Hard Fork on the Midterms, AI Backlash and Tech Political Spending

Guest host Max Read welcomes New York Times technology reporter Erin Griffith and opinion columnist Ezra Klein. They discuss the midterm results and what they may mean for the visible group of Silicon Valley executives and investors associated with Donald Trump.

Read names investors and executives who had been prominent in the coalition, including David Sacks, Marc Andreessen and Elon Musk. He asks whether the week marked the collapse of their political influence.

Read recalls that the so-called tech right appeared influential in the Trump coalition, drawing attention through campaign contributions and government roles. He asks how its position changed after the election.

Klein argues that the group tied itself to an administration whose AI position was unpopular. He cites exit-poll figures in the discussion: 56% called AI and data centers a crucial or major voting factor, while 76% opposed a data center in their community. Those numbers are presented as his account of the polling.

Klein separates AI from other issues on which he says the administration had also become unpopular, including tariffs and the war in Iran. He regards data centers as a local issue that could be tied to individual candidates.

Read asks whether AI harmed the tech right politically or the tech right harmed AI’s reputation. Klein says regulation can create trust by making risks and trade-offs accountable; in his view, the people driving the rollout have not adequately built that trust.

Klein says public anxiety includes possible job losses and safety risks, while the people promoting AI are not always trusted to manage those risks. Read and Griffith discuss the tone in which parts of the industry have presented their plans.

Griffith describes a strain of tech-industry confidence that products will eventually win people over regardless of initial opposition. She questions the assumption that frequent use of existing platforms proves trust or consent to every aspect of AI development.

She also discusses why some investors moved toward Trump, citing opposition to regulation and an online culture that rejected prevailing social norms. Griffith says it remains uncertain whether the election will quickly reverse those cultural commitments.

Griffith says the cultural change extended beyond a few executives as online communities adopted the language and attitudes of prominent investors. She is watching whether the election alters that broader culture.

She notes that investors sometimes point to people’s reliance on phones, online shopping and social platforms as evidence that they will accept AI. In her view, adoption and confidence in its builders are separate questions.

Read separates three issues: general opposition to AI, local opposition to data centers and resentment of billionaire influence. Klein says the industry has explained how it intends to build AI but has not persuaded many people why they should want more of it.

Read plays an excerpt from a victory speech criticizing Silicon Valley leaders and their warnings about AI. The panel treats the explicit language as evidence of AI’s growing place in political rhetoric.

Klein argues that large data centers can face local resistance as industrial facilities, independently of wider AI fears. He says some politicians are developing technical fluency and are not willing simply to accept assurances that the risks will resolve themselves.

Griffith asks whether the industry sees this merely as a messaging problem. Klein distinguishes people who think AI progress is inevitable despite risks from those enthusiastic about rapid development. Griffith adds that some investors focus less on existential risk and more on regulation and access to compute.

Klein says some AI employees tell policymakers that they would prefer a pause but doubt one is feasible amid competition. Griffith contrasts that attitude with investors who see infrastructure and regulation as obstacles to growth.

On data centers, the panel distinguishes issue salience from the outcome of individual ballot questions. Klein says opposition was especially important in states with extensive construction, including Texas and Ohio, while acknowledging it did not explain every race.

Griffith notes that data centers can rank below other issues when voters are asked to name their top concern, while a direct local yes-or-no question can produce strong opposition. She says the distinction matters when interpreting election effects.

Klein emphasizes the geographic variation: he believes the issue was especially salient where many facilities were planned, rather than equally decisive across the country. Read notes that construction may shift to other states rather than stop completely.

Griffith says investors worry that permitting and political changes could delay substantial planned capacity, increasing compute costs. She also describes more dramatic reactions that forecast broader market or geopolitical consequences, treating those as investor fears rather than established outcomes.

The panel describes some extreme investor reactions after the vote. Klein contrasts those fears with the market’s subsequent movement and says a spending race can itself create economic fragility.

