How the midterms reshaped the AI narrative and hit tech markets

07 Nov 2026 04:07 8,771 views
A sharp post-midterm market selloff and new data from CoreWeave and Anthropic are forcing investors to rethink the AI boom. Here’s what changed for AI infrastructure, hyperscalers, neoclouds, and the broader economy—and why slower AI capex could be both a risk and an opportunity.

The AI boom has been one of the strongest forces driving markets, tech strategy, and even GDP growth. But in the span of a few weeks, that story has gotten a lot more complicated. A surprise political outcome, stress in AI infrastructure financing, and new numbers from Anthropic are all pushing investors to rethink who actually benefits from AI—and at what cost.

The midterms shock that rattled the AI trade

The midterm elections delivered a result that markets clearly weren’t positioned for: Democrats sweeping both houses of Congress. At first, stocks drifted higher as investors tried to make sense of it. Then on Thursday, the mood flipped.

The market saw a brutal one-day correction, with the S&P 500 down 4.4% and the Nasdaq down 5.5%. The panic was concentrated in the companies most exposed to AI infrastructure and data center buildout.

The core concern: if Democrats follow through on plans to slow AI data center construction—especially over power usage and environmental concerns—the knock-on effects could be huge. AI has been the single biggest capex story in the economy, and a slowdown would ripple across multiple sectors.

Who gets hit if AI data center construction slows?

AI doesn’t exist in the cloud as an abstract idea. It lives in physical data centers packed with GPUs, power equipment, networking gear, and cooling systems. Slowing that buildout affects an entire ecosystem of companies. Here are some of the main groups exposed:

Power and electrification companies

AI data centers consume enormous amounts of electricity. Companies that build and support power infrastructure have been big winners from the AI boom, including:

• Power suppliers and grid players like GE Vernova and Quanta
• Electrification specialists such as Eaton, Vertiv, and nVent

These names were hit hard in the Thursday selloff as investors quickly repriced their AI growth expectations.

Chipmakers and data center hardware

The AI narrative has been a windfall for semiconductor and data center hardware companies. A slowdown in construction would directly pressure:

• GPU and memory makers like Nvidia, AMD, and Micron
• Server and data center OEMs such as Dell and Super Micro
• Networking and switching providers like Arista and Cisco

Even though current AI capex remains strong, investors are now questioning how long that pace can last and whether political risk will cap future demand.

Neoclouds and AI-focused infrastructure players

Then there are the so-called “neoclouds”—companies whose entire business is building AI data centers and renting out compute capacity. Examples include CoreWeave, Iron, and Nebius.

These players don’t have large, diversified businesses to fall back on. They raise debt, build data centers, and rely on renting GPU time to pay it all back. If data center expansion slows, or borrowing costs spike, their model becomes much more fragile.

Musk-linked companies and political risk

One more group in the crosshairs: Elon Musk’s companies. Given open political hostility from Democrats toward Musk, investors worry regulators could make life harder for Tesla, SpaceX, and other Musk ventures, whether through approvals, subsidies, or policy friction.

Why hyperscalers might quietly benefit

Ironically, a slowdown in AI data center construction might be a long-term positive for the biggest cloud players: Microsoft, Amazon, and Google.

These hyperscalers have been spending at a torrid pace, transforming themselves from capital-light cash machines into capital-intensive, negative free cash flow businesses. Recent quarters tell the story:

• Google and Amazon together reported tens of billions in negative free cash flow
• Meta has been raising roughly $90 billion in equity to help fund its own AI and infrastructure plans

Investors are getting tired of writing blank checks for AI capex. If political pressure forces a slower buildout, the hyperscalers could:

• Return to positive free cash flow sooner, potentially as early as 2027
• Fund AI data center growth more from internal cash flow instead of constant capital raises
• Stretch the AI infrastructure story out over more years, trading “fast and fragile” for “slower but more durable”

In other words, the AI growth curve might flatten, but extend. For long-term investors, that could be healthier than the current boom-and-bust feel.

