The AI industry's super PACs spent most of their money in primaries

The day after the November 3 election, Fortune tallied how candidates backed by the AI industry's super PACs had done. It is a natural question, and a scoreboard is a natural way to answer it. But the FEC records suggest the scoreboard looks at the wrong night. Most of this money was not spent in November at all.

Two rival networks did most of the spending. Leading the Future, which Fortune describes as favoring light regulation of AI, funds a Democratic arm called Think Big and a Republican arm called American Mission. Public First, which Fortune describes as favoring stronger safeguards, has its own pair: Jobs and Democracy PAC on the Democratic side and Defending Our Values PAC on the Republican side. Together the four arms reported $59M in independent expenditures this cycle. Of that, 83% went into primaries and primary runoffs, the contests where a party's own voters choose its nominee.

Mostly primaries

Split by stage, the networks look less like general-election players and more like primary specialists. Leading the Future's two arms put 75% of their $35M into nomination contests. Public First's arms put 95% of their $24M there. Across all four arms, only $7.9M, or 13%, was general-election spending.

That matters for how the money gets judged. A primary often has a handful of candidates from the same party and very little outside spending. A few million dollars there can be a large share of everything said about the race. A general election for a competitive seat draws money from the parties, their super PACs and dozens of other groups. The same check is a smaller voice in a louder room. Our meddling score scores every large outside group on exactly this: its share of money spent in primaries, and how much of the conversation in its races it accounted for.

And nowhere was the primary focus sharper than in one district.

In New York's 12th District, the two networks met head-on. Jobs and Democracy PAC reported $13M supporting Alexander Bores in the June Democratic primary. Think Big reported $8.2M opposing him. For each of the two Democratic arms, this was the single largest contest of the cycle. Together, $21M went into one primary for one House seat, more than a third of everything the four arms spent.

Both networks describe their purpose in terms of AI policy, according to Fortune, yet the contest that drew the most money from either of them was a choice between Democrats for one party's nomination. Our race data covers the general election, not the primary vote, so this essay does not report the primary result; the NY-12 race page has the November race.

Where the money came from

Each network is a funnel: a parent group collects the money, then passes it to its partisan arms. The FEC records show each hop separately, and they reach further back on one side than the other.

On the Leading the Future side, the arms' filings show $45M arriving from the parent super PAC, split evenly between the Democratic and Republican arms. One step further back, Fortune reports, citing Leading the Future's own FEC filings, that the parent is funded largely by Andreessen Horowitz co-founders Marc Andreessen and Ben Horowitz and by OpenAI president Greg Brockman and Anna Brockman. Those gifts to the parent are not in our dataset, so we link to the source rather than draw the line ourselves.

On the Public First side, most of the money, $23M, came from Public First Action Inc, a nonprofit. Nonprofits of this kind do not have to name their donors, so the trail on this side ends one hop earlier. That is a real asymmetry: one network's backers can be named from public filings, the other's largely cannot.

What reached November

The general-election money was small and, for the most part, spread thin. American Mission reported the most, $5.5M across 17 contests. Its largest general-election bet was $1.6M supporting Darline Graham in the South Carolina Senate race; 14 of its other general-election contests drew less than $300,000 each. Think Big's $1.3M went to 4 contests, all supporting Democratic House candidates. Only $599K of all the general-election money opposed a candidate.

Read against Exhibit 1, the pattern is plain. In the general election these groups mostly supported nominees, in modest amounts. Their large sums came earlier, in the contests that chose those nominees.

What this does and doesn't show

The records show where four committees reported spending, at which stage, and who gave to them. They do not show whether that spending changed any result, in a primary or in November, and this essay makes no such claim. A candidate who won with a group's support may well have won without it.

They also do not show everything the AI industry spent. Fortune, citing Tech Influence Watch, puts AI super PAC spending at $79 million across 68 seats; our figure covers only these four arms. Money given directly to candidates, party committees or other groups is not here, and the nonprofit at the top of the Public First network does not disclose its donors. Stage coding depends on how each group filled in its reports; $2.1M could not be placed in a stage and is shown separately.

Finally, these are preliminary figures. Notices run through October 26, and the committees' full post-general reports are due December 3. We will update the numbers then.