‘Doing What It Takes’ in AI. OpenAI, Meta & TSMC. ARD #185
…OpenAI’s ad sprint, Meta’s open Muse on Apple’s Macs, TSMC on its own terms, and Apple’s Ternus era.
It is Thursday, and today’s theme is ‘doing what it takes’, on AI priorities big and small. Putting aside some rivalries in some cases, but everyone keeping an eye on the prize, and executing tactically, step by step. In this AI Tech Wave, everyone is doing what it takes. But each on its own terms.
Three companies today: OpenAI, competing aggressively with its frontier model archrival Anthropic ahead of their mega AI IPOs, and leaning into ads again; Meta, going open again with a frontier-class model, and surprisingly teaming up with an old rival, Apple, to run it locally on a desk; and TSMC, what I call the ‘Fed’ of the AI world, posting a sixth straight record quarter, and working things out with Elon on his Terafab ambitions on its own terms. Then the Gadget AI, where Apple’s new CEO, John Ternus, makes the cover of TIME.
(1) OpenAI leans into ads, as Anthropic sprints to its IPO.
OpenAI is accelerating its ad business, with Meta hires and slashed prices, per The Information in ‘OpenAI Accelerates Ad Sprint With Meta Hires, Slashed Prices’. Ads were a side project when OpenAI started showing them in ChatGPT in February. Not anymore. About 140 people now work on ads at OpenAI, nearly half of them hired since July, with Meta the biggest source of talent.
The numbers are jumping. Advertisers on ChatGPT went to about 60,000 in early October, up from about 20,000 in late August. Prices are way down: a click on a ChatGPT ad costs about $5, versus $30 to $50 on Google for the same business software clients. Ads hit about a $1 billion annual run rate in August, still small next to OpenAI’s overall run rate of nearly $50 billion, and next to the ad kings, Google and Meta. The other side: the ad team has reorganized seven times since February, and may miss the $2.4 billion in 2026 ad revenue OpenAI projected to investors.
Meanwhile, Anthropic showed investors a $16.8 billion quarter, up nearly 50 percent, per The Information in ‘Anthropic’s Rapid But Slowing Growth Could Weigh on IPO Valuation’. Much more on that tomorrow.
MP Take: OpenAI is doing what it takes, and it is doing all of it at once. Back to AI coding and the enterprise against Anthropic, where Codex and ChatGPT are blending together. AI agents, with Dots, launched at its Dev Day earlier this month. And now ads again. Ads were a side quest under Fidji Simo, the prior head of applications, set aside a few months ago when Anthropic was running away with the AI coding and enterprise prize. Now they are back in focus, something I flagged when OpenAI first tested ads in RTZ #993.
Why it matters: ChatGPT is already at about 1.2 billion weekly users, and less than 5 percent of that is subscription based. So a lot of the rest is going to be ad driven, the way Google and Meta were built. Slow and steady progress, shown ahead of Anthropic’s pre-Thanksgiving mega AI IPO, as I discussed in AI-RTZ #1228. The competition between the two continues in earnest, and both are doing what it takes.
Sources: The Information, ‘OpenAI Accelerates Ad Sprint With Meta Hires, Slashed Prices’ · The Information, ‘Anthropic’s Rapid But Slowing Growth Could Weigh on IPO Valuation’.
For longtime readers: AI-RTZ #1228, ‘Anthropic’s IPO Go Time, OpenAI’s $1.4 Trillion, Gemini 4 & More’ · RTZ #993, ‘OpenAI tests its ads in ChatGPT’.
(2) Meta goes open again, and teams up with Apple.
Meta has released Muse Spark Open, which it calls the most capable American open-weight model to date, per Meta in ‘Introducing Muse Spark Open, The Most Capable US Open-Weight Model to Date’. It is big: 816 billion total parameters, 42 billion active, built for AI coding and agents. Alexandr Wang says it is not as advanced as Muse Wonder, Meta’s closed frontier model from Tuesday.
