DeepSeek 2.0? Are we racing toward margin collapse?

The Mirage of AI Abundance and the Race Toward Margin Collapse

Transcript:  

Hey, this is Robert Maciejko. Was this one of the biggest AI weeks so far?   It did feel like a DeepSeek 2.0 moment. Chinese labs keep raising the bar. This week, China's open source, Kimi K2 Thinking arrived with trillion parameter scale and strong public benchmarks. Early reads suggest it tops leading closed US models on reasoning, coding, and agentic tasks. 

Markets were already shaky. A number of the highest flying assets are 10 to 30% off, 52 week highs. Add US government shutdown noise and  tariff tensions, and you get a jittery backdrop. Jobs flashed red. US October layoffs were the highest for that month since 2003.  Cost cutting and AI were both cited. Hiring is slowing across many white collar roles, especially among recent grads.In the past weeks, CapEx news went into overdrive.

OpenAI and partners are reportedly lining up about $1.4 trillion in deals over about eight years for data centers and chips, roughly 30 gigawatts of compute. To put that in perspective, that's the consumption level of Spain, a whole country. The spend is far above today's revenues. In OpenAI's case, they said their run rate at the end of this year will be about $20 billion. So it assumes enormous growth.  

Add to that backstop controversy. The CFO of OpenAI had wording that sparked headlines about government support. Sam Altman had to walk it back posting a clarification saying they're not seeking a bailout. US   AI Czar David Sachs said there will be no federal bailout for AI. Either way, financing risk is now part of the public conversation.

For now, the mode is in infrastructure.   Microsoft, Amazon, and Google are all building NVIDIA chip replacements. The question is though, if supply overshoots demand, will chip and cloud prices also be under pressure? Alliances are shifting. Apple is reportedly set to pay about $1 billion per year to license a custom Google Gemini for Siri; interestingly not OpenAI, which aims to compete with Apple on devices. Microsoft is building more in-house, trying to reduce dependence on OpenAI.

This week geopolitics was not background noise. Jensen Huang said "China is going to win the AI race", although he later walked that back at least a bit. The message stands though, China's developer base it's energy and industrial policy make it a real contender. So Kimi K2 Thinking may end up being the DeepSeek 2.0 moment.

Meanwhile, Elon  Musk's $1 trillion compensation plan was confirmed. If all the people in this picture were millionaires, it would not add up to one Elon Musk who had achieved the trillion dollars. Whatever your view, it shows how concentrated the bets have become.

What does this all mean?

Open source pressure is up, possibly a DeepSeek 2.0 moment. CapEx is up. Prices maybe down across the stack over time unless you can ship must-have products.

That all leaves two mega questions.

If you are an AI optimist and abundance is created, who captures it and on what terms?

If abundance is not created and prices fall faster than demand rises, what's your plan? What's our plan?

Right now, it still feels like there are no adults in the room shaping the markets.

This means we all have to take responsibility for being prepared for what might be coming.    

Thanks for listening.

Hosted by Robert Maciejko