Cross-Border Tech Innovation | Startup Investor via Taiwan CVC Fund | Automotive, Biotech & AI | US-Taiwan-Germany | Ex-Audi Innovation
Meta spent $14.3 billion on Scale AI in June. Two months later, executives are fleeing and Meta's own researchers call Scale's data "low quality." This isn't a partnership failing. It's Zuck's panic-buying his way out of an AI crisis. After Llama 4 flopped in April, Zuckerberg went on a desperate shopping spree - acquiring startups, poaching OpenAI talent, throwing billions at Scale AI. Classic move of a CEO who knows he's losing. Here's the inconvenient truth: OpenAI and Google STOPPED working with Scale AI right after Meta's investment. They saw something Meta didn't. Now Meta's AI team is "increasingly chaotic," top researchers are leaving, and they're quietly working with Scale's competitors instead of their $14 billion investment. Alexandr Wang got the deal of a lifetime. Meta got played. When you're behind in AI and throwing money at problems instead of solving them, you end up like this - $14 billion poorer with the same problems you started with. The biggest risk isn't "not taking any risk." It's taking desperate risks with shareholder money because you're too proud to admit you're behind.
Volker Heistermann The dot-com bubble had its own wave of FOMO acquisitions. Yahoo paid $3.6B for GeoCities, Broadcast dot com went for $5.7B, and Pets dot com burned through nearly $300M. The lesson wasn’t “don’t invest in the future,” it was that panic buying rarely creates durable advantage. The hard part is the tension in the middle. On one side, proper due diligence takes precious time and feels boring. On the other, the market and shareholders are screaming “Do Something!” No CEO wants to get ripped in half between those forces. AI is a far more transformative technology (and the numbers in M&A make the dot com spends look like chump change), but the dynamic is starkly similar. Board-level conviction and a healthy war chest get you a seat at the big table. Panic, on the other hand, can push a player to shove all their chips in when the smarter move would have been folding and waiting for the next round. Which is why right now it might be the best time in history to be on the sell side of AI M&A transactions.
Sam Pepin
3w
Interesting, would love to have you talk more about this and your story on our podcast: https://www.corecentrica.com/the-meaningful-podcast#Podcast-Guests. Let me know.
Diogo Santos
3w
The shaky assumption here is that $14B buys progress just because it buys volume. In AI, scale without curation just compounds noise—garbage multiplied, not solved. From first principles: data quality is multiplicative with model capacity. A trillion low-signal tokens don’t beat a billion high-signal ones. If Meta knew this, the question isn’t “why did they spend?” but “what guardrails failed before the cheque was signed?” There’s a governance lesson in plain sight: when R&D feels like an arms race, execs trade due diligence for speed. But panic capital has the same failure mode as panic coding—you accumulate technical debt, only at board level. Which raises the harder question: if the true bottleneck is trustable data, not GPU spend, what’s the smallest credible step a lagging player can take that isn’t just setting fire to shareholder cash?
Tony Bolland
3w
I agree Alexandr got a fantastic deal but there was a reason all top tier players were working with Scale?? It was predictable that OpenAI and Google would reduce or stop reliance on Scale. No major AI lab wants its core data pipeline dependent on a company with tight ties to a rival. Scale has >1000 employees and according to Time, Scale AI has around 240,000 contract workers—primarily attributed to its Remotasks subsidiary. These workers are coordinated via Scale’s proprietary platform, which adds automation and quality control. Over time, Scale has layered on: • Synthetic data generation • Model evaluation and red-teaming • AI-powered annotation tools (to reduce reliance on humans) • Enterprise data pipelines for government and big tech clients That feels like competitive advantage?
Meta just by virtue of being Meta would have had AI impact doing nothing. Imho
Hiring the best AI experts in the world for frankly insane amounts of money was never going to be the answer and was just going to cause other problems. Same as why you don’t only hire Ronaldo and Messi’s for your football team. Nice graphic by the way, it’s a slow motion train wreck indeed.
$14B for Scale AI wasn’t an “investment.” It was a bailout. Meta didn’t buy quality data. They bought time. And time is the one thing you can’t comp in AI. Here’s the ugly math: OpenAI + Google saw the rot and walked. Meta doubled down because admitting failure costs more than burning billions. Scale wins, shareholders bleed, researchers exit. This isn’t AI strategy. It’s the corporate version of YOLO trading at the top ... panic buying while the insiders quietly rotate into the next play. And the AI arms race will punish laggards. Because in this game, capital doesn’t compound — competence does. — At CryptoShorts Advisory we call this The Desperation Premium: when late players pay billions for assets early players already priced out as junk. Question: Do you think Meta’s “AI train” derails completely… or just keeps dragging shareholders along for the ride?
I spent 15 years between China and HK and became fluent in two Chinese languages so I have full authority to say: “NEVER TRUST A YOUNG WANG”
Zuckerberg’s $14B Scale AI deal is starting to look like the Theranos of training data. When your AI moonshot turns into a group project with the kid no one else wanted, it’s less “strategic bet” and more “panic purchase.”