The twist: Meta worked with Apple, so it runs on two of Apple’s top-end M5 Ultra Mac Studios. On your desk, no cloud. That is per CNBC in ‘Apple and Meta team up to make Muse AI model run on Mac computers’, and Axios in ‘Apple and Meta bury the hatchet, for AI’. It is the first public partnership between the two in years. Tim Cook and Mark Zuckerberg traded barbs for years over privacy, and Meta hired Apple’s head of design and dozens of its people. Now they are making up a bit under Apple’s new CEO, John Ternus, with a former Apple executive now at Meta, Daniel Gross, acting as the diplomat. The catch: each Mac Studio starts at about $10,800, and you need at least two.
MP Take: this is good to see. Meta is going open again. Remember, it was the US champ of open source with Llama, then made a big retrenchment, with founder CEO Mark Zuckerberg spending $14 billion on Scale AI to bring in Alexandr Wang. In about 14 months, the two have put together the Muse AI agent move I discussed in AI-RTZ #1238. Now it wants to be both open and closed, across the spectrum.
The Apple move is timely and clever. Open source is something everyone wants to run on local computers. Nvidia and Microsoft are competing with their hardware, as I noted in AI-RTZ #1104, and AMD just announced its own, at about $1,500 for local computing. So this is going to be a big race over the next two years. Meta doesn’t have the hardware, so it partners with Apple. A pair of Mac Studios is relatively inexpensive next to cloud AI compute bills that can run thousands of dollars a month. And locally, Meta never sees your prompts, sidestepping the old privacy fight.
Practical collaborations are important. I’d like to see Nvidia and Apple bury the hatchet like this as well, after their decade-old feuds. Meta needs distribution for its Muse Spark models relative to everybody else, especially the open source models from China, as I discussed in AI-RTZ #1233. Meta, led by its founder CEO, doing what it takes, as usual.
Sources: Meta, ‘Introducing Muse Spark Open, The Most Capable US Open-Weight Model to Date’ · CNBC, ‘Apple and Meta team up to make Muse AI model run on Mac computers’ · Axios, ‘Apple and Meta bury the hatchet, for AI’.
For longtime readers: AI-RTZ #1238, ‘Meta’s ‘Wonder Full’ Muse Joins the AI Frontier’ · AI-RTZ #1233, ‘Nvidia’s Open Source ‘AI Factory’ Push vs China’ · AI-RTZ #1104, ‘Nvidia & Microsoft focus on local AI computers’ · AI-RTZ #1107, ‘Meta tries mixed AI Agents for consumers & businesses’ · AI-RTZ #1161, ‘Nvidia Assembles AI ‘Avengers’, Open Secure AI Alliance’.
(3) TSMC delivers again, on its own terms.
TSMC posted a record profit for a sixth straight quarter, beating estimates on the AI boom, per CNBC in ‘TSMC posts record profit for a sixth straight quarter, beating estimates on AI boom’. Net profit jumped nearly 70 percent from a year ago, on revenue of about $46.5 billion, up 41 percent. It guided the fourth quarter higher still, to $52 billion to $54 billion, and said spending on new plants from 2026 to 2028 will be significantly higher than the past three years. It is adding another 3 nanometer plant in Taiwan as AI orders pile up, per The Information in ‘TSMC to Add a New 3nm Fab as AI Orders Pile Up’. And it is sending teams to check that its customers can actually power and finance the data centers for all those chips.
TSMC is also in talks with Elon’s SpaceX and xAI on their Terafab project in Texas. But CEO C.C. Wei says TSMC has to run any plant, per Bloomberg in ‘TSMC Open to Working With Terafab But Must Control Plants’. That matters, because Elon has been saying his companies will build and run the fab, no pun intended. TSMC can, at most, sublease a part. So the two are a little separate on the bid and the ask.
MP Take: TSMC is doing what it takes, on its own terms, as usual. I have long called it the ‘Fed of Tech and AI’, as in ARD #120’s section ‘The ‘Fed of the AI Tech Wave’ Reports Strong Numbers and Capacity Hikes’, and in AI-RTZ #1108. It sets the pace for the whole industry, as we discussed just yesterday with ASML’s orders out to 2030 in ARD #184. Its top customers, Nvidia, Apple and others, all need more capacity from TSMC.
