet — rebuilding computers and smartphones from the bottom up, as Jensen keeps saying. Every company has to rethink operating systems we’ve designed for 50 years for humans, now for agents. We’re just bolting it on right now. That’s the core opportunity and challenge — and billions, trillions of dollars of new startups ahead.

Clip 4 — Microsoft’s AI Independence Day

Watch on YouTube Shorts

At Build, Microsoft finally set its roadmap beyond its iconic $13B+ OpenAI partnership — now worth over $200B in equity on its balance sheet. But that’s not what’s important to Satya: he needs a clearer, independent AI strategy. Even with access to OpenAI’s models and IP for the next five years, he still has to build his own large and small models, under aqui-hire Mustafa Suleyman.

MP Take: Microsoft wants a plethora of AI PCs with Copilot across every machine — Qualcomm, Intel, AMD, and now Nvidia’s DGX Spark chips, with Jensen leading the charge in a just-announced partnership. Baked into Azure, going up against Amazon AWS above and Google Cloud behind. They’ve reset their stage and playbook. Now the new game is on again.

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

Today’s AI-RTZ #1112 — AI Siri Ready for Takeoff This Fall — Apple’s WWDC 2026 AI Siri, powered by Google Gemini engines. A lot more detail there on what I discussed today.

Tomorrow — ARD 94 on AI-RTZ #1113.

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.)