Midterms widen the AI divide: CoreWeave falls as hyperscalers gain

Microsoft and Amazon each surged more than 6% on Wednesday while AI infrastructure provider CoreWeave plunged around 10% — a same-day split that shows how U.S. midterm results are widening the gap between cash-rich tech giants and debt-funded AI startups.

The trigger sits in state-level politics. Democratic victories in key gubernatorial races raised the prospect of tougher scrutiny for data center projects, adding to investor concerns about how quickly new capacity can be built and financed. In Texas, Democratic governor-elect Gina Hinojosa has proposed a statewide moratorium on new data center permits, although projects already under construction would reportedly be exempt.

Chipmakers did not escape the selloff either. Micron fell 7% and Nvidia dropped 3.17% as markets digested the results.

Why does this matter beyond a single day of trading? Because the midterm outcome could harden an already-forming split in who can afford to build AI's physical backbone — a divide analysts increasingly describe as K-shaped, where the strong get stronger and the weak get squeezed.

What's driving the split in AI stocks?

Borrowing costs are the mechanism, and it's fairly simple. James Sullivan, managing director and co-head of global fundamental research at JPMorgan, told CNBC's "Squawk Box Asia" on Thursday that borrowing spreads have widened significantly for neocloud companies — the smaller, specialized providers of GPU-heavy cloud capacity. That leaves their cost of additional capital substantially higher than that of hyperscalers, which benefit from more diversified customers and revenue streams.

Sullivan connected the financing gap directly to the equity moves. "The fact that you're seeing this bifurcation in equity performance is largely driven by the view that existing capacity is significantly more valuable now because of incremental supply constraints," he said.

In plain terms: if it gets harder and more expensive to build new data centers, the ones already running become scarcer assets. Companies that own capacity today win. Companies that planned to borrow their way into the market lose.

How serious are the data center risks?

The midterm results could reinforce that view if they make new projects more vulnerable to permitting delays. Research firm BMI, in a post-midterm analysis cited by The Business Times, warned that a sharp slowdown in data center construction and AI investment could produce a greater-than-expected decline in overall investment.

But the timing matters, and the exposure is uneven across the construction pipeline.

• Much of the data center capacity expected to come online in 2027 is already under construction, which limits the near-term impact of new restrictions.
• Projects planned for 2028 face greater exposure to permitting delays and regulatory uncertainty.
• Tighter regulations on data center construction could become a more significant headwind to AI investment from 2028, according to Nomura.

Investment bank Nomura laid out that timeline in a note Wednesday about the midterm results. The takeaway is that a permitting crackdown wouldn't freeze AI capacity overnight — it would bite two to three years out, when today's shovel-ready projects run out and the next wave needs fresh approvals.

How bad is the AI financing gap?

The credit markets started pricing this stress well before the midterms. Shares of Nvidia-backed neoclouds CoreWeave, Nebius and Iren fell at least 30% in October, as tightening credit markets raised questions about how AI infrastructure projects would be financed.

CoreWeave's numbers are the starkest. Its five-year credit default swap spread has surged past 1,100 basis points, reflecting rising costs for investors seeking protection against a potential default. That's a market pricing real distress risk, not just sentiment.

BMI, in its Wednesday note, took a measured view of the downside scenario. The firm said a slowdown in AI infrastructure development was more likely than a nationwide halt, while tighter regulations could raise construction and power procurement costs. So the base case is friction and higher costs, not a stop.

The distinction, then, may come down to how well companies can absorb delays and fund the capacity they do build. JPMorgan Asset Management made exactly that argument in an Oct. 20 research note:

• Hyperscalers have broader cloud businesses and more internally generated cash flow.
• Lower-rated neoclouds face greater exposure to borrowing costs, energy prices and refinancing conditions.
• Financially stronger hyperscalers could also secure cheaper capacity or acquire assets if weaker operators scale back.

That last point is the quiet headline. If struggling neoclouds pull back, the giants don't just survive — they get to buy capacity on better terms. The midterms may have accelerated a consolidation path that the credit markets were already pointing toward.

The stakes here run past the stock tickers. Data center construction is one of the largest capital expenditure stories in the global economy right now, and how it gets financed shapes who controls AI compute — the hyperscalers, or the specialized challengers trying to break in. Politically constrained permitting plus tight credit narrows that field from both sides.

What happens next?

Watch the 2028 pipeline. Nomura's framing puts the real test two years out, when projects that haven't broken ground yet need permits in a potentially stricter regulatory environment — and when weaker operators face refinancing at spreads their business models may not support.