Axios Pro Rata: Midterm predictions

1. Key Themes

Prediction markets' reputational fragility is tied to election outcomes

Prediction markets earn revenue regardless of outcome, but their credibility, and therefore their growth, depends on being seen as accurate.

• "Prediction markets like Kalshi and Polymarket officially don't care which side of a bet wins. There is no 'house.' They get their cut either way. It's a business model that's attracted billions of VC dollars."
• "Prediction markets burst into the public consciousness two years ago, when their users correctly 'picked' Donald Trump to beat Kamala Harris. It bestowed credibility at a time when public trust in traditional polling had cratered."
• "More recently, however, they had electoral misses in Brazil and Israel. Three would be a trend."

Political markets are a smaller revenue line but matter for long-term legitimacy

• "Political predictions generate much less revenue for Kalshi and Polymarket than do sports predictions, but they have a much longer tail."

Refining capacity scarcity is driving consolidation

Wide margins and constrained global capacity are pushing buyers to acquire existing refineries rather than build.

• "Refining margins are wide, due to supply and demand imbalances caused by the Russia/Ukraine war and conflicts in the Middle East."
• "Companies of all sizes have been on the hunt to acquire more refineries to build bigger fuel-making businesses, given the almost intractable constraints in global capacity." — Collin Eaton, WSJ
• The deal itself: "CVR Energy (NYSE: CVI), controlled by Carl Icahn, agreed to buy two U.S. refineries and a related marketing business from Marathon Petroleum (NYSE: MPC) for $4.2 billion," with the businesses generating "around $1.2 billion in EBITDA for the 12 months ending Sept. 30, 2026." That implies roughly 3.5x trailing EBITDA.

AI-layer venture funding is broad: custom models, compute markets, post-training, and agents

Capital is flowing across the AI stack, with Nvidia and top-tier funds repeatedly on cap tables.

• "Applied Compute, a developer of custom open-source AI models for enterprises, raised $350m at a $3.25b valuation. Elad Gil led, joined by Kleiner Perkins, Lux Capital... Nvidia, Blackstone, Shopify Ventures, HubSpot Ventures."
• "Ornn, a compute-trading startup, is raising a triple-tranched $150m in Series A funding led by Thrive Capital and General Catalyst... The final tranche could be at a valuation north of $1b."
• "Trajectory, a tool to post-train models, raised $45m in Series A funding at a $300m post-money valuation. Sequoia Capital led, joined by Nvidia, Bessemer Venture Partners, and Bain Capital Ventures."

Incumbents are buying AI startups, especially in security and customer service

• "Salesforce (Nasdaq: CRM) agreed to acquire Cresta, a Sunnyvale, Calif.-based AI call-center company, for a reported $2.7b. Cresta had raised more than $280m."
• "Cognition, valued by VCs at $46b, acquired Corridor, an AI code-security startup that had raised $31m."
• "CrowdStrike (Nasdaq: CRWD) agreed to acquire Dreadnode, a provider of automated AI security testing that had raised $14m."
• "Alphabet (Nasdaq: GOOG) is in late-stage talks to acquire Howie AI, an AI scheduling assistant for businesses."

The AI buildout is increasingly debt-financed, and bond markets are pushing back

• "Big Tech's debt addiction is coming with increased borrowing costs... The bond market is pushing back on the AI trade, but the AI trade pushes on as Alphabet, Microsoft, Meta, and Oracle all line up billions in new bond offerings."
• Related infrastructure capital: "Brookfield Asset Management and QIA agreed to invest £2.5b into a U.K. data center project with Drax Group."

2. Contrarian Perspectives

Prediction markets should root for Democrats, even though Democrats are more hostile to them

The counterintuitive claim is that the platforms' business interest lies in a Democratic sweep, because that outcome matches market odds and protects their accuracy narrative. A Republican upset would hurt more than a Democratic Congress would.

• "Kalshi and Polymarket should be pulling hard for Democrats in today's midterm elections. Even if that feels counterintuitive."
• "Both platforms currently have markets on Democrats sweeping Congress, with 'yes' at around 75%. If that's how today plays out, it would further cement prediction market legitimacy. A Republican win in either chamber, however, would dump them into the same pot as pollsters — even if 75% probability includes a sizable improbability."