CoreWeave and the limits of AI romance

Even before the midterms, the AI narrative had already shifted once—this time because of CoreWeave. The company’s debt was repriced from a 9% yield to around 15%. At that level, borrowing to build new AI data centers starts to look prohibitive.

This repricing exposed a hard truth: the neoclouds are the weak hands in AI infrastructure. Unlike hyperscalers, they don’t have search, e-commerce, ads, or productivity suites throwing off cash. They’re fully dependent on the bond market to fund expansion.

Fixed income investors don’t care about the romance of AI. They care about:

• Cash flow
• Leverage
• Default risk

Once the bond market loses confidence, the cost of capital spikes, and the entire business model comes under pressure. That’s exactly what we’re now seeing with CoreWeave and its peers.

AI capex is now a macro story

AI spending is no longer just a tech sector issue. AI-related capex is estimated to account for at least half of US GDP growth this year. That’s a staggering number.

If AI data center capex slows, it doesn’t automatically mean a recession, but it does likely mean:

• Slower overall GDP growth
• Pressure on banks, investment banks, and alternative asset managers that finance AI and related projects
• A reset in valuations for AI-adjacent companies

This is also why debates about whether AI progress is stalling or simply normalizing have become so intense. For more on that, see our piece on why everyone is arguing about whether AI progress has slowed down.

Google’s business model under AI pressure

Recent Google results highlight a more subtle but critical shift: the possible replacement of classic search by AI-powered search and assistants.

Google’s latest quarter showed:

• Search revenue growth of 12%, down from 19% in early 2026
• Overall results that missed on several metrics

The worry isn’t just slower growth; it’s the quality of the business. Traditional Google Search is a capital-light, high-margin machine with strong user lock-in. In contrast, AI search and LLM-based assistants are:

• Capital-intensive, requiring huge data center and model training investments
• Lower-margin, given heavy compute costs
• Easier to switch between, since enterprises and developers can move from one model provider to another more easily

If AI search gradually replaces classic search, Google may be trading an extraordinary business for a tougher, more commoditized one.

Anthropic’s amended S-1: growth, “token maxing,” and a price war

Anthropic’s updated S-1 filing gave one of the clearest looks yet at how the economics of frontier AI are evolving. The new numbers confirmed a pattern many suspected: a post-hype slowdown.

From token maxing to normalization

In late 2025 and early 2026, Anthropic benefited from what can be called “token maxing.” Enterprises told employees to use AI aggressively and worry about costs later. That led to:

• Explosive sequential revenue growth of 135% and 147% in Q1 and Q2 2026
• A surge in efficiency, with revenue divided by compute expense rising from 0.69x in late 2025 to over 2x by Q2 2026

But by Q3 2026, the picture changed:

• Sequential revenue growth slowed to 47%
• The revenue-to-compute-expense ratio fell to 1.72x

Token maxing ended as CFOs woke up to the bill and enterprises gained the option to switch to cheaper open-weight models.

A full-blown price war

To re-ignite demand, Anthropic cut token prices by up to 50% in October. Early data from aggregators like OpenRouter suggests usage has increased, but likely not enough to fully offset the price cuts.

OpenAI and Google have followed with their own pricing moves, especially around new models like Gemini. The result is an emerging price war in LLMs, with:

• Falling prices per token
• Rising or flat compute costs
• Pressure on margins across the model provider landscape

At the same time, Anthropic’s amended S-1 revealed roughly $895 billion in total leasing and AI-related commitments. That implies an annual AI compute expense run rate approaching $100 billion in the near future and continued heavy cash burn.

Combine that with the tech correction and higher discount rates, and Anthropic’s upcoming IPO is likely to face a tougher valuation environment than many initially expected. For more background on the lab’s broader risk profile and philosophy, see our coverage of a former Anthropic researcher’s AI extinction warning.

The capital intensity problem across AI

Across the AI stack, several uncomfortable truths are now impossible to ignore:

• Capital intensity is soaring: AI has turned previously capital-light businesses (like cloud and search) into capital-hungry ones.
• Free cash flow is under pressure: Even giants like Google, Amazon, and Meta are raising debt and equity to fund AI capex.
• Neoclouds are at the mercy of bond markets: As CoreWeave’s 15% yield shows, financing can turn on a dime.
• LLMs lack strong moats: Switching between models is often cheap, which makes price wars more likely and durable margins harder to defend.