Klein offers a different possibility: if companies are overbuilding to avoid being left behind, an external slowdown could reduce capital expenditure risk. He argues that markets need some alignment between construction, useful products and adoption; slower construction is not necessarily bad for investors.

In Klein’s view, companies that focus on making existing AI economically useful could benefit from a more measured pace. He frames this as a possibility, not a prediction that every project will be better off delayed.

Read turns to billionaire spending and asks whether Elon Musk’s visible involvement was counterproductive. Klein thinks it made Musk an effective political foil for opponents in Texas, because the spending itself drew attention to claims that a wealthy donor was trying to influence the election.

The panel notes an apparent complication: a California wealth-tax proposition failed despite wider criticism of billionaires. Griffith suggests voters may have doubted the proposal’s design, while Klein points to concerns that the state could lose wealthy residents. Both treat the interpretation as provisional.

Read asks Griffith why California voters rejected the wealth-tax measure in an election the panel sees as broadly critical of billionaire influence. She cites concerns about how the proposal was written and administered.

Klein predicts that national wealth-tax proposals could remain part of future Democratic politics, arguing that a national policy avoids some state-level mobility concerns. The speakers do not treat the California vote as conclusive evidence that all such proposals will fail.

Read raises Vivek Ramaswamy’s Ohio campaign and a controversial advertisement involving his opponent’s family. Klein sharply criticizes the ad and argues that candidates associated with the tech right cannot assume money and online influence will translate into electoral appeal.

The Ohio discussion becomes a disagreement about campaign tactics and candidate appeal. Klein gives a strongly negative assessment of the Ramaswamy advertisement and of attempts to imitate Trump’s political style.

Klein argues that a group with money and influence can still struggle when its own candidates must appeal directly to a broad electorate. The point is his interpretation of the coalition, not a measurement of every voter’s motivations.

Griffith says some wealthy tech figures appear to be learning that political success takes more than spending. Asked whether Silicon Valley will shift back culturally, she expects some lobbying and relationship-building across party lines but does not expect established personal views to disappear overnight.

Griffith notes that Democratic-aligned tech donor Reed Hoffman publicly welcomed the results, while also questioning the tension between advocating less money in politics and making large donations.

Klein suggests a less overtly partisan chief executive may be better positioned than executives who act like political influencers. Read also observes that a newer generation of Democratic politicians may be more technically informed and more critical of the industry than the leaders associated with the earlier social-media boom.

Read contrasts the party’s current relationship with technology with an earlier period when social-media companies were closely associated with Democratic optimism. He asks what kind of politicians and executives will emerge next.

Klein asks what a political party positively wants from technology, not only what harms it wants to avoid. He contrasts earlier optimism about the internet and clean energy with a more ambivalent public response to AI, including concerns about jobs, safety and the kind of future being proposed.

Klein says a positive case for AI must address the future people actually want, not just promise more capability. He raises uncertainty about work and what children should prepare for as examples of questions that need answers.

In a segment on stories to watch, Griffith discusses reporting that Google planned to disable search for schools that did not want its AI products, and says the company later clarified that it would work through the issue. She sees the dispute as part of a wider debate over technology in education.

Read says education could become a major political issue for parents concerned about AI. Griffith sees the school search dispute as an example of a company’s product choices colliding with those concerns.

Klein is watching a bipartisan permitting reform proposal that he says addresses energy construction and some data-center costs. He sees its progress as a test of bipartisan bargaining and of whether lawmakers can agree on how quickly energy and data infrastructure should be built.

Klein says the permitting proposal includes senators from both parties and covers different forms of energy. He is watching whether the lame-duck session produces a deal or reveals a lack of consensus.

Read says prediction markets seemed more integrated into election coverage than in previous cycles. Klein cautions against treating quick price movements as proof of special foresight, and the panel notes that markets can follow polls or rapidly react to new information.

The speakers debate how to interpret prediction-market prices. They agree that a line moving rapidly as actual results arrive is not the same as an independently verified forecast.

After the break, Read speaks with New York Times reporter Theodore Schleifer about technology donors. Schleifer says the top 20 donors contributed about $2 billion to the midterms, more than $800 million of it linked in some way to tech. He estimates Musk’s disclosed spending at roughly $470 million to $500 million.

