Here’s a pretty remarkable chart from last week’s BlackRock Investment Institute report.
Like many great charts, there are several interpretations depending on your perspective.
You might look at this and marvel at how little research and development the US government does. Yes, it has obviously gone up a lot, but it hasn’t really kept pace with the ballooning size of the economy. Over the past three decades, federal R&D has grown approximately threefold, while GDP has roughly quadrupled in size.
You could also conclude that the Mag7 have gone a bit mental lately. It’s important to note that even BlackRock’s chart undersells the AI investment boom. The “Magnificent Seven” have now sketched out plans to spend an estimated $325bn in 2025 on mostly AI infrastructure — just as DeepSeek showed that you might not need to build the computing equivalent of the pyramids to create perfectly fine chatbots.
To think, it wasn’t that long ago that we were all laughing at Facebook/Meta for losing an estimated $21bn $47bn $60bn $70bn on building the “metaverse”.
Despite marvelling at their own chart, BlackRock’s investment strategists remain wedded to their “AI mega force” thesis. Here’s their rationale:
The release of a seemingly more efficient AI model by Chinese startup DeepSeek has renewed questions about AI capex. While these questions are valid, more spending is likely needed to unlock AI innovation — recent developments don’t change our view. Broad AI adoption is still to come, and we have barely scratched the surface of all the potential AI use cases. Yet AI advances mean these models could be evolving faster than expected. That could push AI into the adoption phase sooner and is why the AI narrative and the market’s reaction could change quickly.
We are still in AI’s buildout phase — and even with potential model efficiency gains, big capital spending might still be needed to unlock further innovation, like artificial general intelligence. Strong results and guidance from the magnificent seven show they can support heavy AI capex. Q4 management commentary reveals these companies are comfortable with their AI spend and have long-term conviction in the theme and expect ongoing demand. As the buildout progresses, it opens the door for the set of AI winners to broaden further beyond the magnificent seven, expanding the total AI opportunity set, in our view.
Sure sure, all good . . . Except as BlackRock itself notes: “Beyond the buildout phase, we have yet to see the adoption phase begin in earnest, even with more players in the mix.” Oh.