A Democratic Congress is a limited regulatory threat to prediction markets

• "Democrats are often more hostile to prediction markets than are Republicans, but companies like Kalshi and Polymarket don't have too much to fear from a Democrat-controlled Congress."
• "Yes, there would probably be an uncomfortable hearing or two. But neither chamber would have veto-proof majorities, and the Trump administration has been very sympathetic to the industry."
• "The real constraints, if any, are more likely to come via the Supreme Court."

3. Companies Identified

Kalshi / Polymarket

• Description: Leading prediction market platforms.
• Why mentioned: Lead case study on how a no-house business model still depends on perceived accuracy.
• Quote: "It's a business model that's attracted billions of VC dollars."

CVR Energy (NYSE: CVI)

• Description: Refiner controlled by Carl Icahn.
• Why mentioned: Acquiring two refineries and a marketing business for $4.2B.
• Quote: "CVR Energy... agreed to buy two U.S. refineries and a related marketing business from Marathon Petroleum (NYSE: MPC) for $4.2 billion."

Marathon Petroleum (NYSE: MPC)

• Description: Refining and marketing company.
• Why mentioned: Seller of the Minnesota and North Dakota refineries.
• Quote: "The refineries are located in Minnesota and North Dakota."

Applied Compute

• Description: Developer of custom open-source AI models for enterprises.
• Why mentioned: Largest VC round in the issue.
• Quote: "raised $350m at a $3.25b valuation. Elad Gil led."

Ornn

• Description: Compute-trading startup.
• Why mentioned: Signals a financialization of compute as a commodity.
• Quote: "raising a triple-tranched $150m in Series A funding led by Thrive Capital and General Catalyst."

Trajectory

• Description: Tool to post-train models.
• Why mentioned: Sequoia-led Series A at $300M post-money.
• Quote: "raised $45m in Series A funding at a $300m post-money valuation."

Tutor Intelligence

• Description: Watertown, Mass.-based developer of warehouse robotics.
• Why mentioned: Physical AI/robotics funding.
• Quote: "raised $75m in Series B funding. A16z led, joined by USV and Neo."

Corridian

• Description: Boston-based maker of "ultra-large" autonomous underwater vehicles.
• Why mentioned: Defense-tech seed round.
• Quote: "raised $38m in seed funding, per Axios. Point72 Ventures led."

Cognition

• Description: AI coding company valued at $46B.
• Why mentioned: Acquirer of Corridor.
• Quote: "Cognition, valued by VCs at $46b, acquired Corridor, an AI code-security startup."

Salesforce / Cresta

• Description: Enterprise software giant and AI call-center company.
• Why mentioned: $2.7B reported acquisition.
• Quote: "Salesforce... agreed to acquire Cresta... for a reported $2.7b."

CrowdStrike / Dreadnode

• Description: Cybersecurity leader and AI security-testing startup.
• Why mentioned: Security consolidation around AI.
• Quote: "CrowdStrike... agreed to acquire Dreadnode, a provider of automated AI security testing."

Alphabet / Howie AI

• Description: Tech giant and AI scheduling assistant.
• Why mentioned: Late-stage acquisition talks.
• Quote: "Alphabet... is in late-stage talks to acquire Howie AI."

Grant Thornton / CBIZ

• Description: Accounting firm, backed by New Mountain Capital, and its acquisition target.
• Why mentioned: Large accounting-sector PE roll-up.
• Quote: "Grant Thornton completed its $5b buyout of CBIZ, creating one of the country's largest accounting firms."

Anthropic

• Description: AI lab.
• Why mentioned: Q3 numbers disclosed in an S-1 amendment, a signal that an IPO is in progress.
• Quote: "Digging into Anthropic's Q3 numbers, which were disclosed yesterday in an S-1 amendment."

Cinven / Theramex

• Description: PE firm and women's-health pharma company.
• Why mentioned: €2.6B secondary sale from PAI and Carlyle.
• Quote: "Cinven agreed to buy Theramex, a London-based pharma company focused on women's health... for €2.6b."