On top of that, power constraints and permitting issues are threatening to slow data center timelines, and now the political backdrop may add another layer of friction.

What recent earnings really tell us about AI

A wave of recent earnings reports from AI-adjacent companies was overshadowed by these bigger structural questions. Still, they offer useful snapshots of where we are.

Winners from current AI demand

Several companies showed strong results thanks to today’s still-robust AI capex:

• AMD: Reported 74% EPS growth and 45% revenue growth, with a big Q4 revenue step-up guided. This reflects heavy current demand for AI chips.
• Arista: Delivered 60% EPS growth and 58% revenue growth, plus upbeat guidance, driven by AI networking demand.
• Eaton: As an electrification and power management play with about 20% of revenue tied to AI construction, it posted strong double-digit growth and raised guidance.

These numbers confirm that AI spending today is still extremely strong. The debate is about sustainability, not current strength.

Partial AI plays and market anxiety

Other companies are more indirectly tied to AI and are feeling market nerves:

• Palantir: Nearly doubled revenue (up 96%) and beat on EPS. Its role is to help enterprises organize data and build AI applications, often sitting above the model layer. Despite strong results, the stock traded down as investors broadly de-risked AI names.
• Apollo: Reported solid fee-related earnings and asset growth, but spread-related earnings were softer. As a major financier of AI data centers, it is now seen as partially exposed to any slowdown or repricing in AI infrastructure.

Non-AI names and structural challenges

Some companies in adjacent or unrelated spaces highlight different kinds of risk:

• Apple: Posted solid revenue and EPS growth, but services growth slowed below 10% and guidance disappointed. Apple remains largely outside the core AI model race and appears content to let others bring LLMs to the iPhone platform on its terms.
• Payments players Fiserv and FIS: Both have struggled for years with competition and limited moats. Recent results were mixed, with FIS cutting guidance, pausing buybacks for years to deleverage, and losing its CFO—a combination that hammered the stock.

These cases underscore a theme: in sectors with weak barriers to entry, even modest shocks or strategic missteps can be brutal for equity holders.

Telecom, SpaceX, and a new front in AI-enabled competition

Outside the core AI stack, there’s a brewing battle in telecom. SpaceX announced a deal with Qualcomm to supply network technology for Starlink’s mobile service, signaling a serious move to challenge T-Mobile, AT&T, and Verizon.

With satellite-enabled mobile service on the horizon, the big three US carriers face a new, well-funded competitor that’s not afraid of heavy capex or long time horizons. Given this, and broader sector headwinds, the major mobile carriers look increasingly difficult to own.

FICO and the danger of losing a monopoly

Fair Isaac (FICO) delivered strong historical numbers—27% revenue growth and 43% EPS growth—but did something very rare: it refused to give guidance for the coming fiscal year.

The reason: intense competitive pressure from VantageScore and sweeping changes in mortgage credit scoring. Historically, FICO has only withheld guidance twice, during the global financial crisis and COVID. Doing so now suggests that what was once a near-monopoly franchise is under real threat.

In a world where AI and alternative data can power new credit models, legacy scoring monopolies may find their moats eroding faster than expected.

The AI narrative is still powerful—but no longer simple

For most of this year, the AI story felt straightforward: massive model improvements, sky-high demand, and a belief that “everyone wins.” That narrative is now splintering:

• Not everyone will benefit equally from AI; some business models are fragile.
• Capital markets, not just technology, will increasingly decide who scales and who stalls.
• Political and regulatory risk around power, data centers, and competition is rising.
• The economics of LLMs are under pressure from price wars and high compute costs.

AI remains a transformative technology, but the path forward looks more like a grinding, capital-heavy infrastructure buildout than a simple software supercycle. Understanding where the real bottlenecks, moats, and cash flows lie will matter far more than just betting on “AI” as a theme.

Share:

Comments

No comments yet. Be the first to share your thoughts!

More in Latest News