- OpenTools' Newsletter
- Posts
- 😲Meta's Paid Model Shift
😲Meta's Paid Model Shift
PLUS: Gates Predicts AI Valuation Corrections | China's Data Center Edge
Reading time: 5 minutes
🗞️In this edition
Zuckerberg's avocado model could end Meta's free AI approach
Bill Gates predicts winners and losers in "hyper competitive" AI
Veteran investor predicts China AI dominance by 2035
In other AI news –
Google expands AI mode with better source links
Spotify pilots prompted playlists for deeper personalization
Figma rolls out AI object removal and image extension
4 must-try AI tools
Hey there,
Meta is rethinking its open-source promises, Bill Gates says some AI valuations are headed for a cooldown, and a leading Chinese investor claims China’s infrastructure edge could put it ahead of the US.
We're committed to keeping this the sharpest AI newsletter in your inbox. No fluff, no hype. Just the moves that'll matter when you look back six months from now.
Let's get into it.
What's happening:
Meta's next AI model could mark a change in its "open source" AI strategy that Mark Zuckerberg had called "the path forward." Currently, the company is working on a new AI model, code-named Avocado, that it might charge for access to, Bloomberg reports.
Last year, Meta launched its open source Llama 4 AI model, but it had a disappointing release, with Meta caught gaming AI benchmarks and being forced to delay the planned "Behemoth" version. But Zuckerberg scrapped that "in pursuit of something new," Bloomberg says.
Following Llama 4 launch, Zuckerberg has made sweeping changes to Meta's AI team, including hiring former Scale AI CEO Alexandr Wang after investing $14.3B in the company and spending truckloads of money hiring other top AI talent for newly-named Meta Superintelligence Labs group.
In July 30th memo about "personal superintelligence," Zuckerberg indicated Meta may need to shift approach on open source: to mitigate potential safety risks, companies will have to be "careful about what we choose to open source."
As part of changes, Zuckerberg now "spends much of his time and energy working closely with those new hires, in a group called TBD Lab," Bloomberg says. That team has "siloed space" near Zuckerberg's office at Meta's headquarters.
Why this is important:
Meta abandoning open source for paid model reverses strategy Zuckerberg championed for years as competitive advantage against OpenAI and Google.
Llama 4 caught gaming benchmarks and the Behemoth version scrapped shows Meta's open approach failed to deliver a competitive frontier model.
Hiring Scale AI CEO after $14.3B investment and creating TBD Lab with "siloed space" near Zuckerberg's office signals urgency and strategic importance.
"Careful about what we choose to open source" citing safety risks is convenient justification for monetization strategy after open source didn't produce leads.
Our personal take on it at OpenTools:
This is Meta admitting open source didn't work and pivoting to paid models.
Zuckerberg championed open source as "path forward" positioning Meta as good guys versus closed OpenAI. Now they're considering a paid model because Llama 4 disappointed and they're losing to competitors.
Gaming benchmarks were embarrassing. Getting caught inflating Llama 4 performance metrics then scrapping Behemoth undermined the credibility of entire open strategy.
$14.3B investment in Scale AI plus hiring Wang shows Meta realized open sourcing models without best training data and infrastructure doesn't produce frontier capabilities. They're buying data labeling capacity and expertise they lacked.
The real reason: Meta needs revenue from AI. They've spent billions, Llama hasn't generated meaningful business value, and investors want ROI. Paid model creates revenue stream open source doesn't.
Avocado being "might charge for" not definite suggests internal debate. Some at Meta likely still support an open approach. But Zuckerberg spending time in TBD Lab indicates he's driving a paid strategy.
From Our Partner:
Bring your ideas to life - without tabs, folders, or file names. Kosmik’s AI pulls the right images, videos, and notes into one clean workspace with just a few words.
✨ Work faster. - Find the perfect assets in seconds.
🧠 Stay organized. - Everything you drop in is auto-tagged and categorized.
🤝 Collaborate easily. - Share boards, gather feedback, and move projects forward together.
For Creatives, Designers, and Visual Thinkers
Subscribe to Kosmik Pro today with the code “Opentools40” and get a 40% discount on your first year of subscription.
What's happening:
Bill Gates warned some AI companies with high valuations will lose out in the "hyper competitive" AI industry. "Does it mean all of these companies with high valuations will be winners? No," Gates told CNBC at Abu Dhabi Finance Week Monday.
"AI is only a bubble in the sense that not all of these valuations will end up going up. Some of them will go down," he said, adding AI is "a deeply profound technology that will reshape the world. There's not the slightest doubt about that."
Many AI companies have valuations far beyond averages. Palantir and Tesla have P/E ratios well over 200, compared with the S & P 500 average of about 25. Global markets fell in November as fears of bubble bursting gained traction.
Gates said "a reasonable percentage of those companies won't be worth that much." Despite AI frothiness, Gates said AI would fundamentally change lives for the better in health, education, and agriculture.
Earlier this week, Gates Foundation pledged $1.9B to fight polio. Gates predicted next year would be big for global health, piloting AI tools including "virtual doctor supporting all African dialects" and "farm advisor" to "dramatically raise productivity" for African farmers with small plots.
