The OpenAI mess. AKA: chip fabs are expensive
Superintelligence, so hot right now
Welcome to Cautious Optimism, a newsletter on tech, business and power.
Friday. We made it! As did Polymarket and Kalshi, which finally earned their place in Google Finance. A limited number of users will get access to the prediction markets on Google Finance in the next few weeks. Polymarket and Kalshi do feature interesting betting on economic events, but the services are heavily weighted towards sports betting today. Perhaps being included in a mainstream product like Google Finance will shift their mix? — Alex
📈 Trending Up: Musk’s Tesla pay package … the self-driving wars … paywalls … AI wrongful death suits … flight cuts … the slop trough … betting on fusion
📉 Trending Down: Glaciation … consulting … human mathematics … Roblox’s public image … gay marriage? … Chinese exports … Russian GDP …
Things That Matter
How the turn tables: Investors were pessimistic about tech broadly yesterday, and market commentary blamed the slump on “the artificial intelligence trade [coming] under pressure yet again amid concern about their eye-watering valuations.” In early morning trading today, the Nasdaq was down another 0.8%.
The Nasdaq 100 reached a record high of 26,182.10 recently, so it was just 4.8% lower than all-time highs at 24,918 as of this writing. For months it seemed that the default market movement was higher. Perhaps today marks the early days of the trend changing tack.
On the earnings front: Airbnb’s shares rose after the company reported revenue rose 10% to $4.1 billion in the third quarter, and forecast a further 7%-10% growth for the current quarter. That was better than expected. Block fell 15% after its third-quarter results triggered profitability worries. Doximity crushed expectations and got its knees swept for the effort. D-Wave (quantum) and Nuscale (fission) took beatings after reporting. Dropbox was largely flat after reporting a revenue decline that beat expectations. U.S. affiliate MP Materials got crushed, but JFrog soared after beating expectations and raising guidance.
And then there was Opendoor: Shares of the real estate company fell around 24% after reporting third-quarter results. Here’s the Yahoo Finance summary:
Opendoor reported revenue of $915 million, its lowest quarterly revenue since Q4 2023 but above Wall Street’s expectations for $882 million. A loss per share of $0.12 during the quarter was wider than the expected $0.07 loss, according to S&P Global Market Intelligence consensus estimates.
The same time last year, Opendoor had revenue of $1.38 billion. The company is pitching itself as a reborn entity, re-founded “as a software and AI company.” It’s laid out three goals: buy more houses, accelerate buying with improved margins, and improve operating leverage.
Still, Opendoor was candid that its next quarter will be past-coded:
Our results in the upcoming quarter are largely the outcome of us managing decisions that were made several months ago. We’re focused on making the right long-term decisions for the business, not managing to short-term guidance. What matters, and what we want to be held accountable for, are the actions we take from here and the results they drive over time.
Fair enough. That’s roughly what I expected: Difficult near-term financials, and plans for a brighter future. Honestly given how meme stocks like Opendoor tend to trade, I almost expected it to soar. The old adage, “buy the rumor, sell the news,” may hold even for stocks that appear to be driven by retail interest.
Sweetgreen’s results added another data point to the chart called ‘Just how stretched is the American consumer?’
Superintelligence, so hot right now: Do you want some AGI? How about ASI? No, not your jam? How about HSI? What’s that, you ask? It’s not Artificial General Intelligence (OpenAI), or Artificial Superintelligence (Meta), but Humanist Superintelligence, per Microsoft.
Yes, the long-time OpenAI collaborator and shareholder is spinning up its own superintelligence team. Meanwhile, xAI is expecting the next model in its Grok family, called Grok 5, to be “AGI or something indistinguishable from” it.
Why is everyone piling into the superintelligence game? Because it may hold the true AI prize: the ability to replace lots of useful human work, make new discoveries, and eventually become a copilot for all of humanity — or at least those who can afford the subscription. There’s a feeling in AI-land that we’re getting closer to building artificial intelligence systems that can learn and improve themselves. Once we have that, we can just invest more resources and learn more, right?
You don’t want to be sitting on a powerful LLM when your competitors are shipping self-improving AI!
Speaking of powerful LLMs, Chinese lab Moonshot AI’s latest model in the Kimi family, Kimi K2 Thinking, scored eye-watering benchmark scores. This is great news if you are an open-source AI bull; less so if you are worried about the China-U.S. AI rivalry.