AI Infographics 2026: Trillion-Dollar Capex, the Mythos Block, and Whether AI Is Actually Creating Jobs

TLDR: BigTech AI capital expenditure is on a parabolic trajectory toward $1 trillion by 2027, the labor market effects of AI remain genuinely ambiguous, and Anthropic's Mythos model is at the center of the first-ever informal executive branch restriction on a software product's distribution in US history.

Summary:

Let's start with the money, because that's where the story really begins. Wall Street analysts are now projecting $1 trillion in total AI-related capital expenditure by 2027, and that figure doesn't even capture the full picture. Oracle, CoreWeave, xAI, Nebius, Crusoe — none of these are fully baked into those BigTech numbers. The acceleration isn't theoretical. It's driven by the exploding demand for inference compute, which in turn is fueled by the sheer volume of tokens being generated across the ecosystem. Corporate bonds and venture capital are essentially subsidizing all of it. I find the framing of "burning tokens and heating GPUs" apt — we don't have the real numbers, and I'm not sure anyone does. The ROI for society? Still speculative.

This brings up a question I keep circling back to: what does all of this spending actually produce in terms of productivity? The labor economics here are genuinely messy. There's a fascinating case study in the customer service industry. Call center employment in the Philippines has been rising every year since 2016, through the entire AI boom. That's Jevons Paradox in action — when AI makes an activity cheaper, you don't necessarily do less of it. Companies serve more customers, open more channels, reach more markets. The technology that was supposed to shrink an industry ends up expanding it. I find this more interesting than most of the hot takes about mass displacement.

What the data does suggest is that demographics and immigration policy may matter more than AI for the near-term US labor market. An aging population, reduced immigration, and a low-hire environment together paint a more complicated picture than the simple "AI takes jobs" narrative. Sectors with high AI adoption are showing strong growth in new business applications since 2022, which points to AI lowering the barrier to start a company rather than simply eliminating roles. GenZ college graduates are entering a tighter job market, and more of them appear to be turning toward entrepreneurship. Whether AI-enabled solopreneurs constitute a real structural trend or a social media narrative is genuinely unclear.

Now for the story that cuts through everything else right now. Anthropic is reportedly seeking to raise a new funding round at a valuation approaching $900 billion, possibly exceeding $1 trillion given investor demand. SemiAnalysis put Anthropic's ARR as high as $44 billion. That's a staggering number for a company that didn't exist a few years ago. But the bigger story is Anthropic's Mythos model. The Trump administration has informally opposed Anthropic's plan to expand access to Mythos to roughly 70 additional companies, citing cybersecurity concerns — Anthropic's own researchers apparently acknowledged that Mythos can identify and exploit sweeping security vulnerabilities. The Pentagon designated Anthropic as a national security supply chain risk, which creates the bizarre situation where it's unclear how the DOD could actually use Mythos without violating its own designation.

What makes this genuinely unprecedented is the mechanism. This is, as policy analysts are calling it, the "Mythos Block" — the first time the US executive branch has restricted the distribution of a software product without a specific law or court order. Meanwhile, the Pentagon has inked AI agreements with SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, and Amazon Web Services for classified network integration at Impact Levels 6 and 7. The political dimensions here aren't subtle — Trump's known ties to Peter Thiel and Sam Altman sit awkwardly alongside the selective treatment of Anthropic. I'm not going to pretend the competitive dynamics aren't visible.

Key takeaways:

AI capex is on track for $1 trillion by 2027 across BigTech and emerging cloud providers, with ROI for society still highly uncertain
Jevons Paradox may be cushioning some labor disruption — cheaper AI interactions seem to drive more interactions, not fewer
Demographics and immigration shifts likely outweigh AI's near-term labor market impact in the US
Anthropic is approaching a $1 trillion pre-IPO valuation with ARR estimated at $44 billion
The "Mythos Block" is the first informal executive branch restriction on a software product's distribution in US history, with no law or court order behind it
The Pentagon has separately signed AI deals with 7 companies for classified network integration, noticeably excluding Anthropic

Why do I care:

The Mythos Block is the one to watch closely. We've had plenty of debates about AI regulation, safety frameworks, and voluntary commitments. But an informal executive action that restricts distribution of a specific model, driven by cybersecurity and competitive concerns in an environment with obvious political entanglements — that's a genuinely new pattern. It sets a precedent that has nothing to do with any law or established regulatory structure, and that ambiguity is exactly what makes it so consequential. The trillion-dollar capex story is important, but it's the governance story that will shape whether any of this infrastructure actually delivers value.

What AI Infographics say about the future of AI?