Why this is important:
Gates calling out specific valuations as unsustainable while maintaining AI is "deeply profound" separates technology merit from financial speculation. That's rare candor from tech industry insiders.
Palantir and Tesla having P/E ratios over 200 versus S&P 500 average of 25 shows extreme premium markets are paying for AI exposure. That's 8x normal valuation multiples.
"Reasonable percentage of companies won't be worth that much" from Microsoft founder carries weight. Gates has seen tech bubbles (dot-com, mobile) and knows the difference between transformative technology and overvalued companies.
November market drop cited as evidence bubble fears are affecting actual trading, not just analyst commentary. When markets correct bubble concerns, that's a leading indicator.
Our personal take on it at OpenTools:
Gates is stating the obvious while hedging bets.
Saying "not all high valuations will be winners" in the "hyper competitive" market is true. Of course some companies fail. That's true in any industry, bubble or not.
But Gates pairing valuation warning with "deeply profound technology" qualifier is having it both ways. He's saying AI matters (protecting Microsoft's massive investments) while warning some companies are overvalued (covering himself if market corrects).
Palantir and Tesla P/E ratios over 200 are exhibit A for bubble pricing. Those multiples assume sustained exponential growth for decades. Small deviations from growth expectations trigger massive valuation corrections.
"Next year will be big for global health" prediction about piloting AI tools reveals Gates sees 2026 as an inflection point for practical applications. That's more concrete than valuation warnings.
This is a veteran tech investor saying bubble characteristics exist but technology still matters. That's correct but doesn't help anyone decide which companies are overvalued versus fairly priced.
What's happening:
China will likely overtake the US in AI within a decade because of faster buildout of power and data center infrastructure, according to Allen Zhu Xiaohu, a well-known “unicorn hunter” & managing director at GSR Ventures.
"It's much easier [for China] to catch up on algorithms and AI models than [for the US] to build up data centers and power plants," Zhu said on a recent podcast. "AI competition is really a competition in data centers and electricity supply and China has a significant edge there."
His view echoes growing US concern over approaching power shortages for AI data centers. A recent report said computing demand surge was creating "exceptional" pressure on America's power grid, making it hard to secure reliable electricity for new AI facilities.
Zhu acknowledged Chinese AI startups face an uphill battle against Alibaba and ByteDance, which can offer AI models at very low prices or free, making it harder for startups to earn money.
Zhu's venture fund invested in about a dozen AI application developers this year rather than foundational model developers. He said ByteDance or Alibaba were more likely to become "China's OpenAI" than today's AI startups.
Why this is important:
Recent report confirming the US power grid facing "exceptional" pressure validates Zhu's thesis. The US is struggling to secure reliable electricity for new AI facilities while China accelerates nuclear and solar construction.
ByteDance and Alibaba likely to become "China's OpenAI" rather than startups shows China's AI winner will be established tech giant, not startup. That's different from the US where OpenAI, Anthropic are newer entrants.
Our personal take on it at OpenTools:
Infrastructure advantage is real but algorithmic lead matters more than Zhu acknowledges.
China building data centers and power plants faster than the US is true. Centralized planning enables rapid infrastructure deployment. But "much easier to catch up on algorithms" undersells the difficulty of frontier AI research.
Zhu's "10 years" prediction assumes linear infrastructure advantage translates to AI leadership. But AI progress isn't just about computer availability. It's about research breakthroughs, product-market fit, and ecosystem development. China leads on first, US leads on others.
The US power grid concern is legitimate. Data centers consuming extraordinary amounts of electricity while the US struggles to build new generation capacity creates bottlenecks. But that's solvable with investment and policy changes. It's not a permanent structural disadvantage.
The "China will overtake the US in decade" prediction depends on multiple assumptions: infrastructure advantage persists, algorithmic gap closes, Chinese companies execute better than US competitors, and the US doesn't solve power supply bottlenecks. Any of those failing invalidates prediction.
This is a Chinese investor with vested interest in the Chinese AI ecosystem predicting Chinese victory. Take with appropriate skepticism while acknowledging infrastructure concerns are real.
Google says it will link to more sources in AI Mode – AI Mode will preface its list of sources with an AI-generated snippet about why they’re relevant.
Spotify tests more personalized, AI-powered ‘Prompted Playlists’ – The new tool allows users to describe what they want to hear in a personalized playlist that reflects the “full arc” of their tastes, according to the company.
Figma launches new AI-powered object removal and image extension – The company said that these features will save the hassle of exporting images to other tools for editing and importing them back.
FindWise - An AI-powered search assistant that allows users to ask questions and get answers based on the content of a website
Krizmi - An interactive learning platform that offers auto-generated flashcards and quizzes to help students retain and test their knowledge
Zeliq - An all-in-one sales solution that helps businesses increase their sales and streamline their outreach efforts
Ask Jules - A book discovery companion that helps users find their next book and answers book-related questions
We're here to help you navigate AI without the hype.
What are we missing? What do you want to see more (or less) of? Hit reply and let us know. We read every message and respond to all of them.
– The OpenTools Team
How did we like this version? |
Interested in featuring your services with us? Email us at [email protected] |