Midterm Ballot Rejections Test Data-Centre Expansion and Power Deals

Midterm voters rejected data-centre ballot measures in all but one race, hurting power and neocloud stocks even as big tech platforms rallied.

Market data from the close — the last completed trading session before publication.

The setup — Voters in Tuesday's midterm elections pushed back on data-centre construction in all but one state ballot measure, driving a wedge through the AI infrastructure complex in Wednesday's session (per CNBC). While the largest tech platforms rallied as election uncertainty cleared, power suppliers and debt-reliant cloud builders sold off (per CNBC). That ballot resistance strengthens the case that existing computing capacity is becoming harder to replicate, while directly challenging the assumption that new grid and data-centre projects will move forward on schedule.

What changed for the themes

Our AI compute build-out theme picked up mixed evidence on Wednesday. Before the market opened, Advanced Micro Devices ($AMD) reported a record $13.4 billion third quarter where data-centre revenue nearly doubled on AI chip sales, supporting the thesis that accelerator demand continues to outrun supply (per CNBC and Ars Technica). After the close, Qualcomm ($QCOM) reinforced the narrative that advanced memory is becoming a supply bottleneck, issuing a weak forecast as memory shortages weighed on a 20% drop in smartphone sales, even as it detailed new data-centre partnerships with Amazon ($AMZN) and Meta Platforms ($META) (per CNBC and The Verge). Working against the hardware build-out, voters rejected data-centre expansion in every midterm ballot measure except one, bringing concrete evidence to the narrative that local opposition is limiting where facilities can be built (per CNBC).

Our AI platforms and cloud theme saw its largest companies rally during Wednesday's session as midterm results cleared electoral uncertainty (per CNBC). Following our recent coverage of Meta's planned $20 billion to $25 billion bond sale, commentary published after Wednesday's close noted that local permitting slowdowns could raise the value of existing computing capacity already owned by the cloud giants, while putting further pressure on debt-funded cloud challengers (per CNBC). Still, the narrative that power and local limits are slowing data-centre expansion gained sharper edge from the bipartisan voter pushback (per TechCrunch).

Our AI power and critical minerals theme faced new political friction alongside evidence of long-term utility demand. During Wednesday's session, Utility Dive reported that Arizona Public Service filed a resource plan projecting 3.4% annual peak demand growth over 15 years — expecting sales to double by 2041 on industrial and data-centre loads — while industry groups warned that a federal ban on foreign grid components could inflate grid investment costs (per Utility Dive and Utility Dive). After the close, in a direct response to voter anger over power bills and local grid strain, OpenAI pledged to pay for its own data-centre power, publish construction plans and protect residential ratepayers (per CNBC). That pledge tests the narrative that rising power bills are turning regulators and communities against data centres, which would challenge the theme if infrastructure costs are forced back onto builders or projects are delayed.

Our AI software and agents theme saw two contrasting developments after Wednesday's close. Meta acquired AI coding startup Conductor, which lets software engineers deploy teams of agents simultaneously, supporting the move of enterprise agents into production (per TechCrunch). Conversely, design software maker Figma saw its shares sink after reporting that it will allow clients to use more AI features for the same price, underscoring the risk that vendors may struggle to charge extra for agentic tools (per CNBC). Nothing material changed in the session's filings or coverage for our robotics and physical AI theme.

Risks to watch

The overarching risk across both hardware and cloud platforms is that AI capital spending is outrunning its financial returns. Our dashboard reads the AI infrastructure capital-spending cycle in saturation — with outlays running 107.6% above last year, more than two standard deviations above their ten-year history — while our composite stress gauge sits at an elevated 67 out of 100, where high-yield credit spreads have widened 28 basis points over the past month to 340 basis points.

When evaluating how higher borrowing costs and midterm construction delays strain the AI build-out, investors must separate companies that fund capital spending from operating cash flow from those that borrow to build. By our readings, Microsoft ($MSFT) paid for all of its capital spending over the past year — equal to 39% of revenue — out of operating cash flow and holds more cash than debt. By contrast, Amazon spent 128% of its operating cash flow on capital projects while its long-term debt rose 228%, Oracle ($ORCL) carries an active warning on cash burn with rising debt, and specialised cloud builder CoreWeave ($CRWV) spent 298% of its operating cash flow on capital outlays as its long-term debt rose 185%. If voter-backed local opposition delays data-centre energisation into 2028 (per MarketWatch), companies paying interest on borrowed money for unlit capacity face immediate financial strain.

