Palantir just posted record AI earnings. Its clearest construction proof point is a top-20 GC running field ops on the platform.
Palantir's Q2 revenue grew 93% and the company raised full-year guidance, with U.S. commercial revenue up 149%. Months earlier, McCarthy Building Companies put its entire field-operations workflow on Palantir's AI platform — a real look at what a full 'AI operating system' costs, and what it locks in, next to a point tool.
Palantir reported record second-quarter results this week: revenue up 93% year-over-year to $1.94 billion, U.S. commercial revenue up 149% to $764 million, and raised full-year guidance to as much as $8.16 billion. The stock jumped as much as 29% on the news. The clearest construction-industry read on that growth isn't in the earnings call — it's in a deal Palantir signed months earlier with McCarthy Building Companies, a top-20 U.S. general contractor that's building its entire field-operations system on Palantir's AI platform.
What did Palantir actually report?
Palantir's Q2 2026 numbers, by segment: U.S. commercial revenue hit $764 million, up 149% year-over-year; U.S. government revenue reached $809 million, up 90%. Net income nearly tripled to $1.07 billion from $329 million a year earlier. The company closed 220 deals worth at least $1 million in the quarter — 98 of them worth $5 million or more, 73 worth $10 million or more — and raised its full-year revenue guidance from roughly $7.65 billion to as much as $8.16 billion. CEO Alex Karp told investors demand for what he calls "AI sovereignty" — customers wanting to own and control the data model behind their own decisions — had been "unleashed."
What is McCarthy actually running on Palantir?
McCarthy — founded in 1864, 100% employee-owned since 2002, and generating roughly $4.5 billion in annual revenue — signed a multi-year, multi-million-dollar partnership with Palantir in June 2026. The centerpiece is Pulse, McCarthy's AI Operations Suite, built on Palantir's AIP and modeled in Palantir's Ontology — a data layer meant to tie McCarthy's schedules, crews, equipment, and project history into one operational model instead of scattered spreadsheets and point systems. The pitch, per McCarthy senior superintendent Dave Evans: it's supposed to let field leadership "focus on what matters most for our clients and project teams, rather than getting buried in emails, documents and data." Pulse is aimed at giving superintendents and PMs real-time scenario planning and risk analysis without waiting on the office to compile a report.
Is this the same category as Procore or OpenSpace?
No, and the difference is the part worth understanding before any similar pitch lands on your desk.
| Point tool | Full operating platform | |
|---|---|---|
| Example | Procore, OpenSpace, Buildots | Palantir AIP / Ontology |
| What it needs from you | Connects to data you already have | You model your operations inside its proprietary data layer first |
| Time to value | Weeks | Months, with dedicated build resources |
| Who it fits | Any size GC or sub | Enterprise GCs with in-house data/IT teams |
| Switching cost | Low — export and move on | High — your operating model lives inside the vendor's structure |
A point tool reads your existing data and adds a capability on top. A platform like AIP asks you to build your company's operational model inside its structure before it does anything — which is a bigger lift, and a much stickier one.
Should a mid-size GC or sub act on this?
Not directly — McCarthy's deal is priced and staffed for a multi-billion-dollar contractor, and there's no disclosed version of it sized for a 50-person mechanical sub or a regional GC. But the signal is worth tracking: this is the largest, most concrete example yet of a big AI platform vendor treating a construction company's entire operating data — not one workflow — as the product. If that category keeps growing, expect more owners and large GCs to push similar "connected" platform requirements down onto their subs' reporting and data-sharing obligations.
The takeaway
Before you sign anything pitched as a "connected" or "unified data" AI platform — at any scale — ask one question up front: what does getting your data back out look like on day one if you leave in three years. A tool that reads your existing systems is a low-risk add. A platform that asks you to rebuild your operating model inside its own proprietary layer is a different kind of commitment, and McCarthy's deal is the clearest example on record of what that commitment actually looks like in construction.
We covered the same lock-in dynamic from the acquisition side in July — this is what it looks like when a contractor buys into it directly instead of inheriting it through a vendor's M&A.
Forward this to whoever at your firm is evaluating a "unified data" or "AI operating system" pitch from a software vendor.
- What did Palantir report in its Q2 2026 earnings?
- Revenue of $1.94 billion, up 93% year-over-year. U.S. commercial revenue rose 149% to $764 million, and the company raised full-year guidance to between $8.15 billion and $8.16 billion. Palantir closed 220 deals worth at least $1 million in the quarter, and its stock jumped as much as 29% on the news.
- What is McCarthy Building Companies doing with Palantir?
- McCarthy, a 160-year-old, employee-owned general contractor ranked among the top 20 U.S. builders by revenue, signed a multi-year, multi-million-dollar deal to build its field-operations system, called Pulse, on Palantir's AI platform (AIP). Pulse is meant to give superintendents and PMs real-time scenario planning, risk analysis, and decision support drawn from the company's own project data.
- Can a mid-size GC or a trade sub buy the same thing McCarthy has?
- Not in any practical sense. McCarthy's deal is multi-year and multi-million-dollar, sized for a contractor doing billions in annual revenue with in-house data and IT staff to build out the platform. There's no small-contractor tier disclosed. Most GCs and subs will keep using point tools — Procore, Autodesk Construction Cloud, OpenSpace, Buildots — rather than a full operating-system rebuild.
- What's the real difference between an AI operating system and a construction AI point tool?
- A point tool bolts a feature onto data you already have — it tags photos, drafts an RFI, flags a schedule slip. An operating-system platform like AIP requires you to first model your company's projects, crews, equipment, and workflows inside its own proprietary data layer before any of that reasoning works. That's a much bigger upfront build, and it's much harder to walk away from once your operations are expressed inside someone else's data model.
- What should a contractor ask before signing onto a 'connected' or 'unified data' AI platform?
- What getting your data back out looks like on day one, in a usable format, if you switch vendors in three years. A platform that models your entire operation in its own proprietary structure — rather than reading standard exports from your existing systems — creates real switching costs, and that's worth pricing into the decision before you sign, not after.