Morning Minute: Citadel Cautions Against the AI Trade Ahead of SpaceX IPO

One of Wall Street's most powerful hedge funds is sounding the alarm on the AI trade, and the timing couldn't be more pointed. Citadel Securities — the market-making arm of Ken Griffin's financial empire — released a macro note this month arguing that AI adoption is slowing down due to cost concerns, and the evidence is mounting from some of the biggest names in tech.

In the past month alone, Amazon killed its internal KiroRank token leaderboard after engineers started gaming it. Microsoft cancelled Claude Code for roughly 5,000 employees after token bills blew past budget. Meta scrapped a similar internal AI leaderboard in April. And Uber's CTO admitted the company burned through its entire 2026 AI coding budget in just four months.

The pattern is clear: enterprises have stopped treating AI token consumption as a trophy and started asking what all that spending actually ships. The question Citadel is raising is whether the IPO valuations of the biggest AI companies depend on that spending continuing to grow forever — and what happens if it doesn't.

The IPO Floodgates Are Opening

This debate lands at a pivotal moment. SpaceX is going public today after market close, expected to trade Friday at roughly $1.77 trillion. That massive valuation leans heavily on the xAI compute business SpaceX absorbed in February. A big chunk of the bull case for SpaceX — and for the broader AI infrastructure market — is that demand for AI compute will compound indefinitely.

Anthropic filed a confidential S-1 on June 1, just weeks after a funding round that valued it near $965 billion on the promise of enterprise adoption — during the same stretch that one of its marquee internal customers (Microsoft) walked away over cost. OpenAI is preparing its own filing. The richest IPO slate in history is lining up, and a big part of every pitch is the assumption that enterprise AI spending will only go up.

Citadel's call is that frontier AI gets concentrated among a handful of firms with the balance sheets to absorb the compute cost. Everyone else cuts back. That could create a shaky transition period, especially as these mega-IPOs suck liquidity from the rest of the market.

What the Enterprise Revolt Looks Like in Practice

The Microsoft-Claude Code situation is the most revealing example. Microsoft didn't just scale back its AI coding assistant program — it cancelled it entirely for 5,000 staff because the token costs exceeded the projected budget. This isn't a company struggling to afford AI; this is one of the richest companies on earth deciding that the ROI doesn't justify the spend at current pricing.

Uber's situation is even starker. The company allocated a full-year budget for AI coding tools and spent it all in a third of the time. That's not adoption — that's a cost structure that doesn't work at scale. And if Uber, with its engineering culture and willingness to experiment, can't make the math work, it raises serious questions about how the economics play out for companies with less margin for error.

Amazon killing KiroRank adds another dimension: when your own employees are gaming your internal AI systems, it suggests the tools are creating more complexity than they're resolving. Meta's earlier scrapping of a similar leaderboard points in the same direction.

The Tether-NEURA Signal: AI Gets Physical

Meanwhile, a very different kind of AI investment is accelerating. Stablecoin giant Tether led a $1.4 billion Series C for NEURA Robotics, a German humanoid robotics maker, with Nvidia and Amazon also participating. NEURA will embed crypto payment tools and edge AI directly into its machines.

This is a significant signal. The most profitable company in crypto is pointing its balance sheet at physical AI rather than software-only models. Tether has spent the past year converting USDT profits into stakes across compute, energy, and media. Humanoid robots are the next target.

The investment thesis is straightforward: if enterprise spending on AI software is hitting a cost ceiling, the next growth vector is AI that moves physical goods and performs physical tasks. A humanoid robot that carries its own wallet, processes transactions via edge computing, and settles payments in stablecoins doesn't need enterprise IT budgets to approve its token spend. It operates as its own economic unit.

Mastercard's new Agent Pay system reinforces this trajectory. Launched with Coinbase, Ripple, and Solana, Agent Pay lets AI agents buy services and settle payments using cards, bank accounts, or stablecoins. A card network this size putting stablecoins on equal footing with traditional payment rails for machine-to-machine commerce is meaningful validation. Autonomous agents need money that moves as fast as they do, and stablecoins settle in seconds at any hour.

Friday's SpaceX Open: The Sentiment Test

All of this converges on Friday's SpaceX market debut. If SpaceX opens strong — if a $1.77 trillion valuation holds or expands — it signals that risk appetite for AI infrastructure is still alive and well. The market would be saying that compute demand will find a way, that the companies at the frontier are worth their massive capital requirements, and that the enterprise cost revolt is a temporary speed bump.

If SpaceX flops or trades flat, it's a different signal. It would suggest that Citadel's cost-ceiling thesis has more weight than the bulls want to admit, that liquidity is being pulled toward the biggest names and away from everything else, and that the AI trade may have already priced in growth that can't materialize at current cost structures.

Either way, Friday will be a sentiment read on the entire AI investment landscape — not just SpaceX.

What This Means For You

Whether you're an investor, a tech worker, or just someone trying to understand why everything seems expensive right now, the AI cost ceiling matters for your wallet. Here's why:

If enterprise AI spending really is hitting a wall, it affects more than just tech stocks. The companies building AI infrastructure — chip makers, data center operators, energy companies — have been the primary drivers of market gains for two years. A pullback in AI spending means a pullback in the sectors that have been propping up your 401(k).

For workers, the enterprise AI adoption curve has a direct impact on job security. Companies that are spending recklessly on AI tools today may be the same ones doing layoffs tomorrow when the ROI doesn't materialize. And companies that are being disciplined about costs — like Microsoft cancelling Claude Code — may actually be better positioned long-term.

For consumers, the AI cost question affects pricing. If AI companies need to maintain revenue growth while enterprise customers cut back, they'll need to find new revenue streams. That means more ads, more data harvesting, and more premium tiers in products you already use. The cost of AI gets passed down either way — the only question is whether you pay for it as a taxpayer subsidizing data centers, an investor holding underwater positions, or a consumer paying more for AI-enhanced everything.

Watch SpaceX on Friday. It's not just a stock debut — it's a referendum on whether the AI economy's math actually works.