In power, the risk that demand falls short of utility forecasts — highlighted earlier this week by lower large-load forecasts in the PJM grid — is now compounded by local ballot resistance and component import bans. Confirming that risk would require utilities to trim capital plans in upcoming earnings calls; retiring it would require signed, ratepayer-insulated power contracts moving into construction.

The companies

Microsoft ($MSFT), which rides our AI platforms and AI software themes, rose 6.7% to $555.80 on Wednesday — about 4.9 times its typical daily move after allowing for how the tracked companies moved as a group — as large tech platforms rallied on the removal of midterm election uncertainty (per CNBC). Two current reports on Form 8-K filed after Wednesday's close list no text in our records and cannot explain the session's gain. Wednesday's rally pulls against several of our readings: while Microsoft's contracted backlog — future revenue locked in under customer contracts — grew 72% over the past year against 18% revenue growth, the stock trades at the 100th percentile of its own valuation history, and its price implies about 27% annual cash-flow growth against the 4% annual decline it actually delivered.

Amazon ($AMZN), riding our AI platforms theme, gained 6.1% to $266.03, or 4.5 times its typical daily move relative to the tracked universe, on the same post-election rally in large tech platforms (per CNBC). That move contrasts with our adverse narrative reading on the company, given that its capital spending reached 128% of operating cash flow over the past year. Apple ($AAPL) also climbed 5.3% to $349.51, about 5.5 times its typical relative move, on the post-midterm surge (per CNBC); by our readings, Apple sits at the 97th percentile of its own valuation history and faces adverse narrative pressure from memory shortages squeezing device makers.

Intuitive Surgical ($ISRG), which rides our robotics theme, rose 9.4% to $482.10, about 5.8 times its typical daily move after allowing for the group. Nothing in the day's filings or coverage explains the move; a Form 8-K filed Monday afternoon has no text in our records.

On the downside, Vistra ($VST) and GE Vernova ($GEV), both riding our AI power theme, fell 13.8% to $120.29 (3.5 times Vistra's typical relative move) and 9.3% to $832.83. Data-centre stocks fell during Wednesday's session following the midterm elections (per CNBC), though coverage detailing the specific ballot rejections appeared after the close; Vistra also filed an 8-K early Thursday morning whose text is not in our records. By our readings, Vistra's operating cash flow covered its capital spending over the past year, but its long-term debt still rose 24%. Finally, cloud builders CoreWeave ($CRWV) and Nebius Group ($NBIS) dropped 9.9% to $48.42 and 6.7% to $141.52 alongside the broader pullback in data-centre equities (per CNBC). Our readings place CoreWeave in our risk group, noting warning language in its latest filing and operating profit that fell short of its interest expense, which consumed 21% of revenue over the past year.

The backdrop

Wednesday's session split cleanly between major market indices and the companies building AI infrastructure. The S&P 500 ETF ($SPY) rose 1.7%, the Nasdaq Composite gained 1.4% and the Nasdaq 100 ETF ($QQQ) added 1.0%, while the Volatility Index ($VIX) dropped 20.3% to 16.61 and the 10-year Treasury yield fell six basis points to 4.89%. WTI crude oil fell 8.3% to $88.12 as gold rose 3.4% to $4,073.00. Underneath the index gains, the Semiconductor ETF ($SMH) fell 2.8% while the Software ETF ($IGV) rose 0.9% — where Salesforce ($CRM) still screens on our dashboard as cheap against its own past. Across the 67 companies we track, 48 fell and only 19 rose, with a median drop of 3.2% and an overall median valuation at the 65th percentile of its own history.

What to watch

  • Power generators' earnings and ballot fallout: Vistra ($VST) is expected to report results before Friday's open, followed by Constellation Energy ($CEG) before Monday's open. Investors should watch for commentary on whether local midterm ballot rejections and revised regional load forecasts are delaying data-centre power contracts.
  • Robotics and industrial demand: Rockwell Automation ($ROK) is scheduled to report before Thursday's open, offering a read on factory automation investment for our robotics theme, while USA Rare Earth ($USAR) is expected to report after Monday's close for our critical minerals theme.
  • Debt-funded cloud economics: CoreWeave ($CRWV) is expected to report after Tuesday's close. Watch its interest coverage, cash burn and capital-spending commitments for evidence of whether credit is tightening for leveraged AI cloud builders.
  • Semiconductor equipment orders: Applied Materials ($AMAT) is scheduled to report after the close on Thursday, November 12, testing whether memory bottlenecks and AI chip demand are translating into wafer-fab equipment orders.

Companies in this brief

Research pages with valuation, cash generation, trend and context. Moves are today’s.