What a day in the stock market. Conventional wisdom holds that a Democrat sweep in the midterm election would be bad for stock prices. Boy, did the so-called experts get this one wrong.
Before moving on, let’s work through what people were thinking. The conventional view argued that an expected Democratic sweep would pressure stocks, mainly due to anticipated higher taxes, stricter regulation, and specific headwinds for the AI-driven market rally that had powered much of the recent gains.
Core Policy Concerns
- Taxes and earnings: Analysts commonly linked Democratic control to higher corporate taxes, capital-gains rates, or other levies that would reduce after-tax profits and investment incentives. California Democrats were pushing for a billionaires’ tax. Preliminary results suggest that Proposition 40 failed. Bank of America’s Michael Hartnett framed a sweep as a shift “from populist capitalism to populist socialism,” warning of more than 10% downside for U.S. stocks if it occurred, along with a weaker dollar and lower yields.
- Broader regulation: Expectations of tighter rules on business, finance, energy, and especially technology raised costs and uncertainty for corporate profits. Financials and industrials were frequently flagged as vulnerable; defense stocks faced risks of constrained spending growth. People are particularly worried about regulation that strangles the development of AI and energy.
The 2026 AI-Specific Angle
In the run-up to the 2026 midterms, the dominant market driver was the AI infrastructure boom. Democrats campaigned heavily against rapid data-center expansion, citing energy use, water consumption, noise, local community impacts, and subsidies. Key races, especially Texas, plus Georgia and Ohio, mattered because those states host a large share of planned data-center capacity.
Investors worried a sweep would bring moratoriums, permitting delays, limits on tax incentives, or greater scrutiny of AI/hyperscalers. This was expected to hit semiconductors, data-center operators, AI hardware, and related plays the hardest, while potentially favoring software or hyperscalers that already controlled scarce compute resources. Goldman Sachs and others noted that while a full national moratorium looked unlikely, state-level restrictions could still slow the buildout. This is where we agree with the consensus. If anyone would put the kibosh on data centers, it would be selective state governments.
Additional risks included congressional investigations into tech, crypto, or humans associated with the trump administration.
But Tuesday’s price action tossed all those arguments into the dustbin of history. Our view at the Options Edge was that it does not matter. History shows that the economy and the stock market are not affected by who is in control of the government. It performs about the same under each party. Companies and entrepreneurs find a way to succeed no matter what obstacles are put in their way. That said, everyone is worried about the rise of the Democratic Socialist movement. While the Democratic Party has historically supported high taxes, wealth redistribution, and heavy regulation, it has also been concerned about the private sector. Bill Clinton, for example, was a fiscal conservative and pro-business.
But the results of this election seem to hit a sweet spot. The democrats took over the House and Senate, but now by a wide enough margin to override the President’s veto without the help of republicans. As a result, we can expect a legislative stalemate. We do not expect any significant legislative accomplishments for the next two years. But we do expect robust debate on the budget.
We think the AI phenomenon will continue to move forward. It may do so at a slower pace in some states, but data centers will continue to be built. That being the case, we think it makes sense to review the industry for investment opportunities. The following is an overview of the products (hardware and systems) that go into a modern AI data center, along with the primary companies that make them.
Compute / Accelerators
- GPUs / AI accelerators: Nvidia (dominant, e.g., Blackwell/GB200 series and rack-scale systems); AMD (Instinct series); custom ASICs from Google (TPUs), Amazon (Trainium), and others often involving Broadcom or similar design partners.
- CPUs: Nvidia (Grace), Intel, AMD, Arm-based designs.
- High-bandwidth memory (HBM): SK Hynix, Samsung, Micron (critical for AI performance; prices have risen sharply).
- Server/rack systems and integration: Dell, Supermicro, HPE, Lenovo (OEMs); Foxconn/Hon Hai, Quanta (QCT), Wistron, Wiwynn, Inventec, Celestica (ODMs that design and assemble full racks for hyperscalers). Nvidia also supplies complete systems.