It threads the needle, the pragmatism of how it has run its business for a long time, against the realities of the demand curve in a supply-constrained AI world. That should go on for the rest of the decade, for memory chips from SK Hynix, Samsung, Micron and CXMT too. ChatGPT, Gemini, Muse Spark, Apple’s new Siri, Alexa Plus: all of it needs TSMC chips by the trainload. And with Elon, it is TSMC’s way or not at all, even for the world’s richest man.
Sources: CNBC, ‘TSMC posts record profit for a sixth straight quarter, beating estimates on AI boom’ · The Information, ‘TSMC to Add a New 3nm Fab as AI Orders Pile Up’ · Bloomberg, ‘TSMC Open to Working With Terafab But Must Control Plants’.
For longtime readers: ARD #120, ‘Drumbeats of AI Progress March On’ · AI-RTZ #1108, ‘Taiwan Semi, ‘Fed of the global chip economy’, Speaks’ · ‘It all begins with TSMC in Taiwan’ · RTZ #1034, ‘Elon Musk boosts SpaceX/xAI pre-IPO story with ‘Terafab’’.
My overall take
MP Take: everyone is doing what it takes. They are making friends with old rivals, and executing on the key priorities. At OpenAI and Meta, doing what it takes means more of everything: ads, agents, coding, compute, open and closed, even partnering with old rivals. At TSMC, it means growing only on its own terms. And at Apple, as you’ll see below, it means picking what matters most.
The execution challenges in this AI Tech Wave are bigger than anything I’ve seen in the 30 to 40 years I have been following technology. And these companies are paying attention to the details. With mega AI IPOs weeks away, the markets are watching which approach pays. We will see where this goes.
Gadget AI: Apple’s Ternus era, on TIME’s cover.
Apple’s new CEO, John Ternus, got a cover story in TIME, in ‘John Ternus Is Engineering a New Era for Apple’, part of its Best Inventions of 2026. It is a rite of passage. Ternus took over as CEO from Tim Cook on September 1, after Tim’s 15 years in the job, and Cook stays on as executive chairman. Ternus has been at Apple for 25 years, half of Apple’s age of 50. He is a hardware engineer, and he co-filed a foldable phone patent back in 2011. Now that phone is here: the iPhone Duo, from $1,999, goes on sale tomorrow, with deliveries on the 23rd.
On AI, Apple isn’t chasing the frontier. The new Siri uses Apple’s own model, built on Google’s Gemini, on device and in Apple’s Private Cloud Compute. Apple’s bet is personal, private AI, across about 2.5 billion active devices. And Ternus calls the memory chip shortage a crisis, with all options on the table.
MP Take: more importantly than any specific product, this profile gives you more insight into how Ternus is organizing himself, his team and his company for the AI opportunity. This is a younger fellow who is very earnest about making sure Apple leverages its platforms and ecosystems, hardware and software. And most importantly, its huge global supply chain advantage, the hard work of Tim Cook, who remains chairman of the board. Ternus has Cook helping him through the geopolitical issues and the head of state relationships, whether at the White House or in China. That means Ternus can focus on the things he needs to focus on, and grow into those additional responsibilities over time.
As I said when his era began in AI-RTZ #1196, AI is job one for Ternus. In his own words to TIME, ‘you can’t do everything’, so Apple picks the things that matter most. Worth a read if you are interested in all things Apple.
Sources: TIME, ‘John Ternus Is Engineering a New Era for Apple’.
For longtime readers: AI-RTZ #1196, ‘Apple’s Ternus Era Begins, with AI as Job One’ · RTZ #1063, ‘Long Expected Apple Cook to Ternus CEO shift Activated’ · RTZ #975, ‘New Bounce for Apple Design in AI Age’.
Q&A
What is the one thing I am looking forward to in the Ternus era, over the next 18 months?
The evolution of local, on-device AI across the Apple ecosystem of platforms, whether computers, wearables or of course the iPhones. Both on the devices, and across Apple’s privacy-optimized Private Cloud Compute. I am particularly focused on whether Apple provides better than average support for third-party developers, to build AI applications and services on top of Apple’s hardware and software, with attention to the privacy and trust of Apple’s 2 billion plus users globally. That is a unique opportunity. Only Google, or possibly Meta, could do the same, and Apple has the pole position. Today’s Meta news, with open models running locally on Mac Studios, is a good tactical start.
What is the one thing that could hold Apple back versus its rivals?
