AI's Billion-Dollar Bids & Brute-Force Bets

Zuck is on a warpath

Author

Martin Musiol
July 30, 2025 • Estimated Reading Time: 9 minutes

In partnership with

[Tomorrow for Premium subs: The AI Millionaire Boom in the next 5 years.]

Dear fellow AI enthusiasts,

I was scrolling through Polymarket1 the other day—it's a prediction market where you can bet on almost anything—and a particular question caught my eye:

“Which company has the best AI model by the end of July?”

And I am like: no-brainer, xAI with Grok 4!

Then, I saw this:

Polymarket Probability Chart A dark‑on‑light Polymarket interface shows a multi‑color line chart tracking company probabilities from early June to late July. A blue line (Google) starts around 60 %, dips slightly, then climbs steeply in mid‑July, ending near 98 % at the upper‑right. A red line (OpenAI) drifts downward from about 25 % to 2 %. A purple line (xAI) hovers near 15 %, spikes briefly above 20 %, then collapses below 1 %. A thin green line (DeepSeek) hugs the zero‑percent baseline throughout. The right‑hand y‑axis marks 0 %–100 %. Below the chart, a data table lists outcomes: Google 98 % (+1 %), OpenAI 2 %, xAI <1 %, DeepSeek <1 %, Anthropic <1 %, each with trading buttons (“Buy Yes,” “Buy No”) and dollar‑volume figures. Time‑range toggles (1H, 6H, 1D, 1W, 1M, ALL) sit just above the table; the “ALL” range is active. A small Polymarket logo appears in the top‑right corner.

So, I was confused when I saw Google clearly dominating the bet. But then I looked at how they’re evaluating it: the https://lmarena.ai/leaderboard/text. And that’s where the nuance comes in.

It wasn’t about raw intelligence.

Time to change compliance forever.

We’re thrilled to announce our $32M Series A at a $300M valuation, led by Insight Partners!

Delve is shaping the future of GRC with an AI-native approach that cuts busywork and saves teams hundreds of hours. Startups like Lovable, Bland, and Browser trust our AI to get compliant—fast.

To celebrate, we’re giving back with 3 limited-time offers:

  • $15,000 referral bonus if you refer a founding engineer we hire

  • $2,000 off compliance setup for new customers – claim here

  • A custom Delve doormat for anyone who reposts + comments on our LinkedIn post (while supplies last!)

Thank you for your support—this is just the beginning.

👉️ Get started with Delve

(✨ If you don’t want ads like these, Premium is the solution. It is like you are buying me a Starbucks Iced Honey Apple Almondmilk Flat White a month.)

xAI’s Grok is arguably the model with the highest raw intelligence, especially for scientific and STEM tasks. But as I’ve mentioned before, it’s not yet packaged into a perfectly polished AI product like Google’s Gemini is, but also ChatGPT. To be honest, I also see OpenAI as much stronger than the 2% reflected in this bet.

The discrepancy between the leaderboard and perceived power is huge.

Text‑Arena Leaderboard Screenshot A clean, white webpage titled “Text Arena” displays a leaderboard that ranks large‑language models on versatility, linguistic precision, and cultural context. Under a trophy icon and a dropdown set to “🏆 Overall,” a search bar sits above a table with three columns—“Rank (UB),” “Model,” and “Score.” The rows read: 1  gemini‑2.5‑pro  1463 | 1  o3‑2025‑04‑16  1454 | 2  chatgpt‑4o‑latest‑20250326  1443 | 3  gpt‑4.5‑preview‑2025‑02‑27  1440 | 3  grok‑4‑0709  1432 | 5  claude‑opus‑4‑20250514‑thinking‑16k  1421 | 5  kimi‑k2‑0711‑preview  1420 | 5  deepseek‑r1‑0528  1418 | 6  claude‑opus‑4‑20250514  1414. Grey horizontal lines separate each row, and all text is in a modern sans‑serif font.

But xAI is committed to closing that gap fast. They are one of the youngest players in the foundation model game and are already planning infinite image and video generation with sound.

The key takeaway is this: the Polymarket score is as much about AI product engineering as it is about pure intelligence. 

Quick Hits from the Polymarket Rabbit Hole

BTW, Polymarket is a rabbithole of strong signals. Here are a few others I’m watching:

  • Will Solana hit a new ATH ($295) before 2026?

    • Yes: 27%. This tests the bull strength in the DeFi and GameFi sectors.

  • Israel-Hamas ceasefire before 2026?

    • Yes: 80%. Geopolitical stability here could have an indirect, positive impact on the tech world, especially for semiconductors.

  • Will the US enter a recession in 2025?

    • Yes: 22% (down from 66%!). The soft landing narrative seems to be winning, which is a boost for tech stocks, as recessions hit AI funding hard.

The AI Arms Race: Two Strategies

How are the big players planning to win? We're seeing two wildly different—and equally insane—strategies emerge.

1. xAI’s Brute Force Compute

xAI is scaling its work to brute force its way to raw, crazy intelligence with its Colossus data centers.

  • Colossus One: ~$230,000 GPUs.

  • Colossus Two: >550,000 GPUs, likely using NVIDIA's next-gen GB200/GB300 models.

  • Supporting Infrastructure: Over 1 exabyte of storage and Tesla Megapacks to smooth out the power grid.

This is the largest publicly known AI compute cluster, and xAI has ambitions to scale it far BEYOND a million GPUs by early 2026.

2. Meta’s Money Cannon

While xAI builds hardware, Meta is buying talent. You might have heard about the crazy amount of money the Facebook/Instagram ad machine is using to poach generational talent from places like OpenAI.  

And it doesn’t stop there. According to Wired, Zuckerberg and Meta have approached over a dozen employees at ex-OpenAI CTO Mira Murati's new venture, Thinking Machines Lab, offering massive compensation packages.  

Speaker at a Tech Talk A candid stage photo of a light‑skinned man in his 30s wearing a forest‑green polo shirt with a small navy logo. He smiles broadly while holding a black presentation clicker in his right hand and gesturing with his left. A lavalier microphone is clipped near his shirt placket. Behind him, a projection screen shows blurred black‑on‑white bullet points with phrases such as “If you clone on correct answers…” and “knowledge graph,” indicating a technical slide deck. The lighting highlights the presenter sharply, while the background text remains slightly out of focus, conveying an engaging, professional conference atmosphere.

How massive? They offered John Schulman, a co-founder of the lab, a package reportedly exceeding:

ONE F**KING BILLION DOLLARS!

He rejected it. The reason? A reported mismatch in leadership, culture, and a fundamental divergence in vision from the core ethos of Thinking Machines Lab.

I thought naming Shengjia Zhao as chief scientist might be the final bow on the Meta Superintelligence (MSL) team, but clearly, Zuck isn’t stopping his pursuit of top AI talent at any cost.

And boy, is Meta on fire. Zuck recently spoke about having superintelligence in sight, seeing glimpses of self-improving AI, and a path to automating all valuable work.  

Don't Sleep on Google

And lastly, don’t sleep on Google DeepMind. AlphaEvolve is a masterpiece. [last episode]

They just launched AlphaEarth for climate prediction, and many people completely missed AlphaChip, which has transformed computer chip design.

This is all on top of the entire Alpha series of models they continue to build.  

Exciting times ahead.

I hope you enjoyed it.

Happy weekend!
🙇

Our webpage

1  Side note: Polymarket often reflects reality with surprising accuracy, simply by letting highly informed people put their money where their mouths are.