Networking
- Switches and networking silicon: Broadcom (major), Nvidia (Spectrum Ethernet, InfiniBand/Quantum), Arista Networks, Cisco, Marvell.
- Optical transceivers, optics, and DSPs: Coherent, Lumentum, Zhongji Innolight, Eoptolink; Broadcom and Marvell for related chips.
- Fiber, cables, and connectors: Corning (fiber/cable), Amphenol (high-speed connectors and assemblies), TE Connectivity, and others.
- Interconnect/retimer chips: Astera Labs, Credo, and similar specialists for high-speed links within clusters.
Power Infrastructure
- Internal distribution: Schneider Electric, Vertiv, Eaton, ABB, Siemens, Delta Electronics (strong in high-density/AI power, including 800V designs).
- Transformers and high-voltage gear: GE Vernova, Hitachi Energy, Siemens Energy, HD Hyundai Electric, and others (many face multi-year backlogs).
- On-site/backup generation: GE Vernova, Siemens Energy, Mitsubishi (gas turbines); Caterpillar, Cummins, Rolls-Royce (generators/engines). Some sites use Bloom Energy or similar for specialized power.
Cooling
- Liquid cooling systems: Vertiv, Schneider Electric, CoolIT Systems (Ecolab), Delta Electronics, Asia Vital Components, Auras Technology, Shenzhen Envicool, and others in Nvidia’s ecosystem.
- Chillers, HVAC, and traditional thermal: Carrier, Trane (Trane Technologies), Johnson Controls, Daikin.
- Supporting components: Heat exchangers (Alfa Laval, Dover/SWEP, Kelvion, Danfoss); quick-disconnect couplings (Dover/CPC, Stäubli, Parker Hannifin).
Other Key Components and Foundry
- Advanced packaging and substrates: TSMC, Amkor, Ibiden, Unimicron, Samsung Electro-Mechanics, and others for substrates/packaging capacity (Nvidia has been investing to expand this).
- Printed Circuit Boards: Various Taiwanese, Japanese, and Chinese suppliers (e.g., Unimicron, Ibiden, and others qualified for high-layer AI boards).
- Passives and power electronics: Murata, TDK, Infineon, and similar for capacitors, VRMs, etc.
- Foundry (chip fabrication): Primarily TSMC.
Other Notes and Comments
- Nvidia sits across multiple layers (GPUs, full racks, networking) and has a large partner ecosystem. Hyperscalers (Microsoft, Amazon, Google, Meta) often co-design or specify custom versions and are the biggest buyers.
- Traditional data centers use more air cooling and lower-density power; AI facilities shift heavily toward liquid cooling and denser electrical systems because racks can draw 100–300+ kW (vs. ~10–20 kW historically).
- Supply is constrained in several areas (HBM, power equipment such as transformers/turbines, advanced packaging), contributing to rising costs and long lead times.
- Construction/site work, grid interconnection, and operations involve additional firms (e.g., Quanta Services for transmission, data-center developers like QTS/Vantage), but the list above covers the core equipment bill of materials.
We do not have any specific suggestions at this time. We thought it made sense to review the buildout and the companies that benefit from data center construction, since the share prices of many of these companies are in the dumps. Some stocks have been down over 50% from their highs.
Within the space of the Neoclouds, stocks of companies like CoreWeave and Nebius on the big side, and crypto firms like Hut 8 and Galixy, on the smaller side, are transforming mining operations into AI data centers to capture much higher margins. It makes sense for companies like Hut 8 to pivot as they already have the infrastructure in place. They just have to remove bitcoin mining machines and replace them with GPU servers.
There are other non-tech companies that are involved as well. CAT is supplying construction equipment, for example. We just wanted to list a host of companies, and we suggest our readers focus on those they are familiar with. This is where one can generate an edge and beat the market on a risk-adjusted basis.