As usual for any company, not moving as fast as needed, not doing what it takes. True AI autonomy is needed at the senior levels of every one of Apple’s businesses, giving those leaders almost CEO-level authority, or at least the attention of the CEO. Ternus needs to ramp that up, because he is going up against founder CEOs: Mark Zuckerberg, for whom no detail is too small; Elon Musk, the same; and Jensen Huang at Nvidia, 33 years as founder CEO. The stakes, to do what it takes, are much higher in this AI Tech Wave.
Clips from today
Meta and Apple Bury the Hatchet
Meta’s new open model, Muse Spark Open, runs on Apple’s Macs. Old rivals, working together again.
MP Take: Meta hired Apple’s head of design and dozens of its people, so there has been a rivalry. Now they are making up a bit under Apple’s new CEO, with a former Apple executive at Meta acting as diplomat.
TSMC Delivers, On Its Own Terms
A sixth straight record quarter, and on Elon’s Terafab, TSMC has to run any plant.
MP Take: everyone needs more capacity from TSMC, the Federal Reserve of the AI wave. It is doing what it takes, on its own terms.
Apple’s New CEO on TIME’s Cover
John Ternus, 25 years at Apple, gets a TIME cover as the iPhone Duo goes on sale.
MP Take: more important than any single product, the profile shows how he is organizing himself, his team and his company for the AI opportunity.
Ternus and Cook, Apple’s Tag Team
Tim Cook, as chairman, helps Ternus on the geopolitics, from the White House to China.
MP Take: that frees Ternus to focus, and grow into the rest over time.
Today’s AI-RTZ #1240, ‘Microsoft’s Copilot evolving to ‘a Costco of AI’’ is on Microsoft’s Copilot, led by Satya Nadella, evolving to a ‘Costco of AI’ under a new Copilot leader, Charles Lamanna, a possible successor to Satya himself.
AI Ramblings Daily on AI-RTZ is here to think through AI and reset. Together.
Tomorrow, ARD 186 and AI-RTZ #1241.
Thanks for joining us, AI Curious Folk. Stay tuned.
Full Source Reading
OpenAI and Anthropic
The Information, ‘OpenAI Accelerates Ad Sprint With Meta Hires, Slashed Prices’, Oct 15, 2026
The Information, ‘Anthropic’s Rapid But Slowing Growth Could Weigh on IPO Valuation’, Oct 14, 2026
Meta and Apple
Meta, ‘Introducing Muse Spark Open, The Most Capable US Open-Weight Model to Date’, Oct 15, 2026
CNBC, ‘Apple and Meta team up to make Muse AI model run on Mac computers’, Oct 15, 2026
Axios, ‘Apple and Meta bury the hatchet, for AI’, Oct 15, 2026
TSMC
The Information, ‘TSMC to Add a New 3nm Fab as AI Orders Pile Up’, Oct 15, 2026
Bloomberg, ‘TSMC Open to Working With Terafab But Must Control Plants’, Oct 15, 2026
Gadget AI
The back catalogue
AI-RTZ #1240, ‘Microsoft’s Copilot evolving to ‘a Costco of AI’’
AI-RTZ #1238, ‘Meta’s ‘Wonder Full’ Muse Joins the AI Frontier’
AI-RTZ #1233, ‘Nvidia’s Open Source ‘AI Factory’ Push vs China’
AI-RTZ #1228, ‘Anthropic’s IPO Go Time, OpenAI’s $1.4 Trillion, Gemini 4 & More’
AI-RTZ #1196, ‘Apple’s Ternus Era Begins, with AI as Job One’
AI-RTZ #1161, ‘Nvidia Assembles AI ‘Avengers’, Open Secure AI Alliance’
AI-RTZ #1108, ‘Taiwan Semi, ‘Fed of the global chip economy’, Speaks’
AI-RTZ #1107, ‘Meta tries mixed AI Agents for consumers & businesses’
AI-RTZ #1104, ‘Nvidia & Microsoft focus on local AI computers’
RTZ #1063, ‘Long Expected Apple Cook to Ternus CEO shift Activated’
RTZ #1034, ‘Elon Musk boosts SpaceX/xAI pre-IPO story with ‘Terafab’’
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here.)