Schleifer names a range of donors and committees, including Andreessen Horowitz, Coinbase and Greg Brockman. He separates their stated goals from the question of what any single contribution actually accomplished.

Schleifer says the election illustrates the limits of a simple rule that more money always wins. He notes that disclosed giving does not capture all political spending and cautions that the apparent success of an individual donor is difficult to isolate from the broader political environment.

He reflects on how the perception that Musk’s 2024 spending was decisive helped reinforce a narrative of money buying political success. The new results, he says, show that spending can fail to convert into victory.

They discuss Reed Hoffman’s support for James Talarico in Texas and Sergey Brin’s spending against a California wealth tax. Schleifer says the outcomes aligned with those donors’ preferences, while noting that it is hard to know whether the same results would have occurred without their money.

Schleifer describes Brin as relatively new to major political giving and says a well-funded campaign opposed the California tax. He contrasts that effort with the far smaller amount raised by supporters of the measure.

Schleifer then describes crypto-backed political spending through Fairshake, supported by companies and investors including Coinbase and Andreessen Horowitz. He says the group had mixed primary results and spent heavily against Sherrod Brown in Ohio, where Brown nevertheless won.

Brown’s race is especially meaningful to the crypto industry, Schleifer says, because he has been involved in oversight of financial policy. He cautions that crypto-backed candidates also won some races, so one contest is not the whole picture.

Read asks whether that race weakens the idea that the crypto industry can punish every opponent. Schleifer says a large political fund can also serve as a standing threat even before it is spent, and connects the spending to conflict over the Clarity Act.

The conversation turns to an AI-focused political committee, Leading the Future. Schleifer says some of the same investors applied a model used in crypto to AI, but argues that public opinion about AI is more intense and politically salient.

Schleifer says AI and crypto differ in public salience. He argues that taking industry support can carry a different political cost when voters have stronger opinions about the underlying technology.

Schleifer reports that OpenAI co-founder Greg Brockman and his wife committed $50 million to the group but did not make a planned second $25 million installment. He attributes part of the decision to concern that the support had become a distraction inside the company.

He describes a planned donation being reduced as the politics around AI changed. Read links the episode to internal employee reactions, while Schleifer notes the distinction between a public pledge and money actually paid.

Read sees signs of renewed employee pressure on technology executives over politics. Schleifer adds that their discussion covers disclosed donations; money can also flow through organizations that do not reveal the original donor, limiting what reporters and voters can see.

Both speakers note that dark-money structures can obscure the ultimate funder. Their ranking therefore concerns only the activity visible in records and reporting.

On Musk, Schleifer recalls the 2025 Wisconsin judicial race, after which he says Musk became less publicly visible in his political support. In the current cycle, a large share of spending was concentrated in Texas. Read and Schleifer discuss whether Musk’s prominence made him a convenient opponent regardless of the spending.

Schleifer says Musk sought a lower public profile after a Wisconsin setback, though he continued to spend. The speakers say his status as a national figure may itself have made him part of the Texas contest.

Schleifer stresses that the Texas result cannot be attributed solely to Musk’s money: the candidate might have lost without it. He nevertheless argues that a donor can spend heavily, become identified with a race and fail to achieve the desired result.

Looking ahead to 2028, Schleifer expects donors to remain active but says industry-specific support will depend on candidates’ positions. He is also watching people concerned about AI safety and the potential political influence of newly wealthy employees and investors.

Future giving, Schleifer suggests, could come from people with opposing views of AI: those seeking faster adoption and those focused on safety and slowing development. Candidate positions will affect where that money goes.

He also expects political fundraisers to watch the wealth created by technology companies and their investors. He treats the amount and direction of that future giving as uncertain.

Schleifer closes with a limitation of campaign-finance reporting: more money may pass through newly created nonprofits whose original funders are difficult to identify. He says there can be reasons donors seek privacy, but the trade-off is that the public knows less about who is funding political activity.