Stray Dog
EN中文
EN中文

Prices are Hyperliquid perpetual futures: contracts that track each stock and trade 24/7. They are not exchange prices and can differ from them, most of all while US markets are closed. Change is over 24 hours. For information only.

  • TSLA—
  • PLTR—
  • HIMS—
  • SPCX—
  • SOFI—
  • META—
  • BTC—
…Loading prices
PLTR

Palantir reports Q3 tonight; Hyundai Heavy cites AIP gains

Sources (4)
  • Neutral: Q3 results come after the US close tonight; the company's own revenue guidance is $2.160 billion to $2.164 billion.
  • Good news: Hyundai Heavy says Palantir's AIP agent cut the average robot fault fix time from 21.5 days to 7.9 days, though these are the customer's own figures.
  • Neutral: CTO Shyam Sankar says customers that used AI to cut headcount mostly missed their goal, while those aiming to dominate their market did well (the seller's own account).

What we watchWhether net retention holds at 120%, and the net new customer count.

Read the full notes

Hyundai Heavy's robotics business says that since the Palantir AIP joint agent went live in August, the average time to resolve a robot fault has fallen from 21.5 days to 7.9 days (down 63%), and it expects robot programming costs to fall by more than 50%. These are the customer's own figures, and no contract value has been disclosed.

Read more

It extends industrial physical AI from data analysis to controlling real equipment workflows, and adds to a commercial customer base that is mostly American. In his Stratechery interview, CTO Shyam Sankar set the sales narrative: the unit of the enterprise is the workflow, not the knowledge worker. Almost every customer that wanted to use AI to cut headcount and costs fell short of its goal, while those that wanted to use AI to dominate their market did very well. Palantir still works with Anthropic and uses both frontier labs, and the company has put a great deal of work into letting customers switch models safely (this is the seller's own account, with no data behind it). Earnings come after the US close tonight (call at 6am Singapore time on 3 Nov). Palantir's own Q3 revenue guidance is $2.160 billion to $2.164 billion (Q2 was $1.935 billion, up 93% year on year). The points to watch are whether net retention holds at 120%, net new customers, the gap between revenue growth and headcount growth, GAAP EPS, stock-based compensation as a share of revenue, remaining performance obligations, the Rule of 40, and the wording on UK and European revenue and the tone on the NHS renewal.

Key evidence

  • Since Hyundai Heavy's robotics business put an AI agent co-developed with Palantir into use in August, the average time to resolve a robot fault has fallen from 21.5 days to 7.9 days (down 63%), and robot programming costs are expected to fall by more than 50% (The Korea Herald, 1 Nov)
  • How it works: the agent analyses existing robot programmes and proposes optimisations, and engineers review and change them (a human in the loop, not a replacement). Executives from both sides hold regular steering committee meetings, and chief revenue officer Ryan Taylor backed extending it to new business
  • Palantir's sovereign bootcamps began in July and focus on deploying AI without letting sensitive data leave the customer. Three sessions drew about 250 organisations and 450 executives
  • Sankar's Stratechery interview was published on 29 Oct and picked up by overseas media on 1 Nov: the basic unit of an enterprise is the workflow, not the knowledge worker. Making one employee faster brings no return; the return comes when a whole process measurably improves (coding is the exception)
  • Sankar's observation by type of customer: almost all of those that wanted to use AI to cut headcount and costs missed their goal, while those that wanted to use AI to dominate their market did very well. His reason is that even if AI does 98% of what a person does, the goal of removing that person still fails
  • Palantir still works with Anthropic and uses both frontier labs, and has put a great deal of work into letting customers switch models safely
Show 4 more
  • Palantir's own Q3 revenue guidance is $2.160 billion to $2.164 billion. Q2 was $1.935 billion, up 93% year on year
  • Earnings timing: after the US close on 2 Nov, with the call at 6am Singapore time on 3 Nov
  • On 28 Oct, grants of about 987,000 shares to three executives were booked as stock-based compensation
  • EDGAR: no new Form 4 since 30 Oct

Catalysts to watch

  • Q3 earnings after the US close tonight (call at 6am Singapore time on 3 Nov)
  • Whether net retention holds above 120%
  • Net new customers (and the conversion rate from the 250 organisations and 450 executives at the sovereign bootcamps)
  • The gap between revenue growth and headcount growth
  • GAAP earnings per share
Show 4 more
  • Stock-based compensation as a share of revenue
  • Remaining performance obligations (RPO)
  • The Rule of 40
  • Wording on UK and European revenue and the tone on the NHS renewal

Risks

  • The Hyundai Heavy AIP figures are self-reported by the customer, and no contract value has been disclosed
  • The sovereign bootcamps are a customer acquisition funnel, not revenue, and the conversion rate will have to be read from net new customers in the results
  • Sankar's workflow narrative is the seller's own account, with no data behind it
  • The Department of Defense has stopped using Anthropic, so Palantir's continued work with Anthropic on the commercial side could be used against it politically
  • The point that Cybercab-specific changes are not confirmed to be passed down to the Model Y does not apply to this stock

Sources 4

  1. The Korea Herald, 1 Nov (Hyundai Heavy's robotics business discloses AIP results)
  2. Stratechery, 29 Oct (long interview with Shyam Sankar)
  3. Palantir's own Q3 revenue guidance
  4. EDGAR Form 4 filings (grants to three executives, 28 Oct)

A brief is a lead for our research, not a verified conclusion or advice.