LS Power / EVgo

• Description: Infrastructure investor and EV charging company.
• Why mentioned: Take-private offer.
• Quote: "LS Power offered to buy the stake it doesn't already own in EV charging company EVgo (Nasdaq: EVGO) for $2 per share."

Lonza / Merck KGaA; Shell / XRG

• Description: CDMO manufacturer and chemicals/pharma group; energy major and Abu Dhabi buyer.
• Why mentioned: Asset carve-outs and portfolio pruning.
• Quote: "Lonza agreed to acquire five CDMO manufacturing sites in the U.S. from Merck KGaA." / "Shell... agreed to sell its 16.7% stake in Australia's North West Shelf LNG project for US$1.45b to Abu Dhabi's XRG."

Brookfield / QIA / Drax

• Description: Infrastructure and sovereign investors with a U.K. power company.
• Why mentioned: Data center capital deployment.
• Quote: "agreed to invest £2.5b into a U.K. data center project with Drax Group."

4. People Identified

Dan Primack

• Description: Author of Axios Pro Rata.
• Why mentioned: Argues the prediction-market thesis.
• Quote: "Kalshi and Polymarket should be pulling hard for Democrats in today's midterm elections."

Carl Icahn

• Description: Activist investor who controls CVR Energy.
• Why mentioned: Principal behind the $4.2B refinery purchase.
• Quote: "CVR Energy (NYSE: CVI), controlled by Carl Icahn, agreed to buy two U.S. refineries."

Elad Gil

• Description: Investor.
• Why mentioned: Led Applied Compute's $350M round.
• Quote: "Elad Gil led, joined by Kleiner Perkins, Lux Capital, Hanabi Capital, Nvidia, Blackstone."

Collin Eaton (WSJ)

• Description: Reporter.
• Why mentioned: Source for the refinery-scarcity framing.
• Quote: "given the almost intractable constraints in global capacity."

Ibrahim Ajami

• Description: Mubadala's head of ventures.
• Why mentioned: Leaving to launch his own fund.
• Quote: "is stepping down as Mubadala's head of ventures to launch Noha, his own fund targeting $500m, per the FT."

Yvonne Bajela

• Description: Ex-Phoenix Court investor.
• Why mentioned: Leads a debut VC fund.
• Quote: "Evertrue Capital, a London-based VC firm led by Yvonne Bajela (ex-Phoenix Court), raised $17m for a debut fund."

Kevin Yamashita, Ryan Craig, Ryan Gibson, Michael Laser

• Description: Newly promoted PE/VC leaders.
• Why mentioned: Partner promotions at Bertram Capital and Eclipse.
• Quote: "Bertram Capital promoted Kevin Yamashita and Ryan Craig to managing partners. Eclipse promoted Ryan Gibson and Michael Laser to partners."

5. Operating Insights

Build credibility assets before you need them

Prediction markets show how fast a single accurate call confers legitimacy, and how a few misses can erode it. Founders in probabilistic or forecast-based products should manage expectations around probability versus certainty.

• "A Republican win in either chamber, however, would dump them into the same pot as pollsters — even if 75% probability includes a sizable improbability."

Make revenue independent of outcomes, but protect the trust that drives volume

The no-house model removes outcome risk from the P&L, but the strategic risk is reputational.

• "There is no 'house.' They get their cut either way."

Structure raises in tranches when valuation is uncertain

Ornn's triple-tranched Series A lets the company stage dilution and valuation as it hits milestones.

• "Ornn... is raising a triple-tranched $150m in Series A funding... The final tranche could be at a valuation north of $1b."

6. Overlooked Insights

Anthropic's S-1 amendment signals IPO preparation

A single italicized link reveals that Anthropic has an S-1 on file, with Q3 numbers disclosed via amendment, a notable marker for the AI IPO window.

• "Digging into Anthropic's Q3 numbers, which were disclosed yesterday in an S-1 amendment."

Sub-$20M AI security and observability exits are happening fast

Several small acquisitions suggest tuck-in M&A is an active exit path for early AI security and infrastructure startups.

• "Omen AI... acquired Aquatic Labs... for $16m in cash and stock. Aquatic Labs had raised $5m."
• "CrowdStrike... agreed to acquire Dreadnode... that had raised $14m."