Palantir's AI pitch gets a 30,000-hour proof point

A Palantir client is assigning a number to it.

McCarthy Building Companies says an AI-assisted workflow built with Palantir Technologies (PLTR) and Nvidia (NVDA) technology is saving the construction company about 30,000 work hours annually.

Investors are paying attention to more than just that figure.

L3Harris Technologies (LHX) says an internally tested AI setup built with Palantir and Nvidia models cut costs by 95% in its test environment.

Another implementation, according to Tampa General Hospital, reduced latency 22 times while increasing model performance by 7%.

According to Palantir, the three customers are among several companies that are putting its artificial intelligence platform, Foundry, and Ontology software into real-world operations.

Those figures could be more important to Palantir investors than any other AI presentation.

In its most recent reported quarter, the company’s U.S. commercial sales increased 149% year over year.

Palantir must demonstrate that clients aren’t only experimenting with AI to maintain anything approaching that pace.

To continue growing it, they must get sufficient economic value.

Palantir’s 30,000-hour number makes AI easier to understand

It may seem less obvious to search for cutting-edge AI in the construction industry.

McCarthy’s case is unique because of its scale.

The 160-year-old construction business used over 100,000 production requests for information, or RFIs, to train Nvidia’s Nemotron models.

When plans, specifications, or responsibilities need clarification during a construction project, an RFI serves as a formal query.

If certain issues are not answered quickly, a project may be delayed or expenses may rise.

Before a request reaches a project site, McCarthy uses Palantir’s algorithms to determine whether requests can cause scheduling or cost issues.

Additionally, it developed a subcontractor agreement writing system with AI assistance.

According to Palantir, that workflow is saving McCarthy around 30,000 hours annually while performing at a lower cost than the general-purpose AI models McCarthy tested against.

Related: Palantir stock faces brutal test after crushing expectations

Compared with a benchmark score, that business case is much easier to understand.

Thirty thousand hours is about the same as more than a dozen full-time workers’ yearly labor hours.

McCarthy’s ability to kill those jobs is not necessarily the objective.

It means that thousands of hours that were previously devoted to detecting hazards and converting project information into contracts may be used for other purposes.

When determining if another AI contract is worthwhile, business purchasers might provide a chief financial officer with this kind of analysis.

Other Palantir customers are putting numbers on AI

McCarthy is not the only customer reporting measurable results.

Defense contractor L3Harris has connected more than 30 enterprise resource planning systems through Palantir, creating roughly 3.5 million data connections and more than 5,000 scheduled data feeds.

Instead of depending solely on general-purpose models, the business is now evaluating AI models trained on its own exclusive data.

L3Harris Chief Transformation Officer Heidi Wood said initial internal testing outperformed frontier models in the company’s test environment while cutting costs by 95%.

Those results are company-reported and have not been independently audited.

Another example is Tampa General Hospital.

Since 2021, the hospital has collaborated with Palantir and employs their software for tasks including managing patient flow, staffing nurses, and spotting sepsis symptoms.

Its most recent initiative makes use of an Nvidia model tailored to Tampa General’s specific risk and privacy needs.

Palantir says that by fine-tuning the model, performance increased by 7% and the model became 22 times faster.

Another new client has set similar ambitious objectives.

Battery materials manufacturer POSCO Future M signed an agreement on Oct. 21 to deploy Palantir’s Foundry and AIP platforms.

The firm wants to reduce product development time by 50%, enhance production efficiency by 30%, and improve customer response by over 60%.

Instead of actual outcomes, they are objectives.

Palantir will be used in sales and marketing at first, according to POSCO Future M, before the company decides whether to extend the technology into production.

That difference is crucial.

McCarthy reports an operational savings as of right now.

POSCO is outlining what it believes Palantir will ultimately be able to provide.

Investors should not treat the two figures the same way.

Palantir’s next growth argument comes down to one number

Finn Gomez / Getty Images

Palantir needs customer ROI to keep extraordinary growth going

Palantir is already growing at a rate that few major software firms can match.

Revenue for the second quarter was $1.94 billion, a 93% increase over the same period last year.

While U.S. government income jumped 90% to $809 million, commercial revenue increased 149% to $764 million.

In addition, Palantir completed 73 transactions during the quarter, each totaling at least $10 million.

Palantir reports that by the end of June, the company’s clientele had increased from 849 to 1,049.

These figures clarify the significance of customer outcomes.

Palantir does not necessarily need to explain its long-term growth plans in order to continue obtaining experimental contracts.

When an experiment turns into a deployment, the client spends more money spreading it across the company, and the deployment results in quantifiable savings, the outcome is stronger.

The significance of such growth is already evident in Palantir’s own filings.

In the second quarter, the average trailing-12-month revenue from its top 20 clients was $124 million, a 67% increase from the same period last year.

This implies that current clients are growing their accounts considerably.

A potential explanation for this increase may be found in the latest AI examples.

Purchasing additional Palantir software becomes more justifiable if a military contractor can significantly lower the cost of operating specialist AI models or if a construction business can save 30,000 hours of manual labor.

However, there is still a significant warning.

Palantir shared most of the latest efficiency statistics with its clients, but it did not provide sufficient financial information to determine the ROI for each deployment.

The highest productivity figures for POSCO are still objectives.

Furthermore, a single effective process does not guarantee that every Palantir client will get comparable savings.

That is the countercase investors should watch.

Palantir reports third-quarter results after the market closes on Nov. 2.

The focus will be primarily on revenue growth.

Beneath that could be the more significant long-term question: Are clients getting enough quantifiable value from Palantir’s AI to continue implementing more of it?

Palantir has a convincing response thanks to McCarthy’s 30,000-hour estimate.

It must now make such outcomes reproducible.

Related: Palantir stock faces brutal test after crushing expectations

#Palantir039s #pitch #30000hour #proof #point