Here's a tale of two numbers you should hold in your head at the same time.
Number one: $2 trillion. That's how much has been wiped off AI software market caps in the past two weeks, according to J.P. Morgan. Software stocks cratered as investors confronted the possibility that large language models might replace the very products these companies sell.
Number two: Tens of billions. That's the estimated value of the multiyear deal Nvidia and Meta announced on Tuesday — millions of Blackwell and next-generation Rubin GPUs, plus Nvidia's first standalone CPU deployment, all destined for Meta's AI data centers.
The market is panicking about AI's future. The companies building AI's infrastructure are spending like it's already won.
What the Deal Actually Is
Nvidia will supply Meta with millions of its current Blackwell GPUs and upcoming Rubin GPUs, along with Grace CPUs and networking equipment, over a multiyear period. The chips will power both the training and inference of Meta's AI models across its own data centers and through Nvidia Cloud Partners.
"The deal is certainly in the tens of billions of dollars," chip analyst Ben Bajarin of Creative Strategies told CNBC. "We do expect a good portion of Meta's capex to go toward this Nvidia build-out."
The Grace CPU piece is notable. This marks the first time Meta will deploy Nvidia's standalone CPUs — not just its GPUs — putting Nvidia in direct competition with Intel and AMD in the data center CPU market. It's a quiet land grab happening inside a flashy GPU deal.
Meta's stated goal: build what it calls "personal superintelligence" for its 3+ billion users, with enhanced AI capabilities in WhatsApp, Instagram, and Facebook. The company is building AI infrastructure at a scale that only a handful of entities on Earth can match.
The $2 Trillion Context
The timing makes this deal read differently than it would have a month ago.
Deutsche Bank's Jim Reid, in a note to clients this week, put it plainly: "For months, my published view has been that nobody truly knows who the long-term winners and losers of this extraordinary technology will be. Yet as recently as October, markets were implicitly pricing in a world where almost every tech company would come out a winner."
The $2 trillion wipeout happened because investors are finally confronting a reality that was obvious to anyone paying attention: AI creates winners and losers. The LLMs that power ChatGPT, Claude, and Gemini don't just augment existing software — in many cases, they replace it. Legal tech, IT services, consulting, logistics — companies in all these sectors saw their stocks hammered as the market priced in disruption rather than universal benefit.
Jeremy Siegel of Wharton, writing for WisdomTree, argued this is actually healthy: "When companies talk about $200 billion in capital expenditures, markets should scrutinize payback periods, competitive dynamics, and whether durable moats can be built in an environment where technology is evolving at breakneck speed."
And yet, in the middle of that scrutiny, Meta signs the biggest chip deal in AI history. The message: we are not the companies being disrupted. We are the ones doing the disrupting.
The Split Economy of AI
What's emerging is a two-tier AI economy that will define the next decade:
- Tier 1: The infrastructure layer. Nvidia, the hyperscalers (Meta, Google, Microsoft, Amazon), and the chip supply chain. These companies are spending because they believe AI compute demand will only grow. The Nvidia-Meta deal is a bet that the infrastructure floor hasn't been found yet.
- Tier 2: Everyone built on top. Software companies, SaaS providers, service firms — anyone whose value proposition can be replicated or replaced by a well-prompted model. This is where $2 trillion just evaporated.
The uncomfortable truth: the same technology that's making Nvidia and Meta richer is destroying value for thousands of other companies. This isn't a rising tide. It's a flood with clear high ground, and only a few companies are standing on it.
JPMorgan CEO Jamie Dimon captured the nuance at last year's Fortune Most Powerful Women Summit: "You should be using it," he told businesses. But he added that "some asset prices are high, in some form of bubble territory." The distinction between AI itself (real, transformative) and AI stocks (potentially overvalued) is the gap the market is now trying to price correctly.
What to Watch
- Meta's capex trajectory. This deal likely represents a significant portion of Meta's capital spending for years. If AI product revenue doesn't grow proportionally, investor patience will thin — Mark Zuckerberg's metaverse pivot taught us that much.
- The CPU war. Nvidia's Grace CPUs going into Meta's data centers is a shot across Intel's and AMD's bow. Watch whether Google and Amazon follow with similar Nvidia CPU deployments.
- The software carnage. The $2 trillion wipeout may not be over. As AI capabilities improve, more categories of software become vulnerable. The question for any SaaS company: is your product a feature an LLM can replicate for free?
Why This Matters
You're watching an economy split in real time. One side is spending tens of billions to build AI infrastructure because it believes the revolution is just starting. The other side just lost $2 trillion because investors realized the revolution might destroy their businesses. Both of these things are true simultaneously, and that's the most important thing to understand about AI's economic impact right now: it's not a bubble or a boom. It's a redistribution — of value, of power, and eventually of jobs — on a scale we haven't seen